Title:A computable branching process for the Wigner quantum dynamics

Abstract:A branching process treatment for the nonlocal Wigner pseudo-differential operator and its numerical applications in quantum dynamics is proposed and analyzed. We start from the discussion on two typical truncations of the nonlocal term, i.e., the $k$-truncated and $y$-truncated models. After introducing an auxiliary function $\gamma(\bm{x})$, the (truncated) Wigner equation is reformulated into the integral formulation as well as its adjoint correspondence, both of which can be regarded as the renewal-type equations and have transparent stochastic interpretation. We prove that the moment of a branching process happens to be the solution for the adjoint equation, which connects rigorously the Wigner quantum dynamics to the stochastic branching process, and thus a sound mathematical framework for the Wigner Monte Carlo methods is established. Within the framework, the branching process for the $y$-truncated model recovers the popular signed particle Monte Carlo method which needs a discretization of the momentum space beforehand, whereas the $k$-truncated branching method allows a continuous momentum sampling. More interestingly, with the help of the rigorous connection, we find that a constant auxiliary function $\gamma(\bm{x})\equiv\gamma_0$ performs better from the point of view of both theoretical and numerical aspects. In theory, we obtain a simple but analytical calculation formula for the growth rate of particle number, i.e., $\me^{2\gamma_0 t}$ with $t$ being the time. In numerics, the constant auxiliary function may facilitate the random sampling from the exponential distribution and improve the accuracy. Typical numerical experiments on the Gaussian barrier scattering validate our theoretical findings, as well as demonstrate the accuracy, the efficiency and thus the computability of the Wigner branching process.