A stochastic cellular automaton model of non-linear diffusion and diffusion with reaction (Q2277785)

A probabilistic cellular automaton model is investigated which yields an approximate solution of Fick's equation of diffusion \(du/dt=div({\mathcal D}\cdot \text{grad} u)\) either in the linear case (\({\mathcal D}\) constant) or in the nonlinear case where \({\mathcal D}\) depends on the position and on the solution. To reproduce deterministic macroscopic behaviour, the model uses the following simple microscopic probabilistic dynamics: each particle chooses at random one of its four neighbors on the square lattice and moves to the site it has targeted if this site is free and simply does not move if the site is already occupied. The emphasis is put on the synchronous implementation of this rules which allows a massively parallel algorithm. The authors discuss the deviation between this model and the discretized diffusion equation, and a similar study is done in the case of a reaction-diffusion process in a porous medium on the example of the carbonation in concrete.