Molecular dice: Random number generators a la Boltzmann

We show that collective chaotic behavior emerging from the Boltzmann picture of gases provides an alternate conceptual framework to analyze and create apparent randomness on computers. For instance, at equilibrium the distribution of various physical quantities such as position or velocities or two-dimensional projection of kinetic energy can be used to generate uniform, Gaussian, and exponential distributions. Finally, based on these ideas we present a simplified but highly efficient mesoscopic Boltzmann-type algorithm to generate random distributions at a computational efficiency which is orders of magnitude higher than contemporary algorithms. We illustrate the effectiveness by considering two representative examples: spinodal decomposition of a binary alloy and rare event sampling in biological systems.