Transition times and stochastic resonance for multidimensional diffusions with time periodic drift: a large deviations approach (Q997407): Difference between revisions

Motivated by applications in climate models the authors study \textit{stochastic resonance} for multidimensional diffusions with weakly periodic drift of the form \(dX_t^{\varepsilon}=b(t/T^{\varepsilon},X_t^{\varepsilon})d t+\sqrt{\varepsilon}d W_t\), where \(W\) is d-dimensional Brownian motion, \(b\) is periodic in its first argument and \(T^{\varepsilon}\) is exponentially large in \(\varepsilon^{-1}\). The deterministic system (without the \(d W\) term) with frozen first argument of \(b\) is assumed to possess two stable fixed points which do not depend on the first argument of \(b\). Under appropriate conditions, the authors are able to carefully control the exit times of the solution from a neighborhood of one of the two metastable states to a neighborhood of the other one using refined large deviations estimates. Based on this result they suggest a quality measure for stochastic resonance, i.e. periodic switching between the metastable states in the limit \(\varepsilon \to 0\).