Abstract
We establish the existence of multi-bump solutions for the following class of quasilinear problems
where the nonlinearity \( f :\mathbb R^N \times \mathbb R \rightarrow \mathbb R \) is a continuous function having a subcritical growth and potentials \( V, Z :\mathbb R^N \rightarrow \mathbb R \) are continuous functions verifying some hypotheses. The main tool used is the variational method.
Similar content being viewed by others
1 Introduction
In this paper, we consider the existence and multiplicity of solutions for the following class of problems
where \( \Delta _{ p(x) } \) is the \( p(x) \)-Laplacian operator given by
Here, \( \lambda > 0 \) is a parameter, \( p :\mathbb R^N \rightarrow \mathbb R \) is a Lipschitz function, \( V,Z :\mathbb R^N \rightarrow \mathbb R \) are continuous functions with \( V \ge 0 \), and \( f :\mathbb R^N \times \mathbb R \rightarrow \mathbb R \) is continuous having a subcritical growth. Furthermore, we take into account the following set of hypotheses:
- (\(H_1\)):
-
\( 1 < p_- = \displaystyle \inf _{ \mathbb R^N} p \le p_+ = \sup _{\mathbb R^N} p< N \).
- (\(H_2\)):
-
\( \Omega = \text {int } V^{ -1 } (0) \ne \emptyset \) and bounded, \( \overline{\Omega } = V^{ -1 }(0) \) and \( \Omega \) can be decomposed in \( k \) connected components \( \Omega _1, \ldots , \Omega _k \) with \( \text {dist} \big ( \Omega _i, \Omega _j \big ) > 0, \, i \ne j \).
- (\(H_3\)):
-
There exists \( M > 0 \) such that
$$\begin{aligned} \lambda V(x) + Z(x) \ge M, \, \forall x \in \mathbb R^N, \lambda \ge 1. \end{aligned}$$ - (\(H_4\)):
-
There exists \( K > 0 \) such that
$$\begin{aligned} \big | Z(x) \big | \le K, \, \forall x \in \mathbb R^N. \end{aligned}$$ - (\(f_1\)):
-
$$\begin{aligned} \limsup _{ |t| \rightarrow \infty } \frac{|f(x,t)|}{|t|^{ q(x)-1 }} < \infty , \text { uniformly in } x \in \mathbb R^N, \end{aligned}$$
where \( q :\mathbb R^N \rightarrow \mathbb R \) is continuous with \( p_+ < q_- \) and \(q \ll p^*=\frac{Np}{N-p}\). Here, the notation \(q \ll p^*\) means that \( \displaystyle \inf _{ \mathbb R^N} (p^*-q) > 0\).
- (\(f_2\)):
-
\( f(x,t) = o \big ( |t|^{ p_+ - 1} \big ), \, t \rightarrow 0, \text { uniformly in } x \in \mathbb R^N \).
- (\(f_3\)):
-
There exists \( \theta > p_+ \) such that
$$\begin{aligned} 0 < \theta F(x,t) \le f(x,t) t, \, \forall x \in \mathbb R^N, t > 0, \end{aligned}$$where \( F(x,t) = \int _0^t f(x,s) \, \mathrm{{d}}s \).
- (\(f_4\)):
-
\( \dfrac{f(x,t)}{t^{p_+ - 1}} \) is strictly increasing in \( t \in (0,\infty ) \), for each \( x \in \mathbb R^N \).
- (\(f_5\)):
-
\( \forall a, b \in \mathbb R, \, a < b, \, \mathop {\mathop {\sup }\limits _{x \in \mathbb {R}^N}}\limits _{t \in [a,b]} |f(x,t)| < \infty \).
A typical example of nonlinearity verifying \( (f_1)-(f_5) \) is
where \( p_+ < q_- \) and \( q \ll p^*\).
Partial differential equations involving the \( p(x) \)-Laplacian arise, for instance, as a mathematical model for problems involving electrorheological fluids and image restorations, see [1, 2, 11–13, 29]. This explains the intense research on this subject in the last decades. A lot of works, mainly treating nonlinearities with subcritical growth, are available (see [4–9, 16–18, 20–24, 28] for interesting works). Nevertheless, to the best of the author’s knowledge, this is the first work dealing with multi-bump solutions for this class of problems.
The motivation to investigate problem \( \big ( P_\lambda \big ) \) in the setting of variable exponents has been the papers [3] and [15]. In [15], inspired by del Pino and Felmer [14] and Séré [30], the authors considered \( \big ( P_\lambda \big ) \) for \( p = 2 \) and \( f(u) = u^q , q \in \big (1, \frac{N+2}{N-2} \big ) \) if \( N \ge 3 \); \( q \in (1, \infty ) \) if \( N = 1, 2 \). The authors showed that \( \big ( P_\lambda \big ) \) has at least \( 2^k-1 \) solutions \(u_\lambda \) for large values of \( \lambda \). More precisely, one solution for each non-empty subset \( \Upsilon \) of \( \{ 1,\ldots ,k \} \). Moreover, fixed \( \Upsilon \subset \{ 1,\ldots ,k \}\), it was proved that, for any sequence \( \lambda _n \rightarrow \infty \), we can extract a subsequence \(( \lambda _{n_i}) \) such that \(( u_{ \lambda _{n_i} } )\) converges strongly in \( H^1 \big ( \mathbb R^N \big ) \) to a function \( u \), which satisfies \( u = 0 \) outside \( \Omega _\Upsilon = \bigcup _{ j \in \Upsilon } \Omega _j \) and \( u_{|_{\Omega _j}}, \, j \in \Upsilon \) is a least energy solution for
In [3], employing some different arguments than those used in [15], Alves extended the results described above to the \( p\)-Laplacian operator, assuming that in \( \big ( P_\lambda \big ) \) the nonlinearity \( f \) possesses a subcritical growth and \( 2 \le p < N \). In particular, fixed \( \Upsilon \subset \{ 1,\ldots ,k \}\), for any sequence \( \lambda _n \rightarrow \infty \), we can extract a subsequence \( (\lambda _{n_i}) \) such that \( (u_{ \lambda _{n_i} }) \) converges strongly in \( W^{ 1,p } \big ( \mathbb R^N \big ) \) to a function \( u \), which satisfies \( u = 0 \) outside \( \Omega _\Upsilon \) and \( u_{|_{\Omega _j}}, \, j \in \Upsilon \), is a least energy solution for
In the present paper, we extend the results found in [3] to the \( p(x)\)-Laplacian operator. However, we would like to emphasize that in a lot of estimates, we have used different arguments from that found in [3]. The main difference is related to the fact that for equations involving the \(p(x)\)-Laplacian operator it is not clear that Moser’s iteration method is a good tool to get the estimates for the \(L^{\infty }\)-norm. Here, we adapt some ideas explored in [18] and [25] to get these estimates. For more details see Sect. 5.
Since we intend to find nonnegative solutions, throughout this paper, we replace \( f \) by \( f^+ :\mathbb {R}^N \times \mathbb {R} \rightarrow \mathbb {R} \) given by
Nevertheless, for the sake of simplicity, we still write \( f \) instead of \( f^+ \).
The main theorem in this paper is the following:
Theorem 1.1
Assume that \( (H_1)-(H_4) \) and \( (f_1)-(f_5) \) hold. Then, there exist \( \lambda _0 > 0 \) with the following property: for any non-empty subset \( \Upsilon \) of \( \{1, 2, . . . , k \} \) and \( \lambda \ge \lambda _0 \), problem \( \big ( P_\lambda \big ) \) has a solution \(u_\lambda \). Moreover, if we fix the subset \( \Upsilon \), then for any sequence \( \lambda _n \rightarrow \infty \), we can extract a subsequence \( (\lambda _{n_i}) \) such that \( (u_{ \lambda _{n_i} }) \) converges strongly in \( W^{ 1,p(x) } \big ( \mathbb R^N \big ) \) to a function \( u \), which satisfies \( u = 0 \) outside \( \Omega _\Upsilon = \bigcup _{ j \in \Upsilon } \Omega _j \) and \( u_{|_{\Omega _j}}, \, j \in \Upsilon \), is a least energy solution for
Notations: The following notations will be used in the present work:
-
\(C\) and \(C_i\) will denote generic positive constant, which may vary from line to line;
-
In all the integrals, we omit the symbol \(dx\).
-
If \(u\) is a measurable function, we denote \(u^+\) and \(u^-\) its positive and negative part, i.e., \(u^+(x) = \max \{ u(x), 0 \}\) and \( u^-(x) = \min \{ u(x), 0 \}\).
-
If \( u,v \) are measurable functions, \( u_- = \text {ess} \displaystyle \inf _{ \! \! \! \! \! \mathbb R^N} u , u_+ = \text {ess} \displaystyle \sup _{ \! \! \! \! \! \mathbb R^N} u \) and the notation \( u \ll v \) means that \( \text {ess} \displaystyle \inf _{ \! \! \! \! \! \mathbb R^N} \left( v-u \right) > 0 \) . Moreover, we will denote by \(u^*\) the function
$$\begin{aligned} u^*(x) = {\left\{ \begin{array}{ll} \frac{Nu(x)}{N-u(x)},&{} \text { if } u(x) < N, \\ \infty , &{}\text { if } u(x) \ge N. \end{array}\right. } \end{aligned}$$
2 Preliminaries on variable exponents Lebesgue and Sobolev spaces
In this section, we recall some results on variable exponents Lebesgue and Sobolev spaces found in [8, 19, 21] and their references.
