On spike solutions of singularly perturbed semilinear Dirichlet problem (Q1340452)

The author considers the singularly perturbed semilinear elliptic boundary value problem \[ \varepsilon^ 2 \Delta u+ f(u)=0, \qquad u>0 \text{ in } \Omega, \qquad u=0 \text{ on } \partial\Omega.\tag \(*\) \] \(\Omega\) is a bounded smooth convex domain in \(\mathbb{R}^ n\). The nonlinear term is supposed to have precisely three zeros: \(f(0)= f(z_ 1)= f(z_ 2)=0\) for some \(0<z_ 1< z_ 2\). Further the author assumes \(f'(0)< 0\), \(f'(z_ 2)<0\) and \(\int_ 0^{z_ 2} f(s) ds>0\). As a prototype we may think of \(f(u)=- u(u-1) (u-a)\), \(0<a< {1\over 2}\). It is known e.g. from \textit{Ph. Clément} and \textit{G. Sweers} [Ann. Sc. Norm. Super. Pisa, Cl. Sci., IV. Ser. 14, No. 1, 97-121 (1987; Zbl 0662.35045)], that for \(\varepsilon\) small, there is a maximal solution \(\overline{u}_ \varepsilon >0\). That \(\overline {u}_ \varepsilon\) be maximal means, that for any \(\Phi\in C_ 0^ \infty\) with \(\max_{\overline {\Omega}} \Phi\in (z_ 1, z_ 2)\) and \(\varepsilon\) small, \(\overline {u}_ \varepsilon\) is the only solution to \((*)\) between \(\Phi\) and \(z_ 2\). As \(\varepsilon \searrow 0\), \(\overline {u}_ \varepsilon \to z_ 2\) locally uniformly in \(\Omega\). Furthermore \(\overline {u}_ \varepsilon\) is (for \(\varepsilon\) small) a global minimizer of the corresponding variational functional \(J_ \varepsilon(u)= \int_ \Omega \{{1\over 2}| \nabla u|^ 2- F(u)\}dx\), where \(F(u)= \int_ 0^ u f(s)ds\). As \(u\equiv 0\) is a local minimizer of \(J_ \varepsilon (\cdot)\), there is another solution \(u_ \varepsilon\) to \((*)\) due to the mountain pass lemma. This ``mountain pass solution'' \(u_ \varepsilon\) is investigated in the present paper. It is shown that, for \(\varepsilon\) small: \(0< u_ \varepsilon< \overline {u}_ \varepsilon\), \(\max_{\overline {\Omega}} u_ \varepsilon> u_ 0\), where \(u_ 0\in (0, z_ 2)\) is the unique number such that \(\int_ 0^{u_ 0} f(s)ds =0\), \(| \partial u_ \varepsilon /\partial \nu| <C\exp (-\gamma/\varepsilon)\) on \(\partial \Omega\) with constants \(C\), \(\gamma>0\) independent of \(\varepsilon\), \(\int_ \Omega u_ \varepsilon^ 2 dx<C\varepsilon^ n\) as \(\varepsilon \searrow 0\). Moreover under some mild additional assumption on \(f\) it can be shown, that \(u_ \varepsilon\) has a single maximum point \(x_ \varepsilon\), the spike, and that \(u_ \varepsilon (\varepsilon\cdot+ x_ \varepsilon) \to V\) in \(C^ 2_{\text{loc}} (\mathbb{R}^ n)\), where \(V\) is the (up to translation) unique positive solution of \(-\Delta V= f(V)\) in \(\mathbb{R}^ n\), \(V(x)\to 0\) as \(| x|\to \infty\).