Abrupt convergence for stochastic small perturbations of one dimensional dynamical systems (Q288202)

The authors study the cut-off phenomenon that appears in the asymptotic behavior for a family of stochastic small perturbations of a one dimensional dynamical system. This phenomenon is known to describe abrupt convergence in some random walks, Markov chains and Markov processes. Under regularity and coercivity hypotheses on the potential, they prove that the family of differential equations perturbed by a small Brownian motion presents a profile cut-off phenomenon with respect to the total variation distance. The paper ends with a question of metastability when the potential has two wells of different depths, leading to two statistically different regimes depending on the respective length of the time horizon and the exit time of the shallow well. Then a kind of local cut-off phenomenon is observed.