The CUAGAU Set of Coupled-Cluster Reference Data for Small Copper, Silver, and Gold Compounds and Assessment of DFT Methods.

We have obtained benchmark data for a set of small molecular systems of Cu, Ag, and Au using coupled-cluster methods. Using this collection of reference data (that we termed the CUAGAU set) for assessing DFT-type methods, we find the MN15-L nonhybrid DFT to be cost-effective for geometry optimization [mean absolute deviation (MAD) in bond length = 0.20 Å], with an accuracy that is comparable to that for the double-hybrid (DH) DFT method DSD-PBEP86 (MAD = 0.19 Å). For the computation of thermochemical properties, among "conventional" (non-MP2-based) DFT methods, the best performance is found for the global-hybrid meta-GGA functional MN15, with an MAD of 11.4 kJ mol-1. We also find the nonhybrid method B97M-rV to have a reasonable performance (MAD = 14.4 kJ mol-1), and it may serve as a cost-effective means for qualitative study. If we look beyond conventional functionals, we find DSD-PBEP86 (MAD = 7.3 kJ mol-1) to be more accurate than even MN15. Nonetheless, this level of accuracy is still not sufficient for quantitative studies. In this regard, high-level wave function methods such as composite procedures that are based on coupled cluster are still indispensable for obtaining reliable reference data for transition-metal species.