CRITICAL ANALYSIS OF ELECTRIC-FIELD MODELING - FORMAMIDE

A detailed analysis of ab initio derived electric field and multipole modeling is presented using formamide as an example. Penetration effects, a problem encountered with most currently used modeling methods, are avoided using the Overlap Multipolar Expansion (OME) technique computed up to the hexadecapole level. This method is shown to be grid insensitive. The OME fields and subsequent atomic charge modeling are found to be more sensitive to correlation effects than to basis sets changes. Three different types of error analysis are applied to test the accuracy of regenerated fields from various multipole descriptions. It is shown that atom centered charge models, those most commonly used in the literature, have significant errors. Considerable improvement can be made by using various combinations of charge, dipole, and quadrupole atom- or bond-centered models. For molecular dynamic programs it is argued that the best compromise between accuracy and efficiency is to use both atom and fictive atom charge models calculated from MP2 level density matrices using at least DZP level basis sets.