Let \( h \in L^\infty \big ( \mathbb R^N \big ) \) with \( h_- = \text {ess} \displaystyle \inf _{ \! \! \! \! \! \mathbb R^N} h \ge 1\). The variable exponent Lebesgue space \( L^{ h(x) } \big ( \mathbb R^N \big ) \) is defined by
endowed with the norm
The variable exponent Sobolev space is defined by
with the norm
If \( h_- > 1 \), the spaces \( L^{ h(x) } \big ( \mathbb R^N \big ) \) and \( W^{ 1,h(x) } \big ( \mathbb R^N \big ) \) are separable and reflexive with these norms.
We are mainly interested in subspaces of \( W^{ 1,h(x) } \big ( \mathbb R^N \big ) \) given by
where \( W \in C \big ( \mathbb R^N \big ) \) is such that \( W_- > 0 \). Endowing \( E_W \) with the norm
\( E_W \) is a Banach space. Moreover, it is easy to see that \( E_W \hookrightarrow W^{ 1,h(x) } \big ( \mathbb R^N \big ) \) continuously. In addition, we can show that \( E_W \) is reflexive. For the reader’s convenience, we recall some basic results.
Proposition 2.1
The functional \( \varrho :E_W \rightarrow \mathbb R \) defined by
has the following properties:
-
(i)
If \( \left\| u \right\| _W \ge 1 \), then \( \left\| u \right\| _W^{ h_- } \le \varrho (u) \le \left\| u\right\| _W^{ h_+ } \).
-
(ii)
If \( \left\| u \right\| _W \le 1\), then \( \left\| u \right\| _W^{ h_+ }\le \varrho (u) \le \left\| u \right\| _W^{h_- }\).
In particular, for a sequence \( (u_n) \) in \( E_W \),
Remark 2.2
For the functional \( \varrho _{ h(x) } :L^{ h(x) } \big ( \mathbb R^N \big ) \rightarrow \mathbb R \) given by
an analogous conclusion to that of Proposition 2.1 also holds.
Proposition 2.3
Let \( m \in L^\infty \big ( \mathbb R^N \big ) \) with \( 0 < m_- \le m(x) \le h(x) \text { for a.e. } x \in \mathbb R^N \). If \( u \in L^{ h(x) } \big ( \mathbb R^N \big ) \), then \( |u|^{ m(x) } \in L^{ \frac{h(x)}{m(x)} } \big ( \mathbb R^N \big ) \) and
Related to the Lebesgue space \( L^{ h(x) } \big ( \mathbb R^N \big ) \), we have the following generalized Hölder’s inequality.
Proposition 2.4 (Hölder’s inequality)
[Hölder’s inequality] If \( h_- > 1 \), let \( h' :\mathbb R^N \rightarrow \mathbb R \) such that
Then, for any \( u\in L^{ h( x) } \big ( \mathbb R^N \big ) \) and \( v \in L^{ h'(x) } \big ( \mathbb R^N \big ) \),
We can define variable exponent Lebesgue spaces with vector values. We say \( u = ( u_1, \ldots , u_L ) :\mathbb R^N \rightarrow \mathbb R^L \in L^{ h(x) } \big ( \mathbb R^N, \mathbb R^L \big ) \) if, and only if, \( u_i \in L^{ h(x) } \big ( \mathbb R^N \big ) \), for \( i = 1, \ldots , L \). On \( L^{ h(x) } \big ( \mathbb R^N, \mathbb R^L \big ) \), we consider the norm \( | u |_{ L^{ h(x) }( \mathbb R^N, \mathbb R^L) } = \sum _{i=1}^L | u_i |_{ h(x) } \).
We state below lemmas of Brezis–Lieb type. The proof of the two first results follows the same arguments explored at [26], while the proof of the latter can be found at [8].
Proposition 2.5 (Brezis–Lieb lemma, first version)
[Brezis–Lieb lemma, first version] Let \( \left( u_n \right) \) be a bounded sequence in \( L^{ h(x) } \big ( \mathbb R^N, \mathbb R^L \big ) \) such that \( u_n(x) \rightarrow u(x) \text { for a.e. } x \in \mathbb R^N \). Then, \( u \in L^{ h(x) } \big ( \mathbb R^N, \mathbb R^L \big ) \) and
Proposition 2.6 (Brezis–Lieb lemma, second version)
[Brezis–Lieb lemma, second version] Let \( \left( u_n \right) \) be a bounded sequence in \( L^{ h(x) } \big ( \mathbb R^N, \mathbb R^L \big ) \) with \( h_- > 1 \) and \( u_n(x) \rightarrow u(x) \text { for a.e. } x \in \mathbb R^N \). Then
Proposition 2.7 (Brezis–Lieb lemma, third version)
[Brezis–Lieb lemma, third version] Let \( (u_n) \) be a bounded sequence in \( L^{ h(x) } \big ( \mathbb R^N, \mathbb R^L \big ) \) with \( h_- > 1 \) and \( u_n(x) \rightarrow u(x) \) for a.e. \( x \in \mathbb R^N \). Then
To finish this section, we notice that for any open subset \( \Omega \subset \mathbb R^N \), we can define in the same way the spaces \( L^{ h(x) } \big ( \Omega \big ) \) and \( W^{ 1,h(x) } \big ( \Omega \big ) \). Moreover, all the above propositions have analogous versions for these spaces and, besides, we have the following embedding Theorem of Sobolev’s type.
Proposition 2.8
([21, Theorems 1.1, 1.3]) Let \( \Omega \subset \mathbb R^N \) an open domain with the cone property, \( h :\overline{\Omega } \rightarrow \mathbb R \) satisfying \( 1 < h_- \le h_+ < N \) and \( m \in L^{\infty }_+ \big ( \Omega \big ) \).
-
(i)
If \( h \) is Lipschitz continuous and \( h \le m \le h^{*} \), the embedding \( W^{ 1,h(x) } \big ( \Omega \big ) \hookrightarrow L^{ m(x) } \big ( \Omega \big ) \) is continuous;
-
(ii)
If \( \Omega \) is bounded, \( h \) is continuous and \( m \ll h^{*} \), the embedding \( W^{ 1,h(x) } \big ( \Omega \big ) \hookrightarrow L^{ m(x) } \big ( \Omega \big ) \) is compact.
3 An auxiliary problem
In this section, we work with an auxiliary problem adapting the ideas explored in del Pino and Felmer [14] (see also [3]).
We start noting that the energy functional \( I_\lambda :E_\lambda \rightarrow \mathbb R \) associated with \( \big ( P_\lambda \big ) \) is given by
where \( E_\lambda = \big ( E, \Vert \cdot \Vert _\lambda \big ) \) with
and
being
Thus, \( E_\lambda \hookrightarrow W^{ 1,p(x) } \big ( \mathbb R^N \big ) \) continuously for \( \lambda \ge 1 \) and \( E_\lambda \) is compactly embedded in \( L_{ loc }^{ h(x) } \big ( \mathbb R^N \big ) \), for all \( 1 \le h \ll p^* \). In addition, we can show that \( E_\lambda \) is a reflexive space. Also, being \( \mathcal{O} \subset \mathbb R^N \) an open set, from the relation
for all \( u \in E_\lambda \) with \( \lambda \ge 1 \), writing \( M = ( 1-\delta )^{ -1 } \nu \), for some \( 0 < \delta < 1 \) and \( \nu > 0\), we derive
Remark 3.1
From the above commentaries, in this work the parameter \(\lambda \) will be always bigger than or equal to 1.
We recall that for any \( \epsilon > 0 \), the hypotheses \( (f_1), (f_2) \) and \( (f_5) \) yield
and, consequently,
where \( C_\epsilon \) depends on \( \epsilon \). Moreover, for each \(\nu >0\) fixed, the assumptions \( (f_2) \) and \( (f_3) \) allow us considering the function \( a :\mathbb R^N \rightarrow \mathbb R\) given by
From \((f_2)\), it follows that
Using the function \(a(x)\), we set the function \( \tilde{f} :\mathbb R^N \times \mathbb R \rightarrow \mathbb R \) given by
which fulfills the inequality
Thus
and
where \( \tilde{F}(x,t) = \int _0^t \tilde{f}(x,s) \, \mathrm{{d}}s \).
