Crossing estimates for symmetric Markov processes (Q5944097)

The authors establish a crossing estimate for Dirichlet functions along the paths of general symmetric Markov processes, recurrent or transient, over a finite and infinite time interval, for quasi-every starting point. More precisely, let \(E\) be homeomorphic to a Borel subset of a compact metric space, \({\mathcal B}(E)\) the Borel \(\sigma\)-algebra on \(E\), and \(m\) a \(\sigma\)-finite measure on \({\mathcal B}(E)\) with \(\text{supp}(m)=E\). Let \(X=(X_t, P_x)\) be an \(m\)-symmetric, irreducible right Markov process with state space \(E\), and let \(({\mathcal E}, {\mathcal F})\) be the corresponding symmetric quasi-regular Dirichlet form on \(L^2(E,m)\). Let also \(({\mathcal E}^{\text{ref}},{\mathcal F}^{\text{ref}})\) be the reflected Dirichlet space associated with \(({\mathcal E}, {\mathcal F})\). For \(f\in {\mathcal F}^{\text{ref}}\) let \(\widetilde{f}\) denote a quasi-continuous version of \(f\). For \(a,b \in {\mathbb R}\), \(a<b\), \(U_t\) denotes the number of crossings of the interval \([a,b]\) completed by the process \(s\mapsto \widetilde{f}(X_s)\) during the time interval \([0,t]\). The main result of the paper states that for any smooth measure \(\mu\) on \(E\) and any \(t>0\) \[ {\mathbf E}_{\mu} U_t \leq 2(b-a)^{-2}{\mathcal E}^{\text{ref}}(f,f) \|{\mathbf E}_{\cdot} C_t^{\mu}\|_{\infty} , \] where \(C^{\mu}\) is the positive continuous additive functional of \(X\) with Revuz measure \(\mu\), and \( \|{\mathbf E}_{\cdot} C_t^{\mu}\|_{\infty}\) is the \(L^{\infty}(E,m)\)-norm of \(x\mapsto {\mathbf E}_x C_t^{\mu}\). As a corollary, the authors obtain that in case of symmetrizing measure \(m\), \( {\mathbf E}_m U_t \leq 2t(b-a)^{-2}{\mathcal E}^{\text{ref}}(f,f)\), while in case that \(({\mathcal E}, {\mathcal F})\) is transient, \({\mathbf E}_{\mu} U_{\infty} \leq 2(b-a)^{-2} \|G\mu\|_{\infty}{\mathcal E}^{\text{ref}}(f,f)\) (here \(G\mu\) denotes the 0-potential of the measure \(\mu\)). In the last section of the paper examples for both continuous and discrete-time Markov chains and diffusions on finite and infinite-dimensional spaces are given.