Convergence analysis of trigonometric methods for stiff second-order stochastic differential equations (Q415509)

The authors present an error analysis for a class of stochastic trigonometric schemes associated to the stiff second-order stochastic differential equation (SDE) \[ \ddot{X}_t + \frac{1}{\varepsilon^2}AX_t=g(X_t)+B\dot{W}_t \] where \(\varepsilon \ll1\) and \(A\) is a symmetric positive definite matrix. This equation is rewritten as a system of first order SDEs and it becomes possible to use the variation-of-constants formula. After discretising and choosing two filter functions they obtain a family of explicit trigonometric integrators. The proposed schemes permit the use of large step sizes and have uniform global error bounds in the position (i.e., independent of the large frequencies present in the SDE). In the final section, the authors consider two applications: the stochastic Fermi-Pasta-Ulam problem and a semi-linear stochastic wave equation. Both cases illustrate the robustness of the stochastic trigonometric numerical schemes.