A weak second-order split-step method for numerical simulations of stochastic differential equations (Q466817)

The authors introduce a split-step method for Monte Carlo simulation of stochastic differential equations \[ x_{t_0+h} = x_{t_0} + \int_{t_0}^{t_0+h}b(x(t)) dt + \int_{t_0}^{t_0+h}\sigma(x(t)) d\omega(t). \] It is based on a three-stage scheme, where the first and third steps are \(h/2\)-stage trapezoidal rules for the drift term and the second step is an \(h\)-stage of the martingale term. Combining the consecutive adjacent \(h/2\)-stages, the main result is the second-order weak accurance of this two-stage scheme. Numerical examples on linear complex oscillating systems illustrate the stability and the advantage of the trapezoidal rule.