On the conservativeness and the recurrence of symmetric jump-diffusions (Q1932183)

Let \((X,d,m)\) be a metric measure space such that each metric bounded ball in \(X\) is precompact and \(m\) is a Radon measure with full support. Consider a regular symmetric Dirichlet form \(({\mathcal E},{\mathcal F})\) in \(L^2(X,m)\) such that \[ {\mathcal E}(u,v)= \mu_{uv}(X)/2+ \iint_{x\neq y}(\widetilde u(x)-\widetilde u(y))(\widetilde v(x)- \widetilde v(y))J(dx, dy), \] where \(\mu\) is a bounded signed measure, absolutely continuous with respect to \(m\) with density \(\Gamma^{}(u,v)\), \(\widetilde u\) and \(\widetilde v\) are quasi-continuous versions of \(u,v\in{\mathcal F}_e\), \({\mathcal F}_e\) is the extended Dirichlet space of \(({\mathcal E},{\mathcal F})\), and the jump measure \(J\) has a symmetric kernel such that \(J(dx,dy)= j(x,dy)m(dx)= j(y,dx)m(dy)\). Define \[ \Gamma^{(j)}(u,v)(x):= \int_{x\neq y}(\widetilde u(x)-\widetilde u(y))(\widetilde v(x)-\widetilde v(y))j(x,dy), \] denote by \(\overline B(x,r)\) the closed metric ball with radius \(r\), centered at \(x\), and by \(V(x,r)\), \(V^{(c)}(x,r)\), and \(V^{(j)}(Ax,r)\) the volumes of \(\overline B(x,r)\), \(\overline B(x,r)\cap\{x\in X:\Gamma^{(c)}\neq 0\}\), and \(\overline B(x,r)\cap\{x\in X:\Gamma^{(j)}\neq 0\}\), respectively. Let \(\omega(r)\) be the supremum on \(\{x\in X:\Gamma^{(j)}\neq 0\}\) of \[ \int_{x\neq y}(d(x,y)\wedge r)^2 j(x,dy). \] Under some boundedness and topological consistency assumptions, the authors prove that the condition that \(\liminf_{r\to\infty}(\ln V(x_0, r)/r\ln r)<\infty\) for some \(x_0\in X\) is sufficient for the diffusion process associated with \(({\mathcal E},{\mathcal F})\) to be conservative, and the condition that \[ \liminf_{r\to\infty} r^{-2}(V^{(c)}(x_0, r)+ V^{(j)}(x_0, r)\omega(r)<\infty \] for some \(x_0\in X\) is sufficient for the process to be recurrent. Some examples illustrating the optimality of these conditions are given. Particular attention is paid to the case in which \(X\) is disconnected and the process can jump from one component to another.