Convexity of solutions of semilinear elliptic equations (Q1080066)

Theorem 1. Let \(\Omega\) be a domain in \({\mathbb{R}}^ 2,G,H\in C^{2+\alpha}\), \(\Delta w=G\cdot w+H\cdot w| \nabla w|^ 2>0\) in \(\Omega\), \[ (GG''-2G'{}^ 2)\cdot w+(GH''+G''H-4G'H')\cdot w| \nabla w|^ 2+(HH''-2H'{}^ 2)\cdot w| \nabla w|^ 4\leq 0 \] in \(\Omega\). If \(\phi (w)=2(\partial^ 2_ 1w\partial^ 2_ 2w- (\partial_ 1\partial_ 2w)^ 2)\geq 0\) in \(\Omega\) and \(\phi\) \(\neq 0\), then \(\phi >0\) in \(\Omega\). Theorem 2. Let \(\Omega\) be a strictly convex domain in \({\mathbb{R}}^ 2\) with \(C^{4+\alpha}\) boundary. Let \(G_{\Omega}\) be the Green function for -\(\Delta\) in \(\Omega\), \(G_{\Omega}(x,y)=\ln (1/| x-y|)- k_{\Omega}(x,y)\), \(h_{\Omega}(x)=k_{\Omega}(x,x)\). Then \(h_{\Omega}\) is convex in \(\Omega\) and, if \(\Omega\) is not an infinite strip, then \(h_{\Omega}\) is strictly convex and the level lines of \(h_{\Omega}\) are strictly convex curves. Moreover, if \(\Omega\) is bounded, \(\Delta h_{\Omega}=4e^{2h_{\Omega}}\) in \(\Omega\), \(h_{\Omega}(x)\to 0\) if d(x,\(\partial \Omega)\to 0.\) Theorem 3. Let \(\Omega\) be a bounded convex domain, \(f\in C^{2+\alpha}\), \(f>0\), \(f'(t)+(f(t)/(M-t))>0\) if \(t<M\), \((f^ 2f'{}^ 2/(F(1)-F))+2f'-ff''\geq 0\) where \(F(t)=\int^{t}_{0}f(s)ds\), \(\Delta u=f\cdot u\) in \(\Omega\), \(u=M\) on \(\partial \Omega\), \(M\in {\mathbb{R}}\). Then \(g\cdot u\) is strictly convex in \(\Omega\), where \(g'(t)=1/\sqrt{2F(1)- 2F(t)}\), and the level lines u are strictly convex curves. The same result is true for \(f(t)=(t\vee 0)^ p\) \((0<p<1)\), \(f(t)=| t|^{p-1}t\) \((1\leq p<\infty)\), \(f(t)=(t\vee 0)^{p_ e-\gamma /(\beta +1-\beta (t\vee 0))}\) \((0<p<1;\beta,\gamma >0;f'\geq 0).\) Theorem 4. Let \(\Omega\) be a ring-like domain in \({\mathbb{R}}^ 2\), bounded by an inner boundary \(\Gamma_ 1\) and an outer boundary \(\Gamma_ 2,\Gamma_ i\) are strictly convex \(C^{2+\alpha}\) curves \((i=1,2)\), \(\gamma \in C^{2+\alpha}\) strictly increasing function, \(\gamma (0)=0\), \(\Delta u=\gamma \cdot u\) in \(\Omega\), \(u=M\) on \(\Gamma_ 1\), \(u=m\) on \(\Gamma_ 2\), \(0<m<M\). Then for any \(N>M\) there exists \(n_ 0>0\) (depending on N-M, \(\Omega\) and \(\gamma)\) such that \((N-u)^ n\) is strictly convex in \(\Omega\) for any \(n>n_ 0\).