A justification for the use of approximate transition amplitudes in semiclassical surface hopping

Abstract

Recent work has shown that a certain surface hopping form of the wave function is capable of obtaining highly accurate transition probabilities for nonadiabatic problems. It has also been found that it is necessary to include hops in classically forbidden regions in order to obtain this level of accuracy at low energies. The amplitude for the hops in this surface hopping expansion of the wave function has the typical p−1/2 semiclassical divergence at the turning points in the classical motion. While this singularity is an integrable divergence, the divergent behavior complicates the numerical evaluation of the integrals over hopping points that is present in the surface hopping expressions. Numerical evidence has shown that only small errors are incurred at most energies if these singular hopping amplitudes are replaced with a nonsingular approximation. This agreement is surprising, since the exact and approximate amplitudes differ greatly in the turning point region, and this region is expected to make important contributions to the transition probability at low energies. A numerical analysis is presented in this work that provides a justification as to why this numerically useful approximation works as well as it does.