Can cellular automata methods compete with Monte Carlo semiconductor device simulations?

A cellular-automaton (CA) approach for solving the Boltzmann equation is presented and applied to semiconductor device simulation. The comparison with Drift Diffusion and Monte Carlo (MC) algorithms shows the capabilities of the CA as a modelling tool, even in the presence of complicated geometries and hot carrier effects. With respect to the MC method, the CA exhibits a speed up of two orders of magnitude on scalar processors, coupled to a much more straightforward implementation on parallel processors.