Explicitly correlated coupled cluster benchmarks with realistic-sized ligands for some late-transition metal reactions: basis sets convergence and performance of more approximate methods

Abstract

CCSD(T)-F12b benchmark calculations have been performed for the energetics and barrier heights of three late-transition metal systems, in increasing order of size: oxidative additions at bare Pd, a model for the Grubbs catalyst, and competitive CC/CH activation by a Rh(PCP) pincer complex. The results depend weakly on the basis set on the main-group atoms but are rather more sensitive to the basis set on the metal. An aug-cc-pwCVTZ-PP set on the metal combined with cc-pVTZ-F12 on the main-group elements yields barriers that are effectively converged in the basis set, but even the combination with aug-ccpwCVTZ- PP on the metal and cc-pVDZ-F12 on the main group, or of def2-TZVPP on the metal and def2-TZVP on the main group, works well enough for most benchmark purposes. Inner-shell correlation cannot be neglected for even semi-accurate work. Simple nonempirical (meta-)GGAs with D3BJ dispersion work quite well for the Grubbs and pincer cases but break down for the Pd example, which requires exact exchange. Hybrids of these same functionals, such as PBE0, TPSS0, and B3PW91, are among the best performers through rung four on Perdew’s ladder. For the Grubbs case, dispersion is very important and D3BJ clearly is superior over D2. Only the DSD double hybrids consistently perform well in the absence of dispersion corrections.