Premium
Basis set dependence, precision, and accuracy of full ab initio gradient optimizations of molecular structures of nonstrained hydrocarbons. I. CC bond lengths
Author(s) -
Häfelinger Günter,
Regelmann Claus Ulrich,
Krygowski Tadeusz Marek,
Wozniak Krzysztof
Publication year - 1989
Publication title -
journal of computational chemistry
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.907
H-Index - 188
eISSN - 1096-987X
pISSN - 0192-8651
DOI - 10.1002/jcc.540100306
Subject(s) - basis set , bond length , basis (linear algebra) , ab initio , chemistry , molecular geometry , linear regression , ab initio quantum chemistry methods , computational chemistry , molecular physics , mathematics , molecule , geometry , density functional theory , statistics , organic chemistry
Different highly accurate experimental determinations of molecular structures (ED: r g and r 0 α , MW: r 0 , r s , and r z , and X‐ray distances) of 14 unstrained hydrocarbons have been linearly correlated with corresponding self‐consistent field Hartree Fock (SCF HF) ab initio full gradient‐optimized structures. From the eight applied basis sets ( STO ‐3G, STO ‐6G; 3‐21G, 4‐31G, 6‐31G, 6‐311G; 6‐31G*, and 6‐31G**) the 6‐31G basis set, although not best regarding total energies, yields statistically the most precise regression equation, which allows the prediction of ED r g CC distances as reliable as the best experimental determinations. Surprisingly the accuracy of all calculated CC distances (measured as difference Δ between calculated CC distances and experimental ED r g values) depends linearly on bond distances, with the largest deviations being observed for triple bonds. This seems to be a clear indication of different influences of correlation effects on calculated geometries which are neglected in the applied HF treatment. The linear regression equations presented here allow the prediction of any kind of experimental CC distance parameters for each of the eight basis sets considered. Even experimentally unknown r e CC distances may be predicted from these single determinantal HF optimizations.