Central limit theorems and diffusion approximations for multiscale Markov chain models (Q2448694)

Based on the authors' abstract and introduction: Ordinary differential equations obtained as limits of Markov processes may arise by scaling large systems, i.e., by considering a sequence \(X^N\) of continuous-time Markov chains that converges to a deterministic function \(X\) and gives a distributional limit for the rescaled deviations \(r_N(X^N-X)\), or by averaging rapidly fluctuating systems, i.e., by considering an ergodic Markov process \(Y\) with stationary distribution \(\pi\) so that \[ Z^N(t)={1 \over\sqrt{N}}\int_0^{Nt}h(Y(s))\,ds=\sqrt{N}\int_0^th(Y(Ns))\,ds \] has a distributional limit for \(h\) satisfying \(\int h\,d\pi=0\), or in systems involving multiple time-scales, by a combination of the two. Motivated by models with multiple time-scales arising in systems biology, the authors develop a general approach to prove a central limit theorem by an application of martingale central limit theory. This central limit theorem provides a method for deriving diffusion approximations. In the second part of the paper, the authors study multi-scale, continuous time Markov chains of a type that arises naturally in models of chemical reaction networks. Three examples from systems biology are discussed in detail.