Solutions with multiple spike patterns for an elliptic system (Q2461638)

The authors consider the system \[ -\varepsilon^2 \Delta u + V(x)u=g(v),~-\varepsilon^2 \Delta v + V(x)v=f(u)~\text{in } \Omega,~~u=v=0~\text{on } \partial\Omega \] where \(\Omega \subset \mathbb{R}^N\) is a smooth domain (not necessarily bounded), \(V(x)\) is locally Hölder continuous and \(f,g \in C^1(\mathbb{R})\) satisfy: (fg1) \(f(0)=g(0)=f'(0)=g'(0)=0\); (fg2) \(\lim_{s\to\infty} f(s)/s^{p-1} = \lim_{s\to\infty} f(s)/s^{q-1}=0\) for some \(p,q>2\) such that \[ \frac{1}{p}+\frac{1}{q} > \frac{N-2}{N}; \] (fg3) \(0< (1+\delta)f(s)s \leq f'(s)s^2\) and \(0< (1+\delta)g(s)s \leq g'(s)s^2\), for some \(\delta>0\). Concerning the potential, it is assumed that: {\parindent=9mm \begin{itemize}\item[(V1)] \(V(x) \geq \alpha >0 \), for all \(x\in \Omega\); \item[(V2)] there exist bounded domains \(\Lambda_i\), mutually disjoint, compactly contained in \(\Omega\), \(i=1,\dots,k\), such that \[ \inf_{\Lambda_i} V < \inf_{\partial \Lambda_i} V. \] \end{itemize}} The authors prove that, under assumptions (V1)--(V2) and (fg1)--(fg3), there exists \(\varepsilon_0>0\) such that, for any \(\varepsilon \in (0,\varepsilon_0)\), the system admits classical positive solutions \(u_{\varepsilon},v_{\varepsilon} \in C^2(\Omega) \cap C^1(\overline{\Omega}) \cap H_0^1(\Omega)\) satisfying: {\parindent=7mm \begin{itemize}\item[(i)] there exist \(x_{i,\varepsilon} \in \Lambda_i\), \(i=1,\dots,k\), local maximum points of both \(u_{\varepsilon}\) and \(v_{\varepsilon}\); \item[(ii)] \(u_{\varepsilon}(x_{i,\varepsilon}),v_{\varepsilon}(x_{i,\varepsilon})\geq b>0\) and \(V(x_{i,\varepsilon}) \to \inf_{\Lambda_i} V\) as \(\varepsilon\to 0\); \item[(iii)] \(u_{\varepsilon}(x_{i,\varepsilon}),v_{\varepsilon}(x_{i,\varepsilon})\leq \gamma \exp(-\frac{\beta}{\varepsilon}|x-x_{i,\varepsilon}|)\), for all \(x \in \Omega \setminus \bigcup_{j \neq i} \Lambda_j\); for some positive constants \(b, \gamma, \beta\). \end{itemize}} Moreover, the uniqueness of the local maximum holds in the following sense: {\parindent=7mm \begin{itemize}\item[(iv)] if either \(u_{\varepsilon}\) or \(v_{\varepsilon}\) have a local maximum point at \(z_{\varepsilon} \in \Omega\), \(z_{\varepsilon} \neq x_{i,\varepsilon}\), for all \(i=1,\dots,k\), then \(\lim_{\varepsilon\to 0} u_{\varepsilon}(z_{\varepsilon})=0\) and \(\lim_{\varepsilon\to 0} v_{\varepsilon}(z_{\varepsilon})=0\) hold. \end{itemize}} In the proof the authors use the well known penalization scheme of \textit{M. Del Pino} and \textit{P. L. Felmer} [Ann. Inst. Henri Poincaré, Anal. Non Linéaire 15, No. 2, 127--149 (1998; Zbl 0901.35023)] and some arguments introduced by the the first author in some previous works.