Weak convergence on the first exit time of randomly perturbed dynamical systems with a repulsive equilibrium point (Q1375410)

The following stochastic differential equation \[ dX^{\varepsilon}(t,x)=b(X^{\varepsilon}(t,x))dt+ \varepsilon^{1/2}\sigma((X^{\varepsilon}(t,x))dW(t),\qquad X^{\varepsilon}(0,x)=x, \tag{1} \] is investigated. Here vector-function \(b(\cdot )\) and matrix-function \(\sigma (\cdot )\) are bounded and globally Lipschitz continuous, \(W(t)\) is Wiener process. \(X^{\varepsilon}(t,x)\) is considered as the small perturbation of \(X^0(t,x)\) for small \(\varepsilon \). It is shown that the first exit times of small random perturbations of the dynamical system (described by (1)) from a bounded domain \(D \in {\mathbb{R}}^d\) weakly converge to the life time of an explosive diffusion process. Moreover the mean first exit times converge to the mean explosion time, as random perturbations disappear, when they are appropriately scaled. The case when \(D\) contains only one equilibrium point \(o\) of the dynamical system, \(o\) is polynomially instable and repulsive, is considered.