The Direct-Product Decomposition Approach for Symmetry Exploitation in Many-Body Methods in Case of Non-Abelian Point Groups
Malte Hellmann, Jürgen Gauss
TL;DR
The paper tackles exploiting non-Abelian symmetry in correlated electronic structure calculations by extending the Abelian direct-product decomposition (DPD) approach to the $C_{3v}$ point group. It introduces a dual-representation strategy that uses both reduced and non-reduced representations for four-index quantities, with a dedicated reduction routine to handle transformations, spin-adaptation, and ${O(M^5)}$ contractions in CC methods. A pilot implementation, QUENA, demonstrates substantial computational savings for CCSD in NH$_3$ and PH$_3$ compared with both no symmetry and $C_s$ symmetry, indicating feasibility for large, highly symmetric molecules. These results establish a foundation for general non-Abelian symmetry exploitation in quantum chemistry and outline paths toward broader group coverage and extensions to perturbative triples (CCSD(T)).
Abstract
We demonstrate for the specific case of $C_{3v}$ how the direct-product decomposition scheme for the treatment of symmetry in coupled-cluster (CC) calculations can be extended to non-Abelian point groups. We show that for the two-electron integrals and CC amplitudes a block structure can be obtained by resolving the reducible products of two irreducible representations into their irreducible representations. To deal with the necessary resorts of the ordering of the two-electron integrals and amplitudes, spin-adaptation, and the O(M$^5$) contractions (with M as the number of basis functions) of a CC calculation, we suggest a strategy that uses both the reduced and non-reduced representation of the corresponding quantities and switches back and forth between them. While the reduced representations are the ones used in the O(M$^6$) contractions, the other steps are better carried out in the non-reduced representation. Our pilot implementation of the CC singles and doubles method confirms in test calculations for NH$_3$ and PH$_3$ using different basis sets that significant savings (of more than 20 compared to treatments without symmetry and about 5 compared to treatments using $C_s$ symmetry) are possible and suggest that the exploitation of non-Abelian symmetry would render CC computations on large highly symmetric molecules possible
