Implementation of relativistic coupled cluster theory for massively parallel GPU-accelerated computing architectures

Abstract

In this paper, we report a reimplementation of the core algorithms of relativistic coupled cluster theory aimed at modern heterogeneous high-performance computational infrastructures. The code is designed for efficient parallel execution on many compute nodes with optional GPU coprocessing, accomplished via the new ExaTENSOR back end. The resulting ExaCorr module is primarily intended for calculations of molecules with one or more heavy elements, as relativistic effects on electronic structure are included from the outset. In the current work, we thereby focus on exact 2component methods and demonstrate the accuracy and performance of the software. The module can be used as a stand-alone program requiring a set of molecular orbital coefficients as starting point, but is also interfaced to the DIRAC program that can be used to generate these. We therefore also briefly discuss an improvement of the parallel computing aspects of the relativistic self-consistent field algorithm of the DIRAC program.

Towards highly accurate calculations of parity violation in chiral molecules: relativistic coupled-cluster theory including QED-effects

“…Improvements include: i) Property-optimized basis sets, which go beyond conventional energy-optimized basis sets. ii) We have used a new relativistic coupled cluster code, ExaCorr, geared towards massively parallel calculations [63], allowing extensive calibration studies in order to elaborate a CCSD computational protocol for PV energies and to include H 2 Po 2 among the target molecules iii) Inclusion of QED-effects through effective potentials.…”

“…All calculations employed the molecular-meanfield X2C Hamiltonian ( 2 DC M ) [105] based on the Dirac-Coulomb Hamiltonian. For technical reasons, MP2 and CCSD analytical expectation values were calculated using the RELCCSD [62,136] and ExaCorr [63] modules, respectively. The expectation values were calculated using the Lagrangian-based analytical energy derivative technique described in Ref.…”

“…In the present work we repeat the calculations of Shee et al [62], but with a number of improvements: i) The coupled cluster calculations have been carried out using a new coupled cluster code, ExaCorr, implemented for massively parallel GPU-accelerated computing architectures [63]. ii) We have extended the Dyall-type all-electron basis sets with tight exponents optimized for PV calculations.…”