Now, once that \( \Omega =\text {int } V^{ -1 } (0) \) is formed by \( k \) connected components \( \Omega _1, \ldots , \Omega _k \) with \( \text {dist} \big ( \Omega _i, \Omega _j \big ) > 0, \, i \ne j \), then for each \( j \in \{ 1, \ldots , k \} \), we are able to fix a smooth bounded domain \( \Omega '_j \) such that
From now on, we fix a non-empty subset \( \Upsilon \subset \left\{ 1, \ldots , k \right\} \) and
Using the above notations, we set the functions
and
and the auxiliary problem
The problem \( \big ( A_\lambda \big ) \) is related to \( \big ( P_\lambda \big ) \) in the sense that, if \( u_\lambda \) is a solution for \( \big ( A_\lambda \big ) \) verifying
then it is a solution for \( \big ( P_\lambda \big ) \).
In comparison with \( \big ( P_\lambda \big ) \), problem \( \big ( A_\lambda \big ) \) has the advantage that the energy functional associated with \( \big ( A_\lambda \big ) \), namely, \( \phi _\lambda :E_\lambda \rightarrow \mathbb R \) given by
satisfies the \( (PS) \) condition, whereas \( I_\lambda \) does not necessarily satisfy this condition. In this way, the mountain pass level (see Theorem 3.6) is a critical value for \( \phi _\lambda \).
Proposition 3.2
\( \phi _\lambda \) satisfies the mountain pass geometry.
Proof
for \( \epsilon > 0 \) and \( C_\epsilon > 0 \) be a constant depending on \( \epsilon \). By (3.1), fixing \( \epsilon < \frac{M}{p_+} \) and \( \nu < p_- M \left( \frac{1}{p_+}-\frac{\epsilon }{M} \right) \) and assuming \( \Vert u \Vert _\lambda < \min \left\{ 1, 1/C_q \right\} \), where \( | v |_{ q(x) } \le C_q \Vert v \Vert _\lambda , \, \forall v \in E_\lambda \), we derive from Proposition 2.1
where \( \alpha = \left( \frac{1}{p_+} - \frac{\epsilon }{M} \right) - \frac{\nu }{p_-M} > 0 \). Once \( p_+ < q_- \), the first part of the mountain pass geometry is satisfied. Now, fixing \(v \in C^{\infty }_{0}(\Omega _\Upsilon )\), we have for \(t \ge 0\)
If \(t>1\), by \((f_3)\),
and so,
The last limit implies that \(\phi _{\lambda }\) verifies the second geometry of the mountain pass. \(\square \)
Proposition 3.3
All \( (PS)_d \) sequences for \( \phi _\lambda \) are bounded in \( E_\lambda \).
Proof
Let \( (u_n) \) be a \( (PS)_d \) sequence for \( \phi _\lambda \). So, there is \( n_0 \in \mathbb N \) such that
On the other hand, by (3.8) and (3.9)
which together with (3.2) gives
Hence
from where it follows that \( (u_n) \) is bounded in \( E_\lambda \).\(\square \)
Proposition 3.4
If \((u_n)\) is a \((PS)_d\) sequence for \(\phi _{\lambda }\), then given \(\epsilon >0\), there is \(R>0\) such that
Hence, once that \(g\) has a subcritical growth, if \( u \in E_\lambda \) is the weak limit of \( (u_n) \), then
Proof
Let \( (u_n) \) be a \( (PS)_d \) sequence for \( \phi _\lambda , R > 0 \) large such that \( \Omega '_\Upsilon \subset B_{ \frac{R}{2} }(0) \) and \( \eta _R \in C^\infty \big ( \mathbb {R}^N \big ) \) satisfying
\( 0 \le \eta _R \le 1 \) and \( \big | \nabla \eta _R \big | \le \dfrac{C}{R} \), where \( C > 0 \) does not depend on \( R \). This way,
Denoting
it follows from (3.8),
Using Hölder’s inequality 2.4 and Proposition 2.3, we derive
Since \( (u_n) \) and \( \Big ( \big | \nabla u_n \big | \Big ) \) are bounded in \( L^{ p(x) } \big ( \mathbb {R}^N \big ) \) and \(\frac{\nu }{M}=1-\delta \), we obtain
Therefore
So, given \( \epsilon > 0 \), choosing a \( R > 0 \) possibly still bigger, we have that \( \dfrac{C}{R} < \epsilon \), which proves (3.11). Now, we will show that
Using the fact that \(g(x,u)u \in L^{1}(\mathbb {R}^N)\) together with (3.11) and Sobolev embeddings, given \(\epsilon >0\), we can choose \(R>0\) such that
On the other hand, since \(g\) has a subcritical growth, we have by compact embeddings
Combining the above information, we conclude that
The same type of arguments works to prove that
\(\square \)
Proposition 3.5
\( \phi _\lambda \) verifies the \( (PS) \) condition.
Proof
Let \( (u_n) \) be a \( (PS)_d \) sequence for \( \phi _\lambda \) and \( u \in E_\lambda \) such that \(u_n \rightharpoonup u\) in \(E_{\lambda }\). Thereby, by Proposition 3.4,
Moreover, the weak limit also gives
and
Now, if
and
we derive
Recalling that \(\phi _\lambda '(u_n)u_n=o_n(1)\) and \(\phi _\lambda '(u_n)u=o_n(1)\), the above limits lead to
Now, the conclusion follows as in [8].\(\square \)
Theorem 3.6
The problem \( \big ( A_\lambda \big ) \) has a (nonnegative) solution, for all \( \lambda \ge 1 \).
Proof
The proof is an immediate consequence of the Mountain Pass Theorem due to Ambrosetti and Rabinowitz [10].\(\square \)
4 The \( (PS)_\infty \) condition
A sequence \( (u_n) \subset W^{ 1,p(x) } \big ( \mathbb R^N \big ) \) is called a \( (PS)_\infty \) sequence for the family \( \left( \phi _\lambda \right) _{\lambda \ge 1} \), if there is a sequence \( ( \lambda _n ) \subset [1, \infty ) \) with \( \lambda _n \rightarrow \infty \), as \( n \rightarrow \infty \), verifying
Proposition 4.1
Let \( (u_n) \subset W^{ 1,p(x) } \big ( \mathbb R^N \big ) \) be a \( (PS)_\infty \) sequence for \( \left( \phi _\lambda \right) _{\lambda \ge 1} \). Then, up to a subsequence, there exists \( u \in W^{ 1,p(x) } \big ( \mathbb R^N \big ) \) such that \( u_n \rightharpoonup u \) in \( W^{ 1,p(x) } \big ( \mathbb R^N \big ) \). Furthermore,
-
(i)
\( \varrho _{ \lambda _n } (u_n-u) \rightarrow 0 \) and, consequently, \( u_n \rightarrow u \) in \( W^{ 1,p(x) } \big ( \mathbb R^N \big ) \);
-
(ii)
\( u = 0 \) in \( \mathbb R^N \setminus \Omega _\Upsilon , u \ge 0 \) and \( u_{|_{\Omega _j}}, \, j \in \Upsilon \), is a solution for
$$\begin{aligned} (P_j) \; {\left\{ \begin{array}{ll} - \Delta _{ p(x) } u + Z(x) | u |^{ p(x)-2 } u = f(x,u), \text { in } \Omega _j, \\ u \in W^{ 1,p(x) }_0 \big ( \Omega _j \big ); \end{array}\right. } \end{aligned}$$ -
(iii)
\( \displaystyle \int _{\mathbb R^N} \lambda _n V(x) | u_n |^{ p(x) } \rightarrow 0 \);
-
(iv)
\( \varrho _{ \lambda _n, \Omega '_j } (u_n) \rightarrow \displaystyle \int _{ \Omega _j } \left( \big | \nabla u \big |^{ p(x) } + Z(x) | u |^{ p(x) } \right) , \text { for } j \in \Upsilon \);
-
(v)
\( \varrho _{ \lambda _n, \mathbb R^N \setminus \Omega _\Upsilon } (u_n) \rightarrow 0 \);
-
(vi)
\( \phi _{ \lambda _n } (u_n) \rightarrow \displaystyle \int _{ \Omega _\Upsilon } \frac{1}{p(x)} \left( \big | \nabla u \big |^{ p(x) } + Z(x) | u |^{ p(x) } \right) - \int _{ \Omega _\Upsilon } F(x,u) \).
Proof
Using the same reasoning as in the proof of Proposition 3.3, we obtain that \( \big ( \varrho _{ \lambda _n }(u_n) \big ) \) is bounded in \( \mathbb R \). Then \( \big ( \Vert u_n \Vert _{ \lambda _n } \big ) \) is bounded in \( \mathbb R \) and \( (u_n) \) is bounded in \( W^{ 1,p(x) } \big ( \mathbb R^N \big ) \). So, up to a subsequence, there exists \( u \in W^{ 1,p(x) } \big ( \mathbb R^N \big ) \) such that
Now, for each \( m \in \mathbb N \), we define \( C_m = \left\{ x \in \mathbb R^N \, ; \, V(x) \ge \dfrac{1}{m} \right\} \). Without loss of generality, we can assume \( \lambda _n < 2 ( \lambda _n-1 ), \, \forall n \in \mathbb N \). Thus
By Fatou’s lemma, we derive
which implies that \( u = 0 \) in \( C_m \) and, consequently, \( u = 0 \) in \( \mathbb R^N \setminus \overline{\Omega } \). From this, we are able to prove \((i)-(vi)\).
