Externally corrected CCSD with renormalized perturbative triples (R-ecCCSD(T)) and density matrix renormalization group and selected configuration interaction external sources
Seunghoon Lee, Huanchen Zhai, Sandeep Sharma, Cyrus J. Umrigar, Garnet, Kin-Lic Chan

TL;DR
This paper enhances the ecCCSD method by integrating renormalized perturbative triples corrections using external sources like DMRG and HCI, improving accuracy for molecular potential energy surfaces without instability issues.
Contribution
It introduces a combined approach of ecCCSD with renormalized triples correction using external wavefunctions from DMRG and HCI, demonstrating improved accuracy and stability.
Findings
Triples correction significantly improves ecCCSD results.
No instability observed in the renormalized triples during dissociation.
Approximate wavefunctions can serve as efficient external sources.
Abstract
We investigate the renormalized perturbative triples correction together with the externally corrected coupled-cluster singles and doubles (ecCCSD) method. We take the density matrix renormalization group (DMRG) and heatbath CI (HCI) as external sources for the ecCCSD equations. The accuracy is assessed for the potential energy surfaces of H2O, N2, and F2. We find that the triples correction significantly improves on ecCCSD and we do not see any instability of the renormalized triples with respect to dissociation. We explore how to balance the cost of computing the external source amplitudes with respect to the accuracy of the subsequent CC calculation. In this context, we find that very approximate wavefunctions (and their large amplitudes) serve as an efficient and accurate external source. Finally, we characterize the domain of correlation treatable using the externally corrected…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
