Assessing Spin‐Component‐Scaled Second‐Order Møller–Plesset Theory Using Anharmonic Frequencies

Abstract

Four common parametrisations of spin-component-scaled second-order Møller-Plesset (MP2) theory are benchmarked by calculating the anharmonic vibrational frequencies of a test suite consisting of eighteen diatomic and five small molecules. Of the four methods, the scaled opposite-spin MP2 (SOS-MP2), the variable-scaling opposite-spin MP2 (VOS-MP2) and the spin-component-scaled MP2 (SCS-MP2) methods perform statistically better than standard MP2 theory, while the spin-component scaled for nucleic bases MP2 (SCSN-MP2) performs worse. Vibrations of closed-shell diatomic molecules are slightly more accurately described by the SOS-MP2 method of Head-Gordon (ε(MAD) =51 cm(-1) ) than the SCS-MP2 method of Grimme (ε(MAD) =61 cm(-1)) or the size-consistent parametrisation of VOS-MP2 (ε(MAD) =54 cm(-1)). For open-shell diatomic molecules, the SOS-MP2 (ε(MAD) =83 cm(-1)) and SCS-MP2 (ε(MAD) =81 cm(-1)) methods are of similar accuracy, while VOS-MP2 is slightly better (ε(MAD) =77 cm(-1)). Since the VOS-MP2 and SOS-MP2 methods tend to have smaller deviations from experiment, and they can be made computationally more economical than the SCS-MP2 or MP2 methods, we suggest that they should be the preferred ab initio method for computing vibrational frequencies in large molecules.