- \((i)\) :
-
Since \( u = 0 \) in \( \mathbb R^N \setminus \overline{\Omega } \), repeating the argument explored in Proposition 3.5 we get
$$\begin{aligned} \int \limits _{ \mathbb R^N } \Big ( P_n^1(x) + \big ( \lambda _n V(x) + Z(x) \big ) P_n^2(x) \Big ) \rightarrow 0, \end{aligned}$$where
$$\begin{aligned} P_n^1 (x) = \left( \big | \nabla u_n \big |^{ p(x)-2 } \nabla u_n - \big | \nabla u \big |^{ p(x)-2 } \nabla u \right) \cdot \left( \nabla u_n - \nabla u \right) \end{aligned}$$and
$$\begin{aligned} P_n^2 (x) = \left( | u_n|^{ p(x)-2 } u_n - | u |^{ p(x)-2 } u \right) ( u_n - u ). \end{aligned}$$Therefore, \( \varrho _{ \lambda _n } ( u_n-u ) \rightarrow 0 \), which implies \( u_n \rightarrow u \) in \( W^{ 1,p(x) } \big ( \mathbb R^N \big ) \).
- \((ii)\) :
-
Since \( u \in W^{ 1,p(x) } \big ( \mathbb R^N \big ) \) and \( u = 0 \) in \( \mathbb R^N \setminus \overline{\Omega } \), we have \( u \in W^{ 1,p(x) }_0 \big ( \Omega \big ) \) or, equivalently, \( u_{ |_{\Omega _j} } \in W^{ 1,p(x) }_0 \big ( \Omega _j \big ) \), for \( j = 1, \ldots , k \). Moreover, the limit \(u_n \rightarrow u\) in \(W^{1,p(x)}(\mathbb R^N)\) combined with \(\phi '_{\lambda _n}(u_n)\varphi \rightarrow 0\) for \(\varphi \in C^{\infty }_0 \big ( \Omega _j \big )\) implies that
$$\begin{aligned} \int \limits _{\Omega _j} \left( \big | \nabla u \big |^{ p(x)-2 } \nabla u \cdot \nabla \varphi + Z(x) | u |^{ p(x)-2 } u \varphi \right) - \int \limits _{\Omega _j} g(x,u) \varphi = 0, \end{aligned}$$(4.1)showing that \( u_{ |_{\Omega _j} } \) is a solution for
$$\begin{aligned} {\left\{ \begin{array}{ll} - \Delta _{ p(x) } u + Z(x) | u |^{ p(x)-2 } u = g(x,u), \text { in } \Omega _j, \\ u \in W^{ 1,p(x) }_0 \big ( \Omega _j \big ). \end{array}\right. } \end{aligned}$$This way, if \( j \in \Upsilon \), then \( u_{ |_{\Omega _j} } \) satisfies \( (P_j) \). On the other hand, if \( j \notin \Upsilon \), we must have
$$\begin{aligned} \int \limits _{ \Omega _j } \left( \big | \nabla u \big |^{ p(x) } + Z(x) | u |^{ p(x) } \right) - \int \limits _{ \Omega _j } \tilde{f}(x,u)u = 0. \end{aligned}$$The above equality combined with (3.8) and (3.2) gives
$$\begin{aligned} 0 \ge \varrho _{ \lambda , \Omega _j }(u) - \nu \varrho _{ p(x), \Omega _j }(u) \ge \delta \varrho _{ \lambda , \Omega _j }(u) \ge 0, \end{aligned}$$from where it follows \( u_{|_{\Omega _j}} = 0 \). This proves \( u = 0 \) outside \( \Omega _\Upsilon \) and \( u \ge 0 \) in \( \mathbb {R}^N \).
- \((iii)\) :
-
It follows from (i), since
$$\begin{aligned} \int \limits _{ \mathbb R^N } \lambda _n V(x) | u_n |^{ p(x) } = \int \limits _{ \mathbb R^N } \lambda _n V(x) | u_n-u |^{ p(x) } \le 2 \varrho _{ \lambda _n }(u_n-u). \end{aligned}$$ - \((iv)\) :
-
Let \( j \in \Upsilon \). From (i),
$$\begin{aligned} \varrho _{ p(x), \Omega '_j }( u_n-u ), \varrho _{ p(x), \Omega '_j } \big ( \nabla u_n - \nabla u \big ) \rightarrow 0. \end{aligned}$$Then by Proposition 2.5,
$$\begin{aligned} \int \limits _{ \Omega '_j } \big ( \big | \nabla u_n \big |^{ p(x) } - \big | \nabla u \big |^{ p(x) } \big ) \rightarrow 0 \quad \text{ and } \quad \int \limits _{ \Omega '_j } Z(x) \big ( | u_n |^{ p(x) } - | u |^{ p(x) } \big ) \rightarrow 0. \end{aligned}$$From (iii),
$$\begin{aligned} \int \limits _{ \Omega '_j } \lambda _n V(x) \big ( | u_n |^{ p(x) } - | u |^{ p(x) } \big ) = \int \limits _{ \Omega '_j \setminus \overline{\Omega _j} } \lambda _n V(x) | u_n |^{ p(x) } \rightarrow 0. \end{aligned}$$This way
$$\begin{aligned} \varrho _{ \lambda _n, \Omega '_j } (u_n) - \varrho _{ \lambda _n, \Omega '_j } (u) \rightarrow 0. \end{aligned}$$Once \( u = 0 \text { in } \Omega '_j \setminus \Omega _j \), we get
$$\begin{aligned} \varrho _{ \lambda _n, \Omega '_j } (u_n) \rightarrow \int \limits _{ \Omega _j } \left( | \nabla u |^{ p(x) } + Z(x) | u |^{ p(x) } \right) . \end{aligned}$$ - \((v)\) :
-
By (i), \( \varrho _{ \lambda _n }( u_n-u ) \rightarrow 0 \), and so,
$$\begin{aligned} \varrho _{ \lambda _n, \mathbb R^N \setminus \Omega _\Upsilon }(u_n) \rightarrow 0. \end{aligned}$$ - \((vi)\) :
-
We can write the functional \(\phi _{\lambda _n}\) in the following way
$$\begin{aligned} \phi _{ \lambda _n } (u_n)&= \sum _{ j \in \Upsilon } \int \limits _{ \Omega '_j } \frac{1}{p(x)} \left( \big | \nabla u_n \big |^{ p(x) } + \big ( \lambda _n V(x) + Z(x) \big ) | u_n |^{ p(x) } \right) \\&+ \int \limits _{ \mathbb R^N \setminus \Omega '_\Upsilon } \frac{1}{p(x)} \left( \big | \nabla u_n \big |^{ p(x) } + \big ( \lambda _n V(x) + Z(x) \big ) | u_n |^{ p(x) } \right) - \int \limits _{ \mathbb R^N } G(x,u_n). \end{aligned}$$From \((i)-(v)\),
$$\begin{aligned}&\int \limits _{ \Omega '_j } \frac{1}{p(x)} \left( \big | \nabla u_n \big |^{ p(x) } + \big ( \lambda _n V(x) + Z(x) \big ) | u_n |^{ p(x) } \right) \\&\rightarrow \int \limits _{ \Omega _j } \frac{1}{p(x)} \left( \big | \nabla u \big |^{ p(x) } + Z(x) | u |^{ p(x) } \right) ,\\&\int \limits _{ \mathbb R^N \setminus \Omega '_\Upsilon } \frac{1}{p(x)} \left( \big | \nabla u_n \big |^{ p(x) } + \big ( \lambda _n V(x) + Z(x) \big ) | u_n |^{ p(x) } \right) \rightarrow 0. \end{aligned}$$and
$$\begin{aligned} \int \limits _{ \mathbb R^N } G(x,u_n) \rightarrow \int \limits _{ \Omega _\Upsilon } F(x,u). \end{aligned}$$Therefore
$$\begin{aligned} \phi _{ \lambda _n } (u_n) \rightarrow \int \limits _{ \Omega _\Upsilon } \frac{1}{p(x)} \left( | \nabla u |^{ p(x) } + Z(x) | u |^{ p(x) } \right) - \int \limits _{ \Omega _\Upsilon } F(x,u). \end{aligned}$$
\(\square \)
5 The boundedness of the \( \big ( A_\lambda \big ) \) solutions
In this section, we study the boundedness outside \( \Omega '_\Upsilon \) for some solutions of \( \big ( A_\lambda \big ) \). To this end, we adapt for our problem arguments found in [18] and [25].
Proposition 5.1
Let \( \big ( u_\lambda \big ) \) be a family of solutions for \( \big ( A_\lambda \big ) \) such that \( u_\lambda \rightarrow 0 \) in \( W^{ 1,p(x) } \big ( \mathbb R^N \setminus \Omega _\Upsilon \big ) \), as \( \lambda \rightarrow \infty \). Then, there exists \( \lambda ^* > 0 \) with the following property:
Hence, \(u_{\lambda }\) is a solution for \((P_\lambda )\) for \(\lambda \ge \lambda ^*\).
Before to prove the above proposition, we need to show some technical lemmas.
