On the strong maximum principle for nonlocal operators (Q2272944)

Let \(\Omega\) be a domain in \(\mathbb{R}^N\) and \(c, g \in L^\infty_{\text{loc}}(\Omega)\). The authors are concerned with maximum principles for (suitably defined) variational supersolutions of the equation \[ I u = c(x) u + g \quad \text{in }~ \Omega, \] where \(I\) is a nonlocal linear operator which can be defined poitwisely (for a suitable class of functions) as \[ I u(x) = P.V. \int_{\mathbb{R}^N} (u(x)-u(y)) J(x,y) \,dy. \] Here, \(J: \mathbb{R^N} \times \mathbb{R^N} \to [0,\infty]\) is a measurable kernel function satisfying only the following three assumptions: (J1) \(J(x,y)=J(y,x)\) for all \(x,y \in \mathbb{R}^N\), and \[ \sup_{x \in \mathbb{R}^N}\int_{\mathbb{R}^N} \min\{1,|x-y|^2\} J(x,y) \, dy < \infty; \] (J2) For every \(r>0\), the function \(z \mapsto \text{essinf} \{J(x,x \pm z): x \in \mathbb{R}^N\}\) does not vanish identically on \(B_r(0)\); (J3) There holds \[ \sup_{x \in \mathbb{R}^N}\int_{\mathbb{R}^N} \min\{1,|z|\} |J(x,x+z)-J(x,x-z)| \, dz < \infty. \] The main result of the article is the following strong maximum principle. Let \(u \in \mathcal{V}^j(\Omega)\) be a variational supersolution of the considered equation with \(u \geq 0\) in \(\mathbb{R}^N\), where \[ \mathcal{V}^j(\Omega):=\{u \in L^2_{\text{loc}}(\mathbb{R}^N):~\int_{\Omega}\int_{\mathbb{R}^N}(u(x)-u(y))^2 J(x,y) \, dxdy < \infty\}. \] Then either \(u\equiv 0\) a.e. in \(\Omega\), or \(u\) is strictly positive in \(\Omega\) in the sense that \(\text{essinf}_{K} u>0\) for every compact \(K \subset \Omega\). The usual fractional Laplacian is an example of an admissible operator. Moreover, the result extends also to the regional fractional Laplacian. Several related weak maximum principles are also obtained.