New and Efficient Implementation of CC3
Author(s) -
Alexander C. Paul,
Rolf H. Myhre,
Henrik Koch
Publication year - 2020
Publication title -
journal of chemical theory and computation
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.001
H-Index - 185
eISSN - 1549-9626
pISSN - 1549-9618
DOI - 10.1021/acs.jctc.0c00686
Subject(s) - transpose , jacobian matrix and determinant , coupled cluster , atomic orbital , ground state , excited state , computer science , valence (chemistry) , algorithm , parallel computing , computational science , atomic physics , physics , mathematics , eigenvalues and eigenvectors , quantum mechanics , molecule , electron
We present a new and efficient implementation of the closed shell coupled cluster singles and doubles with perturbative triples method (CC3) in the electronic structure program e T . Asymptotically, a ground state calculation has an iterative cost of 4 n V 4 n O 3 floating point operations (FLOP), where n V and n O are the number of virtual and occupied orbitals, respectively. The Jacobian and transpose Jacobian transformations, required to iteratively solve for excitation energies and transition moments, both require 8 n V 4 n O 3 FLOP. We have also implemented equation of motion (EOM) transition moments for CC3. The EOM transition densities require recalculation of triples amplitudes, as n V 3 n O 3 tensors are not stored in memory. This results in a noniterative computational cost of 10 n V 4 n O 3 FLOP for the ground state density and 26 n V 4 n O 3 FLOP per state for the transition densities. The code is compared to the CC3 implementations in CFOUR, DALTON, and PSI4. We demonstrate the capabilities of our implementation by calculating valence and core excited states of l-proline.
Accelerating Research
Robert Robinson Avenue,
Oxford Science Park, Oxford
OX4 4GP, United Kingdom
Address
John Eccles HouseRobert Robinson Avenue,
Oxford Science Park, Oxford
OX4 4GP, United Kingdom