Lemma 5.2
There exist \( x_1, \ldots , x_l \in \partial \Omega '_\Upsilon \) and corresponding \( \delta _{x_1}, \ldots , \delta _{x_l} > 0 \) such that
Moreover,
where
Proof
From (3.10), \( \overline{\Omega _\Upsilon } \subset \Omega '_\Upsilon \). So, there is \( \delta > 0 \) such that
Once \( q \ll p^* \), there exists \( \epsilon > 0 \) such that \( \epsilon \le p^*(y) - q(y) \), for all \( y \in \mathbb R^N \). Then, by continuity, for each \( x \in \partial \Omega '_\Upsilon \), we can choose a sufficiently small \( 0 < \delta _x \le \delta \) such that
where
Covering \( \partial \Omega '_\Upsilon \) by the balls \( B_{ \frac{\delta _x}{2} }(x), \, x \in \partial \Omega '_\Upsilon \), and using its compactness, there are \( x_1, \ldots , x_l \in \partial \Omega '_\Upsilon \) such that
\(\square \)
Lemma 5.3
If \( u_\lambda \) is a solution for \( \big ( A_\lambda \big ) \), in each \( B_{ \delta _{x_i} }(x_i), \, i = 1, \ldots , l \), given by Lemma 5.2, it is fulfilled
where \( 0 < \overline{\delta } < \widetilde{\delta } < \delta _{ x_i } , k \ge \dfrac{a_-}{4} , C = C \big ( p_-, p_+, q_-, q_+, \nu , \delta _{ x_i } \big ) > 0 \) is a constant independent of \( k \), and for any \( R > 0 \), we denote by \(A_ { k,R,x_i }\) the set
Proof
We choose arbitrarily \( 0 < \overline{\delta } < \widetilde{\delta } < \delta _{x_i} \) and \( \xi \in C^{ \infty } \big ( \mathbb R^N \big ) \) with
For \( k \ge \dfrac{a_-}{4} \), we define \( \eta = \xi ^{ p_+ } ( u_\lambda -k )^+ \). We notice that
on the set \( \left\{ u_\lambda > k \right\} \). Then, writing \( u_\lambda = u \) and taking \( \eta \) as a test function, we obtain
If we set
using that \( \nu \le \lambda V(x) + Z(x), \, \forall x \in \mathbb R^N \), we get
from where it follows
Using Young’s inequality, we obtain, for \( \chi \in (0,1) \),
Writing
for \( \chi \approx 0^+ \) fixed, due to (5.1), we deduce
Therefore
for a positive constant \( C = C \big ( p_-, p_+, q_-, q_+, \nu , \delta _{ x_i } \big ) \) which does not depend on \( k \). Since
we obtain
for a positive constant \( C = C \big ( p_-, p_+, q_-, q_+, \nu , \delta _{ x_i } \big ) \) which does not depend on \( k \).\(\square \)
The next lemma can be found at ([27, Lemma 4.7]).
Lemma 5.4
Let \( (J_n) \) be a sequence of nonnegative numbers satisfying
where \( C, \eta > 0 \) and \( B > 1 \). If
then \( J_n \rightarrow 0 \), as \( n \rightarrow \infty \).
Lemma 5.5
Let \( \big ( u_\lambda \big ) \) be a family of solutions for \( \big ( A_\lambda \big ) \) such that \( u_\lambda \rightarrow 0 \) in \( W^{ 1,p(x) } \big ( \mathbb R^N \setminus \Omega _\Upsilon \big ) \), as \( \lambda \rightarrow \infty \). Then, there exists \( \lambda ^* > 0 \) with the following property:
Proof
It is enough to prove the inequality in each ball \( B_{\frac{\delta _{x_i}}{2}} (x_i), \, i = 1, \ldots , l \), given by Lemma 5.2. We set
Then
From now on, we fix
and \( \xi \in C^1 \big ( \mathbb R \big ) \) such that
Setting
we have \( \xi _n = 1 \) in \( B_{ \widetilde{\delta }_{n+1} }(x_i) \) and \( \xi _n = 0 \) outside \( B_{ \overline{\delta }_n }(x_i) \). Writing \( u_\lambda = u \), we get
Since
writing \( J_{ n+1 }^{ \frac{p^{x_i}_-}{\left( p^{x_i}_- \right) ^*} } = \widetilde{J}_{ n+1 } \), we obtain
Using Lemma 5.3,
From Young’s inequality
Thus
Now, since
it follows that
and so,
Fixing \( \alpha = \big ( p^{ x_i }_- + \left( p^{x_i}_- \right) ^* \big ) \), it follows that
and consequently
where \( C = C \Big ( N, p^{x_i}_-, \delta _{ x_i }, a_-, q_+ \Big ) , B = 2^{ \alpha \frac{\left( p^{x_i}_- \right) ^*}{p^{ x_i }_-} } \) and \( \eta = \frac{\left( p^{x_i}_- \right) ^*}{p^{ x_i }_-} -1 \). Now, once that \( u_\lambda \rightarrow 0 \) in \( W^{ 1,p(x) } \big ( \mathbb R^N \setminus \Omega _\Upsilon \big ) \), as \( \lambda \rightarrow \infty \), there exists \( \lambda _i > 0 \) such that
From Lemma 5.4, \( J_n(\lambda ) \rightarrow 0 , n \rightarrow \infty \), for all \( \lambda \ge \lambda _i\), and so,
Now, taking \( \lambda ^* = \max \{ \lambda _1, \ldots , \lambda _l \} \), we conclude that
\(\square \)
Proof of Proposition 5.1
Fix \( \lambda \ge \lambda ^* \), where \( \lambda ^* \) is given at Lemma 5.5, and define \( \widetilde{u}_\lambda :\mathbb R^N \setminus \Omega '_\Upsilon \rightarrow \mathbb R \) given by
From Lemma 5.5, \(\widetilde{u}_\lambda \in W^{ 1,p(x) }_0 \big ( \mathbb R^N \setminus \Omega '_\Upsilon \big ) \). Our goal is showing that \(\widetilde{u}_\lambda = 0 \) in \( \mathbb R^N \setminus \Omega '_\Upsilon \). This implies
In fact, extending \( \widetilde{u}_\lambda = 0 \) in \( \Omega '_\Upsilon \) and taking \( \widetilde{u}_\lambda \) as a test function, we obtain
Since
and
where
we derive
Now, by (3.7),
This form, \( \widetilde{u}_\lambda = 0 \) in \( \left( \mathbb R^N \setminus \Omega '_\Upsilon \right) _+ \). Obviously, \( \widetilde{u}_\lambda = 0 \) at the points where \( u_\lambda \le a_- \), consequently, \( \widetilde{u}_\lambda = 0 \) in \( \mathbb R^N \setminus \Omega '_\Upsilon \).
6 A special critical value for \( \phi _\lambda \)
For each \( j = 1, \ldots , k \), consider
the energy functional associated to \( (P_j) \), and
the energy functional associated to
It is fulfilled that \( I_j \) and \( \phi _{ \lambda ,j } \) satisfy the mountain pass geometry and let
their respective mountain pass levels, where
and
Invoking the \( (PS) \) condition on \( I_j \) and \( \phi _{ \lambda ,j}\), we ensure that there exist \( w_j \in W^{ 1,p(x) }_0 \big ( \Omega _j \big ) \) and \( w_{ \lambda ,j } \in W^{ 1,p(x) } \big ( \Omega '_j \big ) \) such that
and
Lemma 6.1
There holds that
-
(i)
\( 0 < c_{ \lambda ,j } \le c_j, \, \forall \lambda \ge 1, \, \forall j \in \left\{ 1, \ldots , k \right\} \);
-
(ii)
\( c_{ \lambda ,j } \rightarrow c_j, \text { as } \lambda \rightarrow \infty , \, \forall j \in \left\{ 1, \ldots , k \right\} \).
