Assessment of CCSD(T)-F12 Approximations and Basis Sets for Harmonic Vibrational Frequencies

Abstract

We consider basis set convergence and the effect of various approximations to CCSD(T)-F12 for a representative sample of harmonic frequencies (the HFREQ2014 set). CCSD(T*)(F12*)/cc-pVDZ-F12 offers a particularly favorable compromise between accuracy and computational cost: its RMSD <3 cm(-1) from the valence CCSD(T) limit is actually less than the remaining discrepancy with the experimental value at the valence CCSD(T) limit (about 5 cm(-1) RMSD). CCSD(T)-F12a and CCSD(T)-F12b appear to benefit from error compensation between CCSD and (T).

Toward a W4-F12 approach: Can explicitly correlated and orbital-based ab initio CCSD(T) limits be reconciled?

“…If practiced separately on molecule and separate atoms, this practice is not size-consistent: Marchetti and Werner suggested using the molecule/dimer ratio for all species, restoring size consistency -which we indicate by the suffix "sc" in (T* sc ). In two recent studies, 68,69 we found (T*) to be beneficial for F12 harmonic frequency calculations and for noncovalent interaction energies 69 noncovalent interaction benchmark, 88 we found that the overestimate is mitigated by using the E corr [CCSD-F12b]/E corr [CCSD] correlation energy ratio instead, which we denote by the symbol (Tb), or (Tb sc ) for the size-consistent variant. In the original cc-pV5Z-F12 paper, 67 Unless noted otherwise, the "frozen core" approximation was applied, i.e., all inner-shell orbitals were constrained to be doubly occupied.…”

“…67 ; see also Ref. 68 ), we considered CCSD(F12*) (a.k.a., CCSD-F12c) 83 instead of CCSD-F12b, and found that the difference between the two approaches is only significant for the cc-pVDZ-F12 basis set, which is manifestly inadequate for molecular atomization energies.…”

ExoMol molecular line lists – XVII. The rotation–vibration spectrum of hot SO3

“…The F12b varient is an efficient F12 implementation due to Adler et al (2007). Use of CCSD(T)-F12b methods have been shown to give improved vibrational frequencies compared to standard CCSD(T) calculations (Martin & Kesharwani 2014) but their use for intensity calculations remains relatively untested. We return to this issue below.…”

Linear polarizabilities and second hyperpolarizabilities of streptocyanines: Results from broken‐Symmetry DFT and new CCSD(T) benchmarks

“…Historically, wavefunction‐based techniques, particularly coupled‐cluster methods and Møller–Plesset perturbation theory (MP2 or MP4), have been used to provide accurate predictions for both the linear polarizability ( α ) and second hyperpolarizability ( γ ) in various NLO materials . While, it is generally accepted that the coupled‐cluster method with single and double excitations and perturbative triple excitations (CCSD(T)) is widely viewed as the “gold standard” in quantum chemistry, its usage as a routine computational screening tool is significantly limited due to its immense computational cost (particularly for very large systems) .…”