Title:Addressing the gas kinetics Boltzmann equation with branching paths statistics

Abstract:Several works have shown that it is possible to conceive non-linear Monte-Carlo algorithms, but these works have remained overlooked so far. It is commonly admitted that in non-linear physics it is necessary to start with a linearization step before then calling on the linear Monte-Carlo method. We show here how it is possible to avoid this linearization step in the particular case of the Boltzmann equation. Our method retains the exact character of the Monte-Carlo method (unbiased estimation with a reliable evaluation of its uncertainty). We also extend to non-linear transport a very strong property of Monte-Carlo algorithms in linear transport: the ability to perform probe calculus (evaluate a quantity at a given position and time without computing all the field at previous times). Another striking feature of this algorithm is its ability to focus on rare events, however scarce they may be, including very high energy events that are important from the applicative point of view.