Proof
-
(i)
Once \( W^{ 1,p(x) }_0 \big ( \Omega _j \big ) \subset W^{ 1,p(x) } \big ( \Omega '_j \big ) \) and \( \phi _{ \lambda ,j } \big ( \gamma (1) \big ) = I_j \big ( \gamma (1) \big ) \) for \( \gamma \in \Gamma _j \), we have \( \Gamma _j \subset \Gamma _{ \lambda ,j } \). This way
$$\begin{aligned} c_{ \lambda ,j } = \inf _{ \gamma \in \Gamma _{ \lambda ,j } } \max _{ t \in [0,1] } \phi _{ \lambda ,j } \big ( \gamma (t) \big ) \le \inf _{ \gamma \in \Gamma _j } \max _{ t \in [0,1] } \phi _{ \lambda ,j } \big ( \gamma (t) \big ) = \inf _{ \gamma \in \Gamma _j } \max _{ t \in [0,1] } I_j \big ( \gamma (t) \big ) = c_j. \end{aligned}$$ -
(ii)
It suffices to show that \( c_{ \lambda _n,j } \rightarrow c_j, \text { as } n \rightarrow \infty \), for all sequences \( ( \lambda _n ) \) in \( [1,\infty ) \) with \( \lambda _n \rightarrow \infty , \text { as } n \rightarrow \infty \). Let \( \left( \lambda _n \right) \) be such a sequence and consider an arbitrary subsequence of \( \left( c_{ \lambda _n,j } \right) \) (not relabeled) . Let \( w_n \in W^{ 1,p(x) } \big ( \Omega '_j \big ) \) with
$$\begin{aligned} \phi _{ \lambda _n,j } \big ( w_n \big ) = c_{ \lambda _n,j } \, \text { and } \, \phi '_{ \lambda _n,j } \big ( w_n \big ) = 0. \end{aligned}$$By the previous item, \( \big ( c_{ \lambda _n,j } \big ) \) is bounded. Then, there exists \( \big ( w_{ n_k } \big ) \) subsequence of \( \big ( w_n \big ) \) such that \( \phi _{ \lambda _{ n_k },j } \big ( w_{ n_k } \big ) \) converges and \( \phi '_{ \lambda _{ n_k },j } \big ( w_{ n_k } \big ) = 0 \). Now, repeating the same type of arguments explored in the proof of Proposition 4.1, there is \( w \in W^{ 1,p(x) }_0 \big ( \Omega _j \big ) \setminus \{0\} \subset W^{ 1,p(x) } \big ( \Omega '_j \big ) \) such that
$$\begin{aligned} w_{ n_k } \rightarrow w \text { in } W^{ 1,p(x) } \big ( \Omega '_j \big ), \text { as } k \rightarrow \infty . \end{aligned}$$Furthermore, we also can prove that
$$\begin{aligned} c_{ \lambda _{ n_k },j } = \phi _{ \lambda _{ n_k },j } \big ( w_{ n_k } \big ) \rightarrow I_j(w) \end{aligned}$$and
$$\begin{aligned} 0 = \phi '_{ \lambda _{ n_k },j } \big ( w_{ n_k } \big ) \rightarrow I'_j(w). \end{aligned}$$Then, by \( (f_4) \),
$$\begin{aligned} \lim _k c_{ \lambda _{ n_k },j } \ge c_j. \end{aligned}$$The last inequality together with item (i) implies
$$\begin{aligned} c_{ \lambda _{ n_k },j } \rightarrow c_j, \text { as } k \rightarrow \infty . \end{aligned}$$This establishes the asserted result.
\(\square \)
In the sequel, let \( R > 1 \) verifying
There holds that
Moreover, to simplify the notation, we rename the components \( \Omega _j \) of \( \Omega \) in way such that \( \Upsilon = \{ 1, 2, \ldots , l \} \) for some \( 1 \le l \le k \). Then, we define:
and
Next, our intention is proving that \( b_{ \lambda , \Upsilon } \) is a critical value for \( \phi _\lambda \). However, to do this, we need to some technical lemmas. The arguments used are the same found in [3]; however, for reader’s convenience, we will repeat their proofs
Lemma 6.2
For all \( \gamma \in \Gamma _*\), there exists \( (s_1, \ldots , s_l ) \in [1/R^2,1]^l \) such that
Proof
Given \( \gamma \in \Gamma _*\), consider \( \widetilde{\gamma } :[1/R^2,1]^l \rightarrow \mathbb R^l \) such that
For \( \mathbf t \in \partial [1/R^2,1]^l \), it holds \( \widetilde{\gamma } ( \mathbf t ) = \widetilde{\gamma _0} ( \mathbf t ) \). From this, we observe that there is no \( \mathbf t \in \partial [1/R^2,1]^l \) with \( \widetilde{\gamma } ( \mathbf t ) = 0 \). Indeed, for any \( j \in \Upsilon \),
This form, if \( \mathbf{{t}} \in \partial [1/R^2,1]^l \), then \( t_{j_0} =1 \) or \( t_{j_0} = \frac{1}{R^2} \), for some \( j_0 \in \Upsilon \). Consequently,
Therefore, if \( \phi '_{ \lambda ,j_0 } \big ( \gamma _0 ( \mathbf{{t}} ) \big ) \gamma _0 ( \mathbf{{t}} ) = 0 \), we get \( I_{j_0} ( R w_{j_0} ) \ge c_{j_0} \) or \( I_{j_0} \left( \frac{1}{R} w_{j_0} \right) \ge c_{j_0} \), which is a contradiction with (6.1).
Now, we compute the degree \( \deg \big ( \widetilde{\gamma }, (1/R^2,1)^l, (0, \ldots , 0 ) \big ) \). Since
and, for \( \mathbf t \in (1/R^2,1)^l \),
we derive
This shows what was stated.\(\square \)
Proposition 6.3
If \( c_{ \lambda ,\Upsilon } = \displaystyle \sum _{ j=1 }^l c_{ \lambda ,j } \, \text { and } \, c_\Upsilon = \sum _{ j=1 }^l c_j \), then
-
(i)
\( c_{ \lambda ,\Upsilon } \le b_{ \lambda ,\Upsilon } \le c_\Upsilon , \, \forall \lambda \ge 1 \);
-
(ii)
\( b_{ \lambda ,\Upsilon } \rightarrow c_\Upsilon , \text { as } \lambda \rightarrow \infty \);
-
(iii)
\( \phi _\lambda \big ( \gamma (\mathbf{{t}}) \big ) < c_\Upsilon , \, \forall \lambda \ge 1, \gamma \in \Gamma _*\text { and } \mathbf{{t}} = (t_1, \ldots , t_l ) \in \partial [1/R^2,1]^l \).
Proof
-
(i)
Once \( \gamma _0 \in \Gamma _*\),
$$\begin{aligned} b_{ \lambda ,\Upsilon } \le \max _{ ( t_1, \ldots , t_l ) \in [1/R^2,1]^l } \phi _\lambda \big ( \gamma _0 ( t_1, \ldots , t_l ) \big ) = \max _{ ( t_1, \ldots , t_l )\in [1/R^2,1]^l } \sum _{ j=1 }^l I_j ( t_j R w_j ) = c_\Upsilon . \end{aligned}$$Now, fixing \( \mathbf{s} = (s_1, \ldots , s_l) \in [1/R^2,1]^l \) given in Lemma 6.2 and recalling that
$$\begin{aligned} c_{ \lambda ,j } = \inf \left\{ \phi _{ \lambda ,j } (u) \, ; \, u \in W^{ 1,p(x) } \big ( \Omega '_j \big ) \setminus \{ 0 \} \text { and } \phi '_{ \lambda ,j }(u)u = 0 \right\} , \end{aligned}$$it follows that
$$\begin{aligned} \phi _{ \lambda ,j } \big ( \gamma ( \mathbf{s } ) \big ) \ge c_{ \lambda ,j }, \, \forall j \in \Upsilon . \end{aligned}$$From (3.9),
$$\begin{aligned} \phi _{ \lambda , \mathbb R^N \setminus \Omega '_\Upsilon } (u) \ge 0, \, \forall u \in W^{ 1,p(x) } \big ( \mathbb R^N \setminus \Omega '_\Upsilon \big ), \end{aligned}$$which leads to
$$\begin{aligned} \phi _\lambda \big ( \gamma ( \mathbf{t} ) \big ) \ge \sum _{ j=1 }^l \phi _{ \lambda ,j } \big ( \gamma ( \mathbf{t} ) \big ), \, \forall \mathbf t = (t_1, \ldots , t_l) \in [1/R^2,1]^l. \end{aligned}$$Thus
$$\begin{aligned} \max _{ ( t_1, \ldots , t_l )\in [1/R^2,1]^l } \phi _\lambda \big ( \gamma ( t_1, \ldots , t_l ) \big ) \ge \phi _\lambda \big ( \gamma ( \mathbf s ) \big ) \ge c_{ \lambda ,\Upsilon }, \end{aligned}$$showing that
$$\begin{aligned} b_{ \lambda ,\Upsilon } \ge c_{ \lambda ,\Upsilon }; \end{aligned}$$ -
(ii)
This limit is clear by the previous item, since we already know \( c_{ \lambda ,j } \rightarrow c_j \), as \( \lambda \rightarrow \infty \);
-
(iii)
For \( \mathbf t = ( t_1, \ldots , t_l ) \in \partial [1/R^2,1]^l \), it holds \( \gamma ( \mathbf t ) = \gamma _0 ( \mathbf t ) \). From this,
$$\begin{aligned} \phi _\lambda \big ( \gamma ( \mathbf t ) \big ) = \sum _{j=1}^l I_j ( t_j R w_j ). \end{aligned}$$Writing
$$\begin{aligned} \phi _\lambda \big ( \gamma ( \mathbf t ) \big ) = \mathop {\mathop {\sum }\limits _{j=1}}\limits _{j \ne j_0 }^l I_j ( t_j R w_j ) + I_{j_0} ( t_{j_0} R w_{j_0} ), \end{aligned}$$where \( t_{j_0} \in \left\{ \frac{1}{R^2}, 1 \right\} \), from (6.1) we derive
$$\begin{aligned} \phi _\lambda \big ( \gamma ( \mathbf t ) \big ) \le c_\Upsilon - \epsilon , \end{aligned}$$for some \( \epsilon > 0 \), so (iii).
