Volume 14, Issue 3

The paper describes the deviational particle Monte Carlo method for the Boltzmann equation. The approach is an application of the general “control variates” variance reduction technique to the problem of solving a nonlinear equation. The deviation of the solution from a reference Maxwellian is approximated by a system of positive and negative particles. Previous results from the literature are modified and extended. New algorithms are proposed that cover the nonlinear Boltzmann equation (instead of a linearized version) with a general interaction model (instead of hard spheres). The algorithms are obtained as procedures for generating trajectories of Markov jump processes. This provides the framework for deriving the limiting equations, when the number of particles tends to infinity. These equations reflect the influence of various numerical approximation parameters. Detailed simulation schemes are provided for the variable hard sphere interaction model.

Monte Carlo methods are now widely used in solving various computational fluid dynamics systems. This paper presents an improved scrambled quasi-Monte Carlo method for solving fluid dynamics applications. In our parallel implementation we use an independent scrambled quasirandom sequence for each processor. We explore the use of both scrambled quasi-Monte Carlo and variance reduction methods to improve the accuracy for Monte Carlo schemes. We also present theoretical analyses and numerical experiments to validate our numerical algorithms.