Evaluation of the coupled-cluster correction term, δCCSD(T)MP2, calculated using small valence basis sets

In the focal-point estimation of the coupled cluster with singles, doubles, and perturbative triples [CCSD(T)] interaction energies followed by extrapolation to the complete basis set (CBS) limit, the most critical and challenging step is the rapid and accurate calculation of the coupled-cluster correction term, δMP2CCSDT, which represents the difference between CCSD(T) and MP2 interaction energies at a small basis set. In this study, using calculations with the widely used aug-cc-pVDZ (aVDZ) basis set as the benchmark, we primarily evaluated the reliability of calculating δMP2CCSDT with the much smaller valence basis set vDZP and its diffuse-function-augmented version vDZPD. The results show that with the vDZP basis set, the mean MAD and RMSD of δMP2CCSDT across five benchmark datasets for noncovalent interactions (S22, S66, X40, HSG, and NCIBLIND10) are 0.14 and 0.18 kcal/mol, respectively, indicating relatively large deviations. In contrast, the vDZPD basis set yields much lower average MAD and RMSD of only 0.02 and 0.04 kcal/mol, demonstrating high computational accuracy. Using the vDZPD basis set to calculate noncovalent systems containing around 30-50 atoms, the δMP2CCSDT computation time is saved by about 40%. Furthermore, the results show that the δMP2CCSDT/vDZPD corrections still maintain high accuracy in calculating metal-containing complexes. The CCSD(T)/CBS interaction energy benchmark can be pushed to the hundred-atom scale with vDZPD basis set, holding significant promise for widespread application in the study of large noncovalent systems.