The composite Milstein methods for the numerical solution of Itô stochastic differential equations (Q629486)

Three composite Milstein methods are proposed for approximating strong solutions of the Itô stochastic differential equation (SDE) \[ dy(t)= f(t, y(t))\,dt+ \sum^d_{j=1} g_j(t, y(t))\,dW_j(t), \] where the \(W_j(t)\) are standard Wiener processes. Starting with the case where \(d=1\), the three methods are constructed by combining a semi-implicit Milstein method and an implicit Milstein method. For a linear test equation, the three methods are shown to converge to the exact solution and then mean-square stability, \(T\)-stability, and \(T(A)\)-stability are examined. Summaries of numerical results for a linear SDE and two nonlinear SDEs are provided which suggest that the proposed methods have better accuracy than a known composite Euler method. Then for the case where \(d=2\) similarly favorable stability properties are reported for a linear test equation and for a Duffing-Van der Pol oscillator driven by white noise.