Mapping Coriolis-coupled quantum dynamics onto parallel computer architectures

Abstract

We describe how quantum dynamics calculations with large total angular momentum quantum numbers J can be efficiently carried out on scalable parallel computers. In our approach, different processors i are associated with different allowed values of Ω, the projection quantum number of total angular momentum on a body-fixed axis. Ω can take on values Ω = 0, 1,…, J. One strategy involves associating processors i = 0, 1,…, J directly with the corresponding Ω values. Coriolis coupling permits only Ω ± 1 coupling, which implies only nearest neighbor processors need to communicate. This “Coriolis-coupled” parallel model and certain variations are applied to a triatomic unimolecular fragmentation process. Due to the necessity of communication between processors and load balancing issues, our model does not scale ideally with J. However, for the example considered, one can achieve wall clock times for reasonably high J values that are within about a factor of two of J = 0 calculations.