Basis Set Limit of CCSD(T) Energies: Explicit Correlation Versus Density-Based Basis-Set Correction

A thorough comparison is carried out for explicitly correlated
and density-based basis-set correction approaches, which were primarily
developed to mitigate the basis-set incompleteness error of wave function
methods. An efficient implementation of the density-based scheme is
also presented, utilizing the density-fitting approximation. The performance
of these approaches is comprehensively tested for the second-order
Møller–Plesset (MP2), coupled-cluster singles and doubles
(CCSD), and CCSD with perturbative triples [CCSD(T)] methods with
respect to the corresponding complete basis set references. It is
demonstrated that the density-based correction together with complementary
auxiliary basis set (CABS)-corrected Hartree–Fock energies
is highly robust and effectively reduces the error of the standard
approaches; however, it does not outperform the corresponding explicitly
correlated methods. Nevertheless, what still makes the density-corrected
CCSD and CCSD(T) methods competitive is that their computational costs
are roughly half of those of the corresponding explicitly correlated
variants. Additionally, an incremental approach for standard CCSD
and CCSD(T) is introduced. In this simple scheme, the total energies
are corrected with the CABS correction and explicitly correlated
MP2 contributions. As demonstrated, the resulting methods yield surprisingly
good results, below 1 kcal/mol for thermochemical properties even
with a double-ζ basis, while their computational expenses are
practically identical to those of the density-based basis-set correction
approaches.