Computational aspects of multi-species lattice-gas automata

Abstract

We present computational aspects of a parallel implementation of a multi-species thermal lattice gas. This model, which can be used to simulate reaction-diffusion phenomena in a mixture of different fluids, is analyzed for a fluid system at global equilibrium. Large system sizes combined with long-time simulation makes parallelization a necessity. We show that the model can be easily parallelized, and possesses good scalability. Profiling information shows the random number generator has become a bottleneck. The model can be statistically analyzed by calculating the dynamic structure factor S(k ω). As an illustration, we measure S(k, ω) for a one-component system, and extract the values of transport coefficients from the spectra. Finally, S(k, ω) is shown for a two-component thermal model, where the central peak is more complicated, due to the coupled entropy-concentration fluctuations.