Almost sure asymptotic stability of drift-implicit \(\theta\)-methods for bilinear ordinary stochastic differential equations in \(\mathbb R^1\) (Q1779415)

Asymptotic stability is investigated for numerical methods of the form \[ Y_{n+1}= Y_n+ \theta_n\alpha_{n+1} Y_{n+1}\Delta_n+ (1- \theta_n)\alpha_n Y_n\Delta_n+ \sigma_n Y_n\Delta W_n \] that have variable step size and are intended to be used to approximate solutions of nonautonomous Itô stochastic differential equations of the form \[ dX(t)= \alpha(t)X(t)dt+ a(t)X(t)dW)t), \] where \(W\) is a standard Wiener process. Theorems are proved establishing criteria guaranteeing that the trivial equilibrium solutions of the above equations are globally almost surely asymptotically stable. A definition of region of almost sure exponential stability for these numerical methods is given and discussed.