Accurate and efficient evaluation of transition probabilities at unavoided crossings in ab initio multiple spawning

- Norm-preserving interpolation (NPI) used with ab initio multiple spawning (AIMS).
- NPI eliminates errors in transition probabilities due to discretization of time.
- AIMS-NPI accurately predicts transition probabilities at unavoided crossings.

Recently we introduced a norm-preserving wavefunction interpolation (NPI) method which allows the accurate and efficient integration of the time-dependent Schrödinger equation in regions of rapidly varying time-derivative coupling (TDC), e.g. near conical intersections (CIs) and effectively (N-1)-dimensional trivial unavoided crossings (TUCs). Herein we report the implementation of the NPI scheme into the ab initio multiple spawning (AIMS) method and investigate its behavior in two test systems: one which exhibits a CI and another which exhibits a TUC. In the CI case, AIMS-NPI predicts transfer probabilities that agree to within 0.6% of simulations based on the analytical evaluation of the nonadiabatic couplings, without the need for computationally expensive adaptive integration. In the TUC case, AIMS-NPI results in less than 0.08% error in the population transfer probability when a 0.12 fs time step is used, while simulations based on the analytical approach exhibit catastrophic errors for all finite time steps.