On Poisson equation and diffusion approximation. II. (Q1431482)

[For part I see ibid. 29, 1061--1085 (2001; Zbl 1029.60053).] The final goal of the authors is to study the limt as \(\varepsilon\to 0\) of the slow-fast stochastic differential system \[ \begin{aligned} dX_t^\varepsilon&= \varepsilon^{-2}b(X_t^\varepsilon, Y_t^\varepsilon)dt+ \varepsilon^{-1}\sigma(X_t^\varepsilon,Y_t^\varepsilon)dB_t,\\ dY_t^\varepsilon&= F(X_t^\varepsilon,Y_t^\varepsilon)dt+ \varepsilon^{-1} G(X_t^\varepsilon,Y_t^\varepsilon)dt+ H(X_t^\varepsilon,Y_t^\varepsilon)dB_t. \end{aligned} \] In order to do so, they investigate in detail the parameter-dependent equation \[ dX_t^y=b(X_t^y)dt+\sigma(X_t^y,y)dB_t. \] This kind of equation is asymptotically obtained by the time-change \(t\mapsto \varepsilon^2 t\) in the equation for \(X_t^\varepsilon\). Based on analytic results for Poisson equations they obtain precise regularity and growth bounds for the corresponding invariant density and transition probability density (Green's function) as well as for their parameter-dependence.