Symmetry and linear stability in Serrin's overdetermined problem via the stability of the parallel surface problem

Abstract: We consider the solution of the problem $$ -\Delta u=f(u) \ \mbox{ and } \ u>0 \ \ \mbox{ in } \ \Omega, \ \ u=0 \ \mbox{ on } \ \Gamma, $$ where $\Omega$ is a bounded domain in $\mathbb{R}^N$ with boundary $\Gamma$ of class $C^{2,\tau}$, $0<\tau<1$, and $f$ is a locally Lipschitz continuous non-linearity. Serrin's celebrated symmetry theorem states that, if the normal derivative $u_\nu$ is constant on $\Gamma$, then $\Omega$ must be a ball. In [CMS2], it has been conjectured that Serrin's theorem may be obtained by stability in the following way: first, for a solution $u$ prove the estimate $$ r_e-r_i\le C_\delta\,[u]{\Gamma^\delta} $$ for some constant $C\delta$ depending on $\delta>0$, where $r_e$ and $r_i$ are the radii of a spherical annulus containing $\Gamma$, $\Gamma^\delta$ is a surface parallel to $\Gamma$ at distance $\delta$ and sufficiently close to $\Gamma$, and $[u]{\Gamma^\delta}$ is the Lipschitz semi-norm of $u$ on $\Gamma^\delta$; secondly, if in addition $u\nu$ is constant on $\Gamma$, show that $$ [u]{\Gamma^\delta}=o(C\delta)\ \mbox{ as } \ \delta\to 0^+. $$ In this paper, we prove that this strategy is successful. As a by-product of this method, for $C^{2,\tau}$-regular domains, we also obtain a linear stability estimate for Serrin's symmetry result. Our result is optimal and greatly improves the similar logarithmic-type estimate of [ABR] and the H\"{o}lder estimate of [CMV] that was restricted to convex domains.