Atomization energies were calculated using explicitly correlated coupled cluster methods with correlation consistent basis sets for a series of 19 small molecules containing 3d transition metal atoms. The atomization energies were calculated using a modified Feller-Peterson-Dixon approach in which CCSD(T) complete basis set (CBS) limits were obtained using extrapolations of aVTZ∕aVQZ CCSD(T)-F12b correlation energies, and then a series of additive contributions for relativity, core correlation, higher order correlation, and zero-point vibrations were included. The frozen-core CBS limits calculated with F12 methods closely matched the more computational expensive conventional awCVQZ∕awCV5Z CBS extrapolations, with a mean unsigned deviation of just 0.1 kcal∕mol. In particular, the CCSD(T∗)-F12b∕aVDZ and aVTZ atomization energies were more accurate on average than the conventional CCSD(T)∕aVQZ and aV5Z results, respectively. In several cases the effects of higher order correlation beyond CCSD(T), as judged by CCSDT and CCSDT(Q)Λ calculations, were greater than 1 kcal∕mol, reaching 4.5 kcal∕mol for CrO3. For the 16 molecules of this study with experimental uncertainties of ∼3.5 kcal∕mol or less, the final composite heats of formation have a mean unsigned deviation (MUD) from experiment of just 1.3 kcal∕mol, which is slightly smaller than the average of the experimental uncertainties, 1.8 kcal∕mol. The root mean square deviation (RMS) is only slightly larger at 1.7 kcal∕mol. Without the contributions due to higher order correlation effects, the MUD and RMS rise to 2.1 and 2.8 kcal∕mol, respectively. To facilitate the F12 calculations, new (aug-)cc-pVnZ∕MP2Fit (n = Q, 5) and (aug-)cc-pwCVTZ∕MP2Fit auxiliary basis sets were also developed for the transition metal atoms.

A number of economical modifications to the high-accuracy extrapolated ab initio thermochemistry (HEAT) model chemistry are evaluated. The two resulting schemes, designated as mHEAT and mHEAT+, are designed for efficient and pragmatic evaluation of molecular energies in systems somewhat larger than can be practically studied by the unapproximated HEAT scheme. It is found that mHEAT+ produces heats of formation with nearly subchemical (±1 kJ/mol) accuracy at a substantially reduced cost relative to the full scheme. Total atomization energies calculated using the new thermochemical recipes are compared to the results of the HEAT-345(Q) model chemistry, and enthalpies of formation for the three protocols are also compared to Active Thermochemical Tables. Finally, a small selection of transition states is studied using mHEAT and mHEAT+, which illuminates some interesting features of reaction barriers and serves as an initial benchmark of the performance of these model chemistries for chemical kinetics applications.

No abstract

No abstract

No abstract