We will present a new massively parallel implementation of EOM-CCSD, built on top of TiledArray1 framework and implemented in the Massively Parallel Quantum Chemistry (MPQC4) package.2 The new implementation exhibits excellent parallel performance and allows calculation of excitation energy for systems with up to a hundred atoms and more than one thousand basis functions. To rapidly explore reduced scaling strategies applicable to excited state coupled-cluster methods we introduced representation of excited state eigenvectors in compressed tensor formats similar to those use in the ground state coupledcluster methods based on Pair-Natural Orbitals (PNOs).3,4 Due to the use of TiledArray’s sparse tensor data structures, it is also possible to explore the effect of domain approximations, local density, and other approximations to be featured in production linear scaling response coupled-cluster methods.5 Our presentation will focus primarily on how the tensor compression affects the EOM-CCSD excitation energies.

Exploration of reduced scaling approaches to EOM-CCSD

254th National Meeting (2017)

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