W4-11: A high-confidence benchmark dataset for computational thermochemistry derived from first-principles W4 data

We show that the purely first-principles Weizmann-4 (W4) computational thermochemistry method developed in our group can reproduce available Active Thermochemical Tables atomization energies for 35 molecules with a 3σ uncertainty of under 1 kJ/mol. We then employ this method to generate the W4-11 dataset of 140 total atomization energies of small first-and second-row molecules and radicals. These cover a broad spectrum of bonding situations and multireference character, and as such are an excellent, quasi-automated benchmark (available electronically as Supporting information) for parametrization and validation of more approximate methods (such as DFT functionals and composite methods). Secondary contributions such as relativity can be included or omitted at will, unlike with experimental data. A broad variety of more approximate methods is assessed against the W4-11 benchmark and recommendations are made.