The curious case of DMSO: A CCSD(T)/CBS(aQ56+d) benchmark and DFT study

This work addresses the pathological behavior of the energetics of dimethyl sulfoxide and related sulfur-containing compounds by providing the computational benchmark energetics of R₂E₂ species, where R = H/CH₃ and E = O/S, with bent and pyramidal geometries using state-of-the-art methodologies. These 22 geometries were fully characterized with coupled-cluster with single, double, and perturbative triple excitations [CCSD(T)], second-order Møller-Plesset perturbation theory (MP2), and 22 density functional theory (DFT) methods with 8, 12, and 12, respectively, correlation consistent basis sets of double-, triple-, or quadruple-ζ quality. The relative energetics were determined at the MP2 and CCSD(T) complete basis set (CBS) limits using 17 basis sets up to sextuple-ζ and include augmented, tight-d, and core-valence correlation consistent basis sets. The relative energies of oxygen-/sulfur-containing compounds exhibit exceptionally slow convergence to the CBS limit with canonical methods as well as significant basis set dependence. CCSD(T) with quadruple-ζ basis sets can give qualitatively incorrect relative energies. Explicitly correlated MP2-F12 and CCSD(T)-F12 methods dramatically accelerate the convergence of the relative energies to the CBS limit for these problematic compounds. The F12 methods with a triple-ζ quality basis set give relative energies that deviate no more than 0.41 kcal mol^-1 from the benchmark CBS limit. The correlation consistent Composite Approach (ccCA), ccCA-TM (TM for transition metals), and G3B3 deviated by no more than 2 kcal mol^-1 from the benchmark CBS limits. Relative energies for oxygen-/sulfur-containing systems fully characterized with DFT are quite unreliable even with triple-ζ quality basis sets, and 13 out of 45 combinations fortuitously give a relative energy that is within 1 kcal mol^-1 on average from the benchmark CCSD(T) CBS limit for these systems.