1: \begin{abstract}

2: Langevin simulation provides an effective way to study collisional

3: effects in beams by reducing the six-dimensional Fokker-Planck

4: equation to a group of stochastic ordinary differential equations.

5: These resulting equations usually have multiplicative noise since the

6: diffusion coefficients in these equations are functions of position

7: and time.  Conventional algorithms, e.g. Euler and Heun, give only

8: first order convergence of moments in a finite time interval.

9: In this paper, a stochastic leap-frog algorithm for the numerical

10: integration of Langevin stochastic differential equations with

11: multiplicative noise is proposed and tested. The algorithm has a

12: second-order convergence of moments in a finite time interval and

13: requires the sampling of only one uniformly distributed random variable

14: per time step. As an example,

15: we apply the new algorithm to the study of a mechanical oscillator

16: with multiplicative noise.

17: \end{abstract}

18: