Development of efficient computational techniques and codes for second-order Moller–Plesset perturbation calculation of extended systems

摘要

We have developed the efficient computational techniques and codes that are suitable for the massively parallel calculations of second-order Moller–Plesset perturbation (MP2) calculations of extended systems on supercomputers. In this study, we present (1) the MPI/OpenMP hybrid parallel algorithm of the resolution-of-identity MP2 (RI-MP2) energy calculations, (2) the RI-MP2 method for fragment molecular orbital (FMO) method, and (3) the RI-MP2 method for divide-and-conquer (DC) method. We present outlines of these methods and some examples of massively parallel MP2 calculations of extended systems on supercomputers using the codes developed in this study. The codes presented in this article have been supplied to users of K computer as library programs.