Characterization of bistability for stochastic multistep methods (Q766224)

Bistable numerical methods are studied for stochastic ordinary differential equations of the form \[ dX(t)= b^0(t, X(t))\,dt+ \sum^m_{r=1} b^r(t, X(t))\,dW^r(t),\quad t\in [0,T],\;X(0)= X_0, \] where the processes \(W^r\) are independent standard Brownian motions. The connections between consistency, stability, the strong root condition, and convergence are established. General results for one-step methods and for multistep methods are derived and then applied to the stochastic theta method, higher-order Itô-Taylor schemes, and the backward differentiation formula BDF2-Maruyama method to prove strong convergence, to obtain upper and lower bounds for the error, and to determine the maximum order of convergence.