Abstract
The impact of complete basis set extrapolation schemes (CBS), diffuse functions, and tight weighted-core functions on enthalpies of formation predicted via the DLPNO-CCSD(T1) reduced Feller-Peterson-Dixon approach has been examined for neutral H,C,O-compounds. All tested three-point (TZ/QZ/5Z) extrapolation schemes result in mean unsigned deviation (MUD) below 2 kJ mol(-1) relative to the experiment. The two-point QZ/5Z and TZ/QZ CBS 1/l(max)(3) extrapolation schemes are inferior to their inverse power counterpart (1/( l(max)+1/2 )(4)) by 1.3 and 4.3 kJ mol(-1). The CBS extrapolated frozen core atomization energies are insensitive (within 1 kJ mol(-1)) to augmentation of the basis set with tight weighted core functions. The core-valence correlation effects converge already at triple-zeta, although double-zeta/triple-zeta CBS extrapolation performs better and is recommended. The effect of diffuse function augmentation converges slowly, and cannot be reproduced with double-zeta or triple-zeta calculations as these are plagued with basis set superposition and incompleteness errors.