Basis set convergence and extrapolation of connected triple excitation contributions (T) in computational thermochemistry: the W4-17 benchmark with up to k functions

TL;DR

This study investigates the basis set convergence of the connected triple excitations contribution (T) in thermochemical calculations, demonstrating that extrapolation techniques can achieve high accuracy with manageable basis set sizes.

Contribution

It provides a detailed analysis of basis set convergence for the (T) contribution and proposes effective extrapolation methods to improve computational accuracy in thermochemistry.

Findings

Extrapolation from quintuple- and sextuple-zeta basis sets yields ~0.004 kcal/mol RMS accuracy.

Convergence to ~0.01 kcal/mol is possible with quadruple- and quintuple-zeta basis sets.

Radial flexibility in basis sets is more crucial than angular momentum expansion.

Abstract

The total atomization energy of a molecule is the thermochemical cognate of the heat of formation in the gas phase, its most fundamental thermochemical property. We decompose it into different components and provide a survey of them. It emerges that the connected triple excitations contribution is the third most important one, about an order of magnitude less important than the "big two" contributions (mean-field Hartree-Fock and valence CCSD correlation), but 1-2 orders of magnitude more important than the remainder. For the 200 total atomization energies of small molecules in the W4-17 benchmark, we have investigated the basis set convergence of the connected triple excitations contribution (T). Achieving basis set convergence for the valence triple excitations energy is much easier than for the valence singles and doubles correlation energy. Using reference data obtained from spdfghi and spdfghik basis sets, we show that extrapolation from quintuple-zeta and sextuple-zeta yields values within about 0.004 kcal/mol RMS. Convergence to within about 0.01 kcal/mol is achievable with quadruple- and quintuple-zeta basis sets, and to within about 0.05 kcal/mol with triple- and quadruple-zeta basis sets. It appears that radial flexibility in the basis set is more important here than adding angular momenta L: apparently, replacing nZaPa basis sets with truncations of 7ZaPa at L=n gains about one angular momentum for small values of n. We end the article with a brief outlook for the future of accurate electronic structure calculations.