Almost sure exponential stability of numerical solutions for stochastic delay differential equations (Q981648)

A theorem is proved that gives sufficient conditions for almost sure exponential stability (ASES) of Euler-Maruyama method numerical solutions of the \(n\)-dimensional nonlinear stochastic delay differential equation \[ dx(t)= f(x(t), x(t-\tau),t)\,dt+ g(x(t), x(t-\tau), t)\,dw(t),\quad t\geq 0. \] A counterexample is presented to show that without the linear growth condition on \(f\) of the theorem, ASES may be lost. Then for the backward Euler-Maruyama method ASES is proved when a one-sided Lipschitz condition on \(f\) in \(x\) replaces the linear growth condition on \(f\).