Ergodic degrees for continuous-time Markov chains (Q1777538)

Let \(\{X_t, t\geq 0\}\) be a continuous-time and irreducible Markov chain on a countable state space \(E\) with transition matrix \(| | p_{ij}(t)| | , i,j\in E\). Assuming the stationary distribution \(\{\pi_j, j\in E\}\) exists, the paper studies the rate of convergence of \(p_{ij}(t)-\pi_j\) to zero in usual and in an integral sense. The main result (Theorem 1.3) states that given \(l\geq 0\) the integrals \(\int_0^{\infty}t^l(p_{ij}(t)-\pi_j)dt\) are finite iff \(E^j\sigma_j^{l+2}<\infty\) for some \(j\in E\), where \(\sigma_j\) is the return time to state \(j\). Also the author points out two equivalent conditions which ensure that, given integer \(l\geq 0\), \(E^j\sigma_j^l<\infty\). To illustrate the results the author gives examples where explicit calculation of the rate of convergence is possible. Special attention is given to birth and death processes.