\(\square \)
Corollary 6.4
\( b_{ \lambda ,\Upsilon } \) is a critical value of \( \phi _\lambda \), for \( \lambda \) sufficiently large.
Proof
Assume \( b_{ \widetilde{\lambda },\Upsilon } \) is not a critical value of \( \phi _{\widetilde{\lambda }} \) for some \( \widetilde{\lambda }\). We will prove that exists \( \lambda _1 \) such that \( \widetilde{\lambda } < \lambda _1 \). Indeed, by item (iii) of Proposition 6.3, we have seen that
This way
Since \( b_{ \lambda ,\Upsilon } \rightarrow c_\Upsilon \) (item (ii) of Proposition 6.3), there exists \( \lambda _1 > 1 \) such that if \( \lambda \ge \lambda _1 \), then
So, if \( \widetilde{\lambda } \ge \lambda _1 \), we can find \( \tau = \tau ( \widetilde{\lambda } ) > 0 \) small enough, with the ensuing property
From the deformation’s lemma [31, Page 38], there is \( \eta :E_\lambda \rightarrow E_\lambda \) such that
Then, by (6.2),
Now, using the definition of \( b_{ \widetilde{\lambda },\Upsilon } \), there exists \( \gamma _*\in \Gamma _*\) satisfying
Defining
due to (6.3), we obtain
But since \( \widetilde{\gamma } \in \Gamma _*\), we deduce
a contradiction. So, \( \widetilde{\lambda } < \lambda _1\).\(\square \)
7 The proof of the main theorem
To prove Theorem 1.1, we need to find nonnegative solutions \( u_\lambda \) for large values of \( \lambda \), which converges to a least energy solution in each \( \Omega _j \) \( (j \in \Upsilon ) \) and to \( 0 \) in \(\Omega _\Upsilon ^{c}\) as \( \lambda \rightarrow \infty \). To this end, we will show two propositions which together with the Propositions 4.1 and 5.1 will imply that Theorem 1.1 holds.
Henceforth, we denote by
and
Moreover, for small values of \( \mu \),
We observe that
showing that \( \mathcal{A}_\mu ^\lambda \cap \phi _\lambda ^{ c_\Upsilon } \ne \emptyset \). Fixing
we have the following uniform estimate of \( \big \Vert \phi '_{ \lambda }(u) \big \Vert \) on the region \( \left( \mathcal{A}_{ 2 \mu }^\lambda \setminus \mathcal{A}_\mu ^\lambda \right) \cap \phi _\lambda ^{ c_\Upsilon } \).
Proposition 7.1
Let \( \mu > 0 \) satisfying (7.1). Then, there exist \( \Lambda _*\ge 1 \) and \( \sigma _0 >0 \) independent of \( \lambda \) such that
Proof
We assume that there exist \( \lambda _n \rightarrow \infty \) and \( u_n \in \left( \mathcal{A}_{ 2 \mu }^{\lambda _n} \setminus \mathcal{A}_\mu ^{\lambda _n} \right) \cap \phi _{\lambda _n}^{ c_\Upsilon } \) such that
Since \( u_n \in \mathcal{A}_{ 2 \mu }^{ \lambda _n } \), this implies \( \big ( \varrho _{ \lambda _n } (u_n) \big ) \) is a bounded sequence and, consequently, it follows that \( \big ( \phi _{ \lambda _n }(u_n) \big )\) is also bounded. Thus, passing a subsequence if necessary, we can assume \(\phi _{ \lambda _n }(u_n) \) converges. Thus, from Proposition 4.1, there exists \( 0 \le u \in W^{ 1,p(x) }_0 \big ( \Omega _\Upsilon \big ) \) such that \( u_{ |_{ \Omega _j } }, \, j \in \Upsilon \), is a solution for \( (P_j) \),
We know that \( c_j \) is the least energy level for \( I_j \). So, if \( u_{ |_{ \Omega _j } } \ne 0 \), then \( I_j(u) \ge c_j \). But since \( \phi _{ \lambda _n } (u_n) \le c_\Upsilon \), we must analyze the following possibilities:
-
(i)
\( I_j(u) = c_j, \, \forall j \in \Upsilon \);
-
(ii)
\( I_{ j_0 }(u) = 0 \), for some \( j_o \in \Upsilon \).
If (i) occurs, then for \( n \) large, it holds
So \( u_n \in \mathcal{A}_\mu ^{\lambda _n} \), a contradiction.
If (ii) occurs, then
which is a contradiction with the fact that \( u_n \in \mathcal{A}_{ 2 \mu }^{\lambda _n} \). Thus, we have completed the proof.\(\square \)
Proposition 7.2
Let \( \mu > 0 \) satisfying (7.1) and \( \Lambda _*\ge 1 \) given in the previous proposition. Then, for \( \lambda \ge \Lambda _*\), there exists a solution \( u_\lambda \) of \( (A_\lambda ) \) such that \( u_\lambda \in \mathcal{A}_\mu ^\lambda \cap \phi _\lambda ^{ c_\Upsilon } \).
Proof
Let \( \lambda \ge \Lambda _*\). Assume that there are no critical points of \( \phi _\lambda \) in \( \mathcal{A}_\mu ^\lambda \cap \phi _\lambda ^{ c_\Upsilon } \). Since \( \phi _\lambda \) is a \( (PS) \) functional, there exists a constant \( d_\lambda > 0 \) such that
From Proposition 7.1, we have
where \( \sigma _0 > 0 \) does not depend on \( \lambda \). In what follows, \( \Psi :E_\lambda \rightarrow \mathbb R \) is a continuous functional verifying
We also consider \( H :\phi _\lambda ^{ c_\Upsilon } \rightarrow E_\lambda \) given by
where \( Y \) is a pseudo-gradient vector field for \( \Phi _\lambda \) on \( \mathcal{K} = \left\{ u \in E_\lambda \, ; \, \phi '_\lambda (u) \ne 0 \right\} \). Observe that \( H \) is well defined, once \( \phi '_\lambda (u) \ne 0 \), for \( u \in \mathcal{A}_{2 \mu }^\lambda \cap \phi _\lambda ^{ c_\Upsilon } \). The inequality
guarantees that the deformation flow \( \eta :[0, \infty ) \times \phi _\lambda ^{ c_\Upsilon } \rightarrow \phi _\lambda ^{ c_\Upsilon } \) defined by
verifies
and
We study now two paths, which are relevant for what follows:
\( \bullet \) The path \( \mathbf{t} \mapsto \eta \big ( t, \gamma _0( \mathbf{t} ) \big ), \text { where } \mathbf t = (t_1,\ldots ,t_l) \in [1/R^2, 1]^l \).
The definition of \( \gamma _0 \) combined with the condition on \( \mu \) gives
Since
from (7.5), it follows that
So, \( \eta \big ( t, \gamma _0( \mathbf{t} ) \big ) \in \Gamma _*\), for each \( t \ge 0 \).
\( \bullet \) The path \( \mathbf{t} \mapsto \gamma _0( \mathbf{t} ), \text { where } \mathbf t = (t_1,\ldots ,t_l) \in [1/R^2, 1]^l \).
We observe that
and
forall \( \mathbf{t} \in [1/R^2, 1]^l \). Moreover,
and
Therefore
is independent of \( \lambda \) and \( m_0 < c_\Upsilon \). Now, observing that there exists \( K_*> 0 \) such that
we derive
for \( T > 0 \) large.
In fact, writing \( u = \gamma _0( \mathbf{t} ) , \mathbf{t } \in [1/R^2,1]^l \), if \( u \notin A_\mu ^\lambda \), from (7.3),
and we have nothing more to do. We assume then \( u \in A_\mu ^\lambda \) and set
Now, we will analyze the ensuing cases:
Case 1: \( \widetilde{\eta }(t) \in \mathcal{A}_{\frac{3}{2} \mu }^\lambda , \, \forall t \in [0,T] \).
Case 2: \( \widetilde{\eta }(t_0) \in \partial \mathcal{A}_{\frac{3}{2} \mu }^\lambda , \text { for some } t_0 \in [0,T] \).
Analysis of Case 1
In this case, we have \( \Psi \big ( \widetilde{\eta }(t) \big ) = 1 \) and \( \big \Vert \phi '_\lambda \big ( \widetilde{\eta }(t) \big ) \big \Vert \ge \widetilde{d_\lambda } \) for all \( t \in [0,T] \). Hence, from (7.3),
that is,
showing (7.6).
