Gold-substrate interaction is essential in gold-catalyzed organic transformations. This study uses high-level coupled cluster calculations with core-valence correlation and complete basis set (CBS) limit extrapolation as a reference, for assessing the performance of popular density functional theory (DFT) approximations for a variety of Au(I)/Au(III) complexes with unsaturated aliphatic hydrocarbon CnHm substrates (ethene, ethyne, and allene). The tested functionals cover from LDA to GGA and meta-GGA, and to hybrids and double hybrids (LSDA, PBE, M06-L, TPSS, B3LYP, PBE0, M06, M06-2X, TPSSh, B2-PLYP, B2GP-PLYP). Both the geometry and bond dissociation energy (De) of the Au-CnHm complexes are studied. Our findings show that B2GP-PLYP, PBE0, and B2-PLYP are the best performing functionals for this set of Au-CnHm complexes. DFT dispersion correction (DFT-D3), though very helpful for some functionals (e.g., B3LYP and B2-PLYP), does not uniformly improve the results of all functionals. Ab initio methods like MP2 and SCSMP2 are also tested. MP2 is found to be the worst performing method, and while SCSMP2 greatly improves the results, still its accuracy is lower than that of the best functionals, B2GP-PLYP, PBE0, and B2-PLYP.

To improve the accuracy of local coupled cluster (LCC) methods in computing activation energies, we propose herein a new computational scheme. Its applications to various types of late-transition-metal-catalyzed reactions involving Au, Pt, and Ir indicate that the new corrective approach for LCC methods can downsize the mean unsigned deviation and maximum deviation, from the CCSD(T)/CBS reference, to about 0.3 and 0.9 kcal/mol. Using this method, we also calibrated the performance of popular density functionals, with respect to the same test set of reactions. It is concluded that the best functional is the general-purpose double hybrid functional B2GP-PLYP. Other well-performing functionals include the "kinetic" functionals M06-2X and BMK, which have a large percentage of HF exchange, and general-purpose functionals like PBE0 and wB97X. Comparatively, general-purpose functionals like PBE0 and TPSSh perform much better than the tested "kinetic" functionals for Pt-/Ir-catalyzed reactions, while the opposite is true for Au-catalyzed reactions. In contrast, wB97X performs more uniformly in these two classes of reactions. These findings hint that even within the scope of late transition metals, different types of reactions may require different types of optimal DFT methods. Empirical dispersion correction of DFT was found to have a small or no effect on the studied reactions barriers.

Mononuclear Ru-based water oxidation catalysts (WOCs) are an important class of WOCs for water splitting. In this work, through high-level coupled cluster calculations (CCSD(T)/CBS), we have examined a variety of density functionals for their performances in the whole catalytic cycle of water oxidation catalyzed by mononuclear Ru-based WOCs. The tested functionals cover a wide range from pure GGA and meta-GGA to hybrids and double hybrids (TPSS, OLYP, BP86, M06-L, B3LYP, PBE0, M06, M06-2X, TPSSh, CAM-B3LYP, wB97X, B2-PLYP, B2GP-PLYP). Depending on different reaction types and species in the catalytic cycle, the performances of different DFTs vary severely, whose trends are summarized in the paper. Our results indicate that using a single approximate functional to accurately model all reactions involved in the whole Ru-based WOC catalytic cycle is still a very challenging task. In the current status, PBE0 and M06 may be recommended for the whole catalytic cycle. Generally, this study provides a guide for selecting an appropriate DFT method in modeling each of the various steps in water oxidation catalyzed by Ru-based WOCs. The sensitivity of DFT and ab initio results upon the degree of basis set completeness found in this work is also worthy of attention in the future theoretical study of mononuclear Ru-based WOCs.