Two-sided eigenvalue estimates for subordinate processes in domains (Q2566077)

Let \(X=(X_t)_{t\geq 0}\) be a symmetric Markov process on \(E\), \(D\) an open subset of \(E\) and \(S=(S_t)_{t\geq 0}\) an independent subordinator with Laplace exponent \(\varphi\). Let \(X^\varphi=(X_{S_t})\) and \((X^D)^\varphi=(X^D_{S_t})\) be the subordinate processes of \(X\) and the subprocess \(X^D\) on \(D\), respectively. Under the hypothesis of the existence of the eigenvalues \(\{-\mu_n,n\geq 1\}\) and \(\{-\lambda_n,n\geq 1\}\) of the generators of \((X^\varphi)^D\) and \(X^D\), respectively, the authors give their relations. More precisely, assume that \(\varphi\) is a complete Bernstein function and the eigenvalues are arranged in decreasing order. Then, by using the relations of their Dirichlet forms, they prove that \(\mu_n\leq \varphi(\lambda_n)\). Furthermore, when \(X\) is a spherically symmetric \(\alpha\)-stable process in \(R^d\) with index \(\alpha\in (0,2]\) and \(D\) satisfies the exterior cone condition, it is proved that there exists a constant \(C\in (0,1)\) such that \((1-C)\varphi(\lambda_n)\leq \mu_n\leq \varphi(\lambda_n)\). In particular, if \(D\) is a bounded convex domain in \(R^d\), \(C\) can be taken as \(C={1\over 2}\). The cases that \(X\) is a Brownian motion in \(R^d\), \(S\) is an \(\alpha/2\)-subordinator with \(\alpha \in(0,2)\) and \(D\) is a bounded domain satisfying exterior cone condition or bounded convex domain are given as examples.