Critical benchmarking of popular composite thermochemistry models and density functional approximations on a probabilistically pruned benchmark dataset of formation enthalpies.

摘要: First-principles calculation of the standard formation enthalpy, ΔHf° (298 K), in such a large scale as required by chemical space explorations, is amenable only with density functional approximations (DFAs) and certain composite wave function theories (cWFTs). Unfortunately, accuracies popular range-separated hybrid, “rung-4” DFAs, cWFTs that offer best accuracy-vs-cost trade-off have until now been established for datasets predominantly comprising small molecules; their transferability to larger systems remains vague. In this study, we present an extended benchmark dataset structurally electronically diverse molecules. We apply quartile-ranking based on boundary-corrected kernel estimation filter outliers arrive at probabilistically pruned enthalpies 1694 compounds (PPE1694). For dataset, rank prediction G4, G4(MP2), ccCA, CBS-QB3, 23 DFAs using conventional probabilistic error metrics. discuss systematic errors highlight role empirical higher-level correction plays G4(MP2) model. Furthermore, comment uncertainties associated reference data atoms stemming from these grow molecular size. believe findings will aid identifying meaningful application domains quantum thermochemical methods.