A massively parallel implementation of the conventional and the Laplace transform based CCSD(T) energy has been developed in the Massively Parallel Quantum Chemistry (MPQC4) package.1 Just as the strongly-scalable CCSD-F12 implementation reported recently,2 upon which this work is based, the (T) correction demonstrates excellent scaling. Fora (H2O)18 cluster in the cc-pVDZ basis the (T) correction attains a 80% parallel efficiency on 32 nodes relative to a single 16-core node. The standard algorithm thus allows routine CCSD(T) treatment of systems with more than one hundred atoms and one thousand basis functions.3 To reduce the computational cost of the CCSD(T) to O(N6) we employ the Laplace transform formulation of the (T) correction.4 Only two to three quadrature points are sufficient for routine applications, except for small-gap problems such as non-covalent interactions of extended conjugated systems; nevertheless, the crossover with the standard (T) method are rapidly achieved even for modestly-sized systems.5 A massively parallel implementation of the LT-CCSD(T)-F12 method thus allows routine benchmark applications of the CCSD(T) method near the basis set limit for systems with more than one hundred atoms.

Extending the reach of the CCSD(T) method by massive parallelism and reduced scaling

254th National Meeting (2017)

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