A sharp stability result for the Gauss mean value formula (Q2227145)

Let $\Omega\subseteq\mathbb{R}^N$ be an open set, and let $\mathcal{H}(\Omega)$ denote the linear space of the \textit{harmonic functions} in $\Omega$. As is very well known, in the particular case when $\Omega = B(x_0,r)$ is a Euclidean ball, then the following \textit{mean-value formula holds}: \[ u(x_0) = \fint_{\Omega}u\,\mathrm{d} y, \qquad\forall\,\,u\in\mathcal{H}(\Omega)\cap L^1(\Omega). \tag{1} \] In 1972, \textit{Ü. Kuran} [Bull. Lond. Math. Soc. 4, 311--312 (1972; Zbl 0257.31006)] proved the \textit{rigidity} of the above mean-value formula: \textit{if $\Omega\subseteq\mathbb{R}^N$ is an open set with finite measure and if there exists $x_0\in D$ such that \[ u(x_0) = \fint_{\Omega}u\,\mathrm{d} y, \qquad\forall\,\,u\in\mathcal{H}(\Omega)\cap L^1(\Omega), \] then $\Omega$ is a Euclidean ball centered at $x_0$}. In the present paper, the authors improve Kuran's result by proving the \textit{stability} of the mean-value formula (1). Broadly put, the problem of the \textit{stability} of (1) can be stated as follows: if $\Omega\subseteq\mathbb{R}^N$ is an open set with finite measure and if there exists $x_0\in\Omega$ such that \[ \text{$u(x_0)$ is \textit{close} to $\fint_{\Omega}u\,\mathrm{d} y$ for every $u\in \mathcal{H}(\Omega)\cap L^1(\Omega)$}, \] is it true that $\Omega$ is \textit{close} to a Euclidean ball (centered at $x_0$)? In order to measure \textit{the closeness} of $u(x_0)$ to its integral mean-value on $\Omega$, the authors introduce the notion of \textit{Gauss mean value gap} (of $\Omega$ relative to $x_0$): \[ G(\Omega,x_0) := \sup_{ \begin{subarray}{c} u\in\mathcal{H}(\Omega)\cap L^1(\Omega) \\ u\not\equiv 0 \end{subarray}} \frac{|u(x_0)-m_\Omega(u)|}{m_\Omega(|u|)}, \] where we have used the simplified notation \[ m_\Omega(f) := \fint_\Omega f\,\mathrm{d} y \qquad (\text{for $f\in L^1(\Omega)$}). \] Moreover, \textit{the closeness} of $\Omega$ to an Euclidean ball centered at $x_0\in\Omega$ is measured in terms of the so-called \textit{Fraenkel asymmetry}: \[ \alpha(\Omega) := \min_{x\in\mathbb{R}^N}\frac{|\Omega\Delta B(x,r_\Omega)|}{|\Omega|}, \] where $r_D$ is the radius of a ball with the same measure of $\Omega$. Using these notations, the problem of the stability of (1) can be rephrased as follows: \[ \text{if $G(D,x_0)$ is small, is it true that $\alpha(D)$ is small?} \] The authors provide a positive answer to this question by establishing a \textit{stronger result}, namely the following stability inequality. \textbf{Theorem 1}. There exists a constant $C = C(N) > 0$ such that, if $\Omega\subseteq\mathbb{R}^N$ is an open set with finite measure and $x_0\in\Omega$, then \[ \frac{|\Omega\setminus B(x_0,r_{x_0})|}{|\Omega|} \leq C(N)\,G(\Omega,x_0), \tag{2} \] where $r_{x_0} := \operatorname{dist}(x_0,\mathrm{d}_e\Omega)$. As a consequence, one also has \[ \alpha(D) \leq 2C(N)\,G(\Omega,x_0). \] The proof of Theorem 1 is very elegant in its simplicity: in fact, it relies essentially on a careful `qualitative' study of the \textit{Kuran function} \[ h(x) := 1+r_{x_0}^{N-2}\frac{|x-x_0|^{2}-r_{x_0}^2}{|x-x_1|^N}, \] where $r_{x_0} = \operatorname{dist}(x_0,\mathrm{d}_e\Omega)$ and $x_1\in \mathrm{d}_e B(x_0,r_{x_0})\cap\mathrm{d}_e\Omega$. Moreover, the estimate from below of the Gauss mean value gap in Theorem 1 is \textit{sharp}: in fact, the authors construct a family $\{\Omega_\varepsilon\}_{\varepsilon}$ of ellipsoids in $\mathbb{R}^N$ such that \[ \frac{1}{c(N)}\frac{|\Omega_\varepsilon\setminus B(0,1)|}{|\Omega_\varepsilon|} \leq G(\Omega_\varepsilon,0)\leq c\frac{|\Omega\Delta B(x_0,r_\Omega)|}{|\Omega|}, \] where $c > 0$ is a suitable constant independent of $\varepsilon$. In order to prove the above two-side estimate, the authors establish an interesting continuity-type result for the Gauss mean value gap with respect to the $C^{1,\alpha}$-convergence of domains; moreover, they also show that this continuity property \textit{cannot be improved} by replacing the $C^{1,\alpha}$-convergence with a weaker one.