On preserving long-time features of a linear stochastic oscillator (Q878207)

A numerical method is proposed for solving the stochastic differential equation (SDE) \[ dX_t= a(t, X_t)\,dt+ \sum^m_{k=1} b^k(t, X_t)\,dW^k_t,\quad X_{t_0}= X_0, \] where \(a\), \(b^k\) are \(\mathbb{R}^d\)-valued functions and \(W_t\) is a standard Wiener process. The method is shown to have mean square order \({1\over 2}\) in general and order 1 for SDEs with additive noise. Then the method is applied to the linear stochastic oscillator \[ \ddot X_t+ X_t= \alpha\dot W_t, \] where \(\alpha\) is a positive constant. The resulting approximate solution is shown to have linear growth of the second moment and to oscillate infinitely often.