Subgeometric ergodicity for continuous-time Markov chains (Q972471)

In this paper the authors study subgeometric ergodicity for continuous-time Markov chains. Let \(X=(X_t)_{t\geq 0}\) be an ergodic Markov chain on a countable state space \(E\) with transition probabilities \(P^t(i,j)\) and stationary distribution \(\pi=(\pi_i)_{i\in E}\). Let \(r:[0,\infty)\to [1,\infty)\) be a subgeometric rate function (roughly speaking, \(r\) is increasing and satisfies \(\lim_{t\to \infty}(\log r(t))/t=0\)). Markov chain \(X\) is said to be subgeometrically ergodic if \[ \lim_{t\to \infty} r(t)\sum_{j\in E} |P^t(i,j)-\pi_j|=0\, \qquad \mathrm{for all } i \in E\, . \] The main result of the paper is a characterization of subgeometric ergodicity through three equivalent conditions: (i) an integral-type functional of the first hitting time has a finite moment; (ii) there exists a nonnegative drift function; and (iii) the subgeometric moment of the first hitting time has a finite expectation under the stationary distribution. Several examples illustrating the results are given.