In this talk I will present our recent development of theory and programs for CASPT2 on-the-fly surface hopping dynamics, which has utilized a code generation approach. Our code allows for two-level parallelization of dynamics simulations: Each trajectory calculation (i.e., computation of gradients and nonadiabatic couplings) takes advantage of modern parallel computer hardware, in addition to trivial parallelization over trajectories. Technical challenges toward exascale computing will be discussed (e.g., memory-size aware DAG scheduler).

Parallel computing enables predictive on-the-fly photodynamics simulations

255th National Meeting (2018)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.