Rigidity and sharp stability estimates for hypersurfaces with constant and almost-constant nonlocal mean curvature (Q1656827): Difference between revisions

Let \(\Omega\) be an open set in \(\mathbb{R}^n\) with smooth boundary \(\partial\Omega; \partial \Omega\in C^{1,\alpha}\), \(\alpha>2s\). The non-local mean curvature of \(\partial\Omega\), at \(p\in\partial\Omega\) is defined by \[ H_s^\Omega(p)=\frac{1}{\omega_{n-2}}\int_{\mathbb{R}^n}\frac{\tilde{\chi}_\Omega(x)}{|x-p|^{n+2s}}dx, \quad \tilde{\chi}_\Omega(x)=\chi_{\Omega^c}(x)-\chi_\Omega(x). \] Here, \(\chi_E\) denotes the characteristic function of \(E\) and \(\Omega^c=\mathbb{R}^n\setminus\Omega\). By using the moving frame method, the following three theorems are proved. {Theorem 1.1}. If \(\Omega\) is a bounded open set of class \(C^{1,2s}\) (not assumed to be connected) and \(H_s^\Omega\) is constant on \(\partial\Omega\), then \(\partial\Omega\) is a sphere. This is a nonlocal version of \textit{A. D. Aleksandrov}'s Theorem [Transl., Ser. 2, Am. Math. Soc. 21, 412--416 (1962; Zbl 0119.16603)]. The authors remark that an analogous result is also proved by \textit{X. Cabre}et al. [``Curves and surfaces with constant nonlocal mean curvature: meeting Alexandrov and Delaunay'', Preprint, \url{arXiv:1503.00469}]. The other two theorems measure the difference between \(\partial\Omega\) and sphere and give a corresponding stability inequality. For this purpose, the Lipschitz constant \[ \delta_s(\Omega)=\sup_{p,q\in\partial \Omega, p\neq q}\frac{|H_s^\Omega(p)-H_s^\Omega(q)|}{|p-q|},\] and the uniform distance from being a ball \[ \rho(\Omega)=\inf\left\{\frac{t-s}{\mathrm{diam}(\Omega)}:p\in\Omega, B_s(p)\subset \Omega\subset B_t(p)\right\}. \] are introduced and the following result is shown. {Theorem 1.2}. \begin{itemize} \item[(i)] If \(\Omega\) is a bounded open set with \(C^{2,\alpha}\) boundary, \(\alpha>2s\), then there is a dimensional constant \(\hat{C}(n)\) such that \[ \rho(\Omega)\leq \hat{C}(n)\eta_s(\Omega), \quad \eta_s(\Omega)=\frac{\mathrm{diam}(\Omega)^{2n+2s+1}}{|\Omega|^2}\delta_s(\Omega). \] \item[(ii)] There exists \(\eta(n)>0\) such that if \(\eta_s(\Omega)\leq\eta(n)\) then up to rescaling \(\Omega\), we can find a bi-Lipschitz map \(F:\partial B_1(0)\to\partial \Omega\) satisfying \[ \left(1-\bar{C}(n)\sqrt{\eta_s(\Omega)}\right)|x-y|\leq |F(x)-F(y)|\leq \left(1+\bar{C}(n)\sqrt{\eta_s(\Omega)}\right)|x-y|, \] for all \(x,y\in\partial B_1(0)\), for some dimensional constant \(\bar{C}(n)\). \end{itemize} Since both \(\rho(\Omega)\) and \(\eta_s(\Omega)\) are scaling invariant, the following theorem improves (ii). {Theorem 1.5}. Let \(\Omega\) be an open set such that \(B_{1-2\rho(\Omega)}(0)\subset \Omega\subset B_1(0)\) with \(C^{2,\alpha}\) boundary for some \(\alpha>2s\). Then there is a constant \(\eta(n,s)>0\) such that if \(\eta_s(\Omega)\leq \eta(n,s)\), there exists a map \(F:\partial(B_1(0))\to \mathbb{R}^n\) of class \(C^{2,\tau}\) for any \(\tau<2s\) such that \(F\left(\partial(B_1(0))\right)=\partial\Omega\) and \[ \| F-\mathrm{Id}\|_{C^{2,\tau}(\partial B_1(0))}\leq C(\mu,s,\tau)\eta_s(\Omega). \] In particular, if \(\eta_s(\Omega)\) is sufficiently small, then \(\Omega\) is a convex set. Besides the moving frame method, to evaluate \(H_s^\Omega\) and related invariants, it is useful to express \(H_s^\Omega\) as a limit of integrals having smooth kernels. This is done in Lemma 2.1. Theorem 1.1 is proved by Lemma 2.1 and a geometric description of moving frames following [Zbl 0119.16603] in \S3. After showing that for \(\delta_s(\Omega)\) small, then up to a translation, all critical planes from the moving plane method pass close in the origin (Lemma 4.1), Theorems 1.2 and 1.5 are proved via delicate estimates of \(H_s^\Omega\) in \S4, the last section. The authors remark that the regularity criterion, crucial in the proof of Theorem 1.5 is implicitly contained in [\textit{M. C. Caputo} and \textit{N. Guillen}, ``Regularity for non-local almost minimal boundaries and applications'', Preprint, \url{arXiv:1003.2470}].