Nonsymmetric sign-changing solutions to overdetermined elliptic problems in bounded domains

Abstract

In 1971 J. Serrin proved that, given a smooth bounded domain $\Omega \subset {\mathbb{R}}^N$ and $u$ a positive solution of the problem: \begin{equation*} \begin{array}{ll} -Δu = f(u) &\mbox{in $\Omega$, } u =0 &\mbox{on $\partial \Omega$, } \partial_ν u =\mbox{constant} &\mbox{on $\partial \Omega$, } \end{array} \end{equation*} then $\Omega$ is necessarily a ball and $u$ is radially symmetric. In this paper we prove that the positivity of $u$ is necessary in that symmetry result. In fact we find a sign-changing solution to that problem for a $C^2$ function $f(u)$ in a bounded domain $\Omega$ different from a ball. The proof uses a local bifurcation argument, based on the study of the associated linearized operator.

26 pages, 1 figure. In this new version I have corrected some minor mistakes, and I thank Antonio Jesús Fernández for pointing me out. Comments are welcome