Invariant measure for the Markov process corresponding to a PDE system (Q2581167)

Let \((X^\varepsilon,Z^\varepsilon)\) be Markov process on \(R^d\times\{1, \dots,n_0\}\) such that \(dX^\varepsilon(t)=b(X^\varepsilon(t),Z^\varepsilon(t))dt+\sqrt{\varepsilon}dB(t)\), where \(B\) is a standard Brownian motion, and \(Z^\varepsilon\) has infinitesimal transition probabilities of the form \[ \mathbf{P}(Z^\varepsilon(t+h)=l\mid Z^\varepsilon(t)=k, X^\varepsilon(t)=x)= \begin{cases} q_{kl}(x)h+o(h), k\neq l,\\ 1+q_{kk}(x)h+o(h), k= l, \end{cases} h\downarrow 0, \] see [\textit{A.\ V.\ Skorokhod}, ``Asymptotic methods in the theory of stochastic differential equations'' (1989; Zbl 0695.60055)]. The author shows that the family of transition probabilities of \((X^\varepsilon,Z^\varepsilon)\) is Feller continuous and establishes the existence of an invariant measure \(\mu^\varepsilon\). For \(\mu^\varepsilon\), large deviations principle is proved. See also \textit{A.\ Eizenberg} and \textit{M. Freidlin} [Ann. Probab. 21, No.~2, 1015--1044 (1993; Zbl 0776.60037)].