Prospects for rank-reduced CCSD(T) in the context of high-accuracy thermochemistry.

This article focuses on developing and assessing five novel approximation schemes for the perturbative triples correction [(T)] in coupled-cluster single double triple [CCSD(T)] calculations, a key component in high-accuracy computational thermochemistry. Using a subset of molecules from the W4-17 dataset, the study combines density fitting of electron repulsion integrals, rank-reduced compression of T2 amplitudes, Tucker-3 compression of T3 intermediates, and either Laplace transform or orthogonal rotation of orbital energy denominators to reduce the computational scaling from O(N⁷) to O(N⁶). All five methods achieve sub-0.1 kJ mol⁻¹ mean absolute errors relative to canonical density-fitted CCSD(T), with the approximation labeled Z̃T offering the best balance of accuracy and computational efficiency. These findings suggest that such reduced-scaling techniques can significantly lower the cost of CCSD(T) calculations in thermochemical modeling while maintaining sub-chemical accuracy, though further work is needed to optimize projector selection for individual reactions.