Analysis of Case 2
In this case, there exist \( 0 \le t_1 \le t_2 \le T \) satisfying
and
We claim that
Setting \( w_1 = \widetilde{\eta }(t_1) \) and \( w_2 = \widetilde{\eta }(t_2) \), we get
for some \( j_0 \in \Upsilon \). We analyze the latter situation, once that the other one follows the same reasoning. From the definition of \( \mathcal{A}_\mu ^\lambda \),
consequently,
Then, by mean value theorem, \( t_2-t_1 \ge \frac{1}{2 K_*} \mu \) and, this form,
implying
which proves 7.6. Fixing \( \widehat{\eta } (t_1, \ldots , t_l) = \eta \big ( T, \gamma _0 (t_1,\ldots ,t_l) \big ) \), we have that \( \widehat{\eta } \in \Gamma _*\) and, hence,
which contradicts the fact that \( b_{ \lambda , \Upsilon } \rightarrow c_\Upsilon \).\(\square \)
Proof of Theorem 1.1
According Proposition 7.2, for \(\mu \) satisfying (7.1) and \( \Lambda _*\ge 1 \), there exists a solution \( u_\lambda \) for \( (A_\lambda ) \) such that \( u_\lambda \in \mathcal{A}_\mu ^\lambda \cap \phi _\lambda ^{ c_\Upsilon } \), for all \(\lambda \ge \Lambda _*\).
Claim: There are \(\lambda _0 \ge \Lambda _*\) and \(\mu _0>0\) small enough, such that \(u_\lambda \) is a solution for \( \big ( P_\lambda \big )\) for \(\lambda \ge \Lambda _0\) and \(\mu \in (0, \mu _0)\).
Indeed, assume by contradiction that there are \( \lambda _n \rightarrow \infty \) and \( \mu _n \rightarrow 0 \), such that \((u_{\lambda _n})\) is not a solution for \((P_{\lambda _n})\). From Proposition 7.2, the sequence \( (u_{\lambda _n}) \) verifies:
-
(a)
\( \phi '_{ \lambda _n }(u_{\lambda _n}) = 0, \, \forall n \in \mathbb {N}\);
-
(b)
\( \varrho _{ \lambda _n, \mathbb {R}^N \setminus \Omega _\Upsilon }(u_{\lambda _n}) \rightarrow 0\);
-
(c)
\( \phi _{ \lambda _n,j } (u_{\lambda _n}) \rightarrow c_j, \, \forall j \in \Upsilon . \)
The item (b) ensures we can use Proposition 5.1 to deduce \( u_{\lambda _n} \) is a solution for \( \big ( P_{\lambda _n} \big ) \), for large values of \( n \), which is a contradiction, showing this way the claim.
Now, our goal is to prove the second part of the theorem. To this end, let \((u_{\lambda _n})\) be a sequence verifying the above limits. Since \( \phi _{ \lambda _n }(u_{ \lambda _n } ) \) is bounded, passing a subsequence, we obtain that \( \phi _{ \lambda _n }(u_{ \lambda _n } ) \rightarrow c \). This way, using Proposition 4.1 combined with item (c), we derive \( u_{ \lambda _n } \) converges in \( W^{ 1,p(x) } \big ( \mathbb {R}^N \big ) \) to a function \( u \in W^{ 1,p(x) } \big ( \mathbb {R}^N \big ) \), which satisfies \( u = 0 \) outside \( \Omega _\Upsilon \) and \( u_{|_{\Omega _j}}, \, j \in \Upsilon \), is a least energy solution for
References
Acerbi, E., Mingione, G.: Regularity results for stationary electrorheological fluids. Arch. Ration. Mech. Anal. 164, 213–259 (2002)
Acerbi, E., Mingione, G.: Regularity results for electrorheological fluids: stationary case. C. R. Math. Acad. Sci. Paris 334, 817–822 (2002)
Alves, C.O.: Existence of multi-bump solutions for a class of quasilinear problems. Adv. Nonlinear Stud. 6, 491–509 (2006)
Alves, C.O.: Existence of solutions for a degenerate \( p(x)\)-Laplacian equation in \(\mathbb{R}^N \). J. Math. Anal. Appl. 345, 731–742 (2008)
Alves, C.O.: Existence of radial solutions for a class of \( p(x)\)-Laplacian equations with critical growth. Differ. Integral Equ. 23, 113–123 (2010)
Alves, C.O., Barreiro, J.L.P.: Existence and multiplicity of solutions for a \( p(x) \)-Laplacian equation with critical growth. J. Math. Anal. Appl. 403, 143–154 (2013)
Alves, C.O., Ferreira, M.C.: Nonlinear perturbations of a \(p(x)\)-Laplacian equation with critical growth in \(\mathbb{R}^N \). Math. Nach. 287(8–9), 849–868 (2014)
Alves, C.O., Ferreira, M.C.: Existence of solutions for a class of \(p(x)\)-Laplacian equations involving a concave-convex nonlinearity with critical growth in \(\mathbb{R}^N\). Topol. Methods Nonlinear Anal. (2014, to appear)
Alves, C.O., Souto, M.A.S.: Existence of solutions for a class of problems in \(\mathbb{R}^N \) involving \(p(x)\)-Laplacian. Prog. Nonlinear Differ. Equ. Their Appl. 66, 17–32 (2005)
Ambrosetti, A., Rabinowitz, P.H.: Dual variational methods in critical point theory and applications. J. Funct. Anal. 14, 349–381 (1973)
Antontsev, S.N., Rodrigues, J.F.: On stationary thermo-rheological viscous flows. Ann. Univ. Ferrara Sez. VII Sci. Mat. 52, 19–36 (2006)
Chambolle, A., Lions, P.L.: Image recovery via total variation minimization and related problems. Numer. Math. 76, 167–188 (1997)
Chen, Y., Levine, S., Rao, M.: Variable exponent, linear growth functionals in image restoration. SIAM J. Appl. Math. 66, 1383–1406 (2006)
del Pino, M., Felmer, P.L.: Local mountain passes for semilinear elliptic problems in unbounded domains. Calc. Var. PDE 4, 121–137 (1996)
Ding, Y.H., Tanaka, K.: Multiplicity of positive solutions of a nonlinear Schrödinger equation. Manuscr. Math. 112(1), 109–135 (2003)
Fan, X.L.: On the sub-supersolution method for \( p(x) \)-Laplacian equations. J. Math. Anal. Appl. 330, 665–682 (2007)
Fan, X.L.: \( p(x) \)-Laplacian equations in \(\mathbb{R}^N \) with periodic data and nonperiodic perturbations. J. Math. Anal. Appl. 341, 103–119 (2008)
Fan, X., Zhao, D.: A class of De Giorgi type and Hölder continuity. Nonlinear Anal. 36, 295–318 (1999)
Fan, X.L., Zhao, D.: On the Spaces \( L^{ p(x) } \big ( \Omega \big ) \) and \( W^{ 1, p(x)} (\Omega ) \). J. Math. Anal. Appl. 263, 424–446 (2001)
Fan, X.L., Zhao, D.: Nodal solutions of \( p(x) \)-Laplacian equations. Nonlinear Anal. 67, 2859–2868 (2007)
Fan, X.L., Shen, J.S., Zhao, D.: Sobolev embedding theorems for spaces \( W^{ k, p(x) } (\Omega )\). J. Math. Anal. Appl. 262, 749–760 (2001)
Fernández, Bonder J., Saintier, N., Silva, A.: On the Sobolev embedding theorem for variable exponent spaces in the critical range. J. Differ. Equ. 253, 1604–1620 (2012)
Fernández Bonder, J., Saintier, N., Silva, A.: On the Sobolev trace theorem for variable exponent spaces in the critical range. Ann. Mat. Pura Appl. (2014, to appear)
Fu, Y., Zhang, X.: Multiple solutions for a class of \(p(x)\)-Laplacian equations in involving the critical exponent. Proc. R. Soc. Edinb. Sect. A 466, 1667–1686 (2010)
Fusco, N., Sbordone, C.: Some remarks on the regularity of minima of anisotropic integrals. Commun. Partial Differ. Equ. 18(1–2), 153–167 (1993)
Kavian, O.: Introduction à la théorie de points critiques et applications aux problèmes elliptiques. Springer, Paris (1993)
Ladyzhenskaya, O.A., Ural’tseva, N.N.: Linear and quasilinear elliptic equations. Academic Press, New York (1968)
Mih\(\breve{a}\)ilescu, M., R\(\breve{a}\)dulescu, V.: On a nonhomogeneous quasilinear eigenvalue problem in Sobolev spaces with variable exponent. Proc. Am. Math. Soc. 135(9), 2929–2937 (2007)
Ruzicka, M.: Electrorheological fluids: modeling and mathematical theory. Lecture Notes in Mathematics, vol. 1748, Springer, Berlin (2000)
Séré, E.: Existence of infinitely many homoclinic orbits in Hamiltonian systems. Math. Z. 209, 27–42 (1992)
Willem, M.: Minimax Theorems. Birkhäuser, Boston (1996)
Acknowledgments
The authors would like to thank the anonymous referee for their valuable suggestions.
Author information
Authors and Affiliations
Corresponding author
Additional information
Partially supported by INCT-MAT and PROCAD.
C. O. Alves was partially supported by CNPq/Brazil 303080/2009-4.
Rights and permissions
About this article
Cite this article
Alves, C.O., Ferreira, M.C. Multi-bump solutions for a class of quasilinear problems involving variable exponents. Annali di Matematica 194, 1563–1593 (2015). https://doi.org/10.1007/s10231-014-0434-2
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10231-014-0434-2