Determination of the spectral index of ergodicity of a birth-and-death process (Q2722124)

Let us consider a Markov birth-death process \(X_{t}\) with discrete phase space \(E\) and continuous time. It is defined by its intensities of birth \(\lambda_{n}\) and death \(\mu_{n}\) in the state \(n.\) Let \(Q\) be its infinitesimal matrix, \(Q=\{q_{ij}\); \(i,j \in E\},\) and let \(P(t)=\{p_{ij}(t)\); \(i,j \in E\}\) be a matrix of transition probabilities, and \(\Pi=\{\pi_{i}\); \(i\in E\}\) be a matrix of stationary probabilities. The nonpositive index of ergodicity NEWLINE\[NEWLINEe(G)=\sup_{i,j}\lim_{t\to+\infty}(1/t)\log|p_{ij}(t)-\pi_{j}|NEWLINE\]NEWLINE as a function of matrix \(G\) is studied. \textit{E. van Doorn} [``Stochastic monotonicity and queueing applications of birth-death processes'' (1980; Zbl 0454.60069)] showed that for the exponential ergodic birth-death process the index \(e(G)\) coincides with the spectral index of ergodicity NEWLINE\[NEWLINE\alpha(G)=\sup_{t\geq 0}(1/t)\log\|P(t)-\Pi\|.NEWLINE\]NEWLINE A new explicit relation for the calculation of the spectral index of ergodicity of birth-death process with continuous time is considered. The calculation of the index is reduced to the solution of the optimization nonlinear programming problem which contains the infinitesimal matrix of the process. As an example, the new method for finding exact values of indices of exponential ergodicity for some Markov queueing systems is proposed.