Small random perturbations of dynamical systems: Exponential loss of memory of the initial condition (Q1112455)

The authors consider two types of random perturbations of dynamical systems: the discrete time case of the type studied by \textit{P. Gora} [Z. Wahrscheinlichkeitstheor. Verw. Geb. 69, 137-160 (1985; Zbl 0551.60064)] when a Markov chain \(X_ t^{\epsilon}\) is obtained by composing randomly two maps T and \(T^{\epsilon}\) with \(T^{\epsilon}\) close to identity, and the continuous time case when \(X_ t^{\epsilon}\) is a solution of a stochastic differential equation, \[ dX_ t^{\epsilon}=b(X_ t^{\epsilon})dt+\epsilon dw_ t. \] In addition to a Ventzel'-Freidlin type condition that the limit set of the unperturbed system consists of a finite number of equivalence classes the authors assume that attractors consist only of a finite number of points. Denoting by \(X_ t^{\epsilon}(x)\) the process starting at x they prove that for x, y belonging to small neighbourhoods of attractors \(dist(X_ t^{\epsilon}(x),X_ t^{\epsilon}(y))\) tends to zero at an exponential rate as \(t\to \infty\) with probability close to one when \(\epsilon\) is small.