ABSTRACT
Second-order Møller-Plesset (MP2) perturbation theory is among the simplest and most widely used quantum chemical methods for incorporating electron correlation. Although MP2 theory is computationally less expensive than most other correlated electronic structure methods, its computational cost scales steeply with the molecular size. For conventional implementations of MP2 theory, which are based on canonical orbitals obtained in a HartreeFock calculation, the scaling is formally O(N5), where N is the size of the molecule. Parallel implementation of MP2 theory, therefore, is an important step towards extending the applicability of the method to larger molecular systems. Another potential bottleneck in applications of conventional MP2 theory is the rather large storage requirement, which arises in the transformation of the two-electron integrals from the atomic orbital to the molecular orbital basis. To mitigate this bottleneck, a parallel implementation of MP2 theory should use a distributed data model in which the integral arrays are distributed across processes.
