Stable solutions to semilinear elliptic equations are smooth up to dimension \(9\) (Q2663749)

In this very interesting paper the authors give a full proof of the following long-standing conjecture. Assume \(\Omega\subset\mathbb{R}^n\) is a bounded domain, \(f:\mathbb{R}\rightarrow\mathbb{R}\) is positive, non-decreasing, convex, superlinear at \(+\infty\), and \(n\le 9\). If \(u\in W^{1,2}_0(\Omega)\) is a stable solution to \[ -\Delta u=f(u) \text{ in } \Omega, \] then \(u\) is bounded. This conjecture has been the topic of intense research in the past decades, having been proved to hold under various additional conditions: [\textit{G. Nedev}, C. R. Acad. Sci., Paris, Sér. I, Math. 330, No. 11, 997--1002 (2000; Zbl 0955.35029); the first author and \textit{A. Capella}, J. Funct. Anal. 238, No. 2, 709--733 (2006; Zbl 1130.35050); the first author, Commun. Pure Appl. Math. 63, No. 10, 1362--1380 (2010; Zbl 1198.35094); the first author, Discrete Contin. Dyn. Syst. 39, No. 12, 7249--7264 (2019; Zbl 1425.35095); \textit{S. Villegas}, Adv. Math. 235, 126--133 (2013; Zbl 1479.35471); the first and the third author, Commun. Partial Differ. Equations 38, No. 1--3, 135--154 (2013; Zbl 1268.35059); \textit{J. Spruck} et al., Discrete Contin. Dyn. Syst. 36, No. 2, 601--609 (2016; Zbl 1326.35141)]. The authors of this paper conclude this study. To that effect, they first obtain interior boundedness of solutions without requiring convexity or monotonicity of \(f\), generalizing a result known previously only for \(n\le 4\) [the first author, Commun. Pure Appl. Math. 63, No. 10, 1362--1380 (2010; Zbl 1198.35094)]. Furthermore, when \(n\le 9\), both in the interior and global settings, the authors prove that \(W^{1,2}\) stable solutions are bounded in terms of their \(L^1\) norms, with a constant that does not depend on \(f\). More precisely, the first main result of the paper provides a universal interior a-priori bound on the \(C^{\alpha}\) norm of stable solutions. Assume \(u\in C^2(B_1)\) is a stable solution of \(-\Delta u=f(u) \text{ in } B_1\) with \(n\le 9\), and \(f\) is locally Lipschitz and non-negative. Then \[ ||u||_{C^{\alpha}(\overline{B}_{1/2})}\le C||u||_{L^1(B_1)},\] where \(\alpha=\alpha(n)>0\) and \(C=C(n)\). Combining this result with the moving planes method, the authors obtain a universal bound on \(u\) when \(\Omega\) is convex: assume \(n\le 9\), \(\Omega\subset\mathbb{R}^n\) is a bounded, convex and \(C^1\) domain, and \(f:\mathbb{R}\rightarrow\mathbb{R}\) is locally Lipschitz, and non-negative. If \(u\in C(\overline{\Omega})\cap C^2(\Omega)\) is a stable solution of \[ -\Delta u=f(u) \text{ in } \Omega, \ \ \ u=0 \text{ on } \partial\Omega, \] then there exists \(C=C(\Omega)\) such that \[||u||_{L^{\infty}(\Omega)}\le C||u||_{L^1(\Omega)}. \] The second main result from this paper addresses the global regularity of stable solutions in \(C^3\) domains, when the non-linearity is convex and non-decreasing. The authors obtain a-priori estimates working with classical solutions: let \(\Omega\subset\mathbb{R}^n\) be a bounded domain of class \(C^3\), and \(f:\mathbb{R}\rightarrow\mathbb{R}\) be non-negative, non-decreasing and convex. If \(u\in C(\overline{\Omega})\cap C^2(\Omega)\) is a stable solution of \[ -\Delta u=f(u) \text{ in } \Omega, \ \ \ u=0 \text{ on } \partial\Omega, \] then \[||\nabla u||_{L^{2+\gamma}(\Omega)}\le C||u||_{L^1(\Omega)}, \] where \(\gamma=\gamma(n)>0\) and \(C=C(\Omega)\). Moreover, if \(n\le 9\), there exist \(\alpha =\alpha(n), C=C(\Omega)\) such that \[||u||_{C^{\alpha}(\overline{\Omega})}\le C||u||_{L^1(\Omega)}. \] This result completely solves two open problems posed by \textit{H. Brezis} [in: Morse theory, minimax theory and their applications to nonlinear differential equations. Proceedings of the workshop, Beijing, China, April 1--September 30, 1999. Somerville, MA: International Press. 23--33 (2003; Zbl 1200.35144)] and \textit{H. Brézis} and \textit{J. L. Vázquez} [Rev. Mat. Univ. Complutense Madr. 10, No. 2, 443--469 (1997; Zbl 0894.35038)]. As a consequence of the above estimates, the authors prove the long-standing conjecture described initially. The key ingredient is to prove the bounds from the second main result for classical solutions. Having proved this, an approximation argument [\textit{L. Dupaigne}, Stable solutions of elliptic partial differential equations. Boca Raton, FL: CRC Press (2011; Zbl 1228.35004)] allows one to conclude that the same bounds hold for every \(W^{1,2}_0(\Omega)\) stable weak solution \(u\). Regarding the techniques employed, a key ingredient is the choice of test function in the stability inequality. The authors choose \(\xi=(x\cdot\nabla u)|x|^{(2-n)/2}\zeta\), where \(0\le\zeta\le 1\) is a smooth cut-off function, equal to \(1\) in a ball and vanishing outside a slightly larger ball. With such choice, they are able to show that the radial derivative of \(u\) in a ball is controlled by the total gradient in an annulus. As this is not enough to obtain \(L^{\infty}\) bounds for general solutions, the authors use stability once more, showing that the radial derivative and the total derivative have comparable size in \(L^2\) (when the integral of \(|\nabla u|^2\) on balls satisfies a doubling property). This relies on a delicate compactness argument that, in turn, relies on a-priori estimates. Regarding the boundary estimate, the authors are able to prove that whenever the non-linearity is convex and non-decreasing, the class of stable solutions is closed under \(L^1_{\mathrm{loc}}\) convergence. This allows one to reduce oneself to a flat-boundary configuration, control the gradient by its radial component, and finish the second main result.