Premium
Informatics‐Based Energy Fitting Scheme for Correlation Energy at Complete Basis Set Limit
Author(s) -
Seino Junji,
Nakai Hiromi
Publication year - 2016
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.24455
Subject(s) - extrapolation , basis (linear algebra) , gaussian , mathematics , curve fitting , limit (mathematics) , algorithm , energy (signal processing) , set (abstract data type) , statistical physics , computer science , statistics , mathematical analysis , geometry , quantum mechanics , physics , programming language
Energy fitting schemes based on informatics techniques using hierarchical basis sets with small cardinal numbers were numerically investigated to estimate correlation energies at the complete basis set limits. Numerical validations confirmed that the conventional two‐point extrapolation models can be unified into a simple formula with optimal parameters obtained by the same test sets. The extrapolation model was extended to two‐point fitting models by a relaxation of the relationship between the extrapolation coefficients or a change of the fitting formula. Furthermore, n ‐scheme fitting models were developed by the combinations of results calculated at several theory levels and basis sets to compensate for the deficiencies in the fitting model at one level of theory. Systematic assessments on the Gaussian‐3X and Gaussian‐2 sets revealed that the fitting models drastically reduced errors with equal or smaller computational effort. © 2016 Wiley Periodicals, Inc.