RECON: An algorithm for molecular charge density reconstruction using atomic charge density fragments

A computer algorithm is developed for the rapid construction of molecular charge densities and charge density-based properties using atomic charge density fragments. This algorithm, developed for modeling electronic properties of molecules, is based on the theory of atoms in molecules which defines the properties of atoms in molecules.;A library of atomic charge density fragments is built in a form that allows for the rapid retrieval of the fragments and molecular assembly. Presently, there are 170 atomic charge density fragments available for the modeling of molecular systems. There is a surface and data file associated with each atomic charge density fragment in the library. The surface file is a numerical representation of the atomic charge density's shape in three-dimensional space. The data files contain information that describes important topological features of the atomic charge densities. This topological information is used to orientate the atomic charge density fragments into their proper molecular coordinate-space positions. The data files also contain atomic charge density-based descriptors. These descriptors are designed to encode electronic and structural information deemed important for the study of chemical and physical behavior controlled by intermolecular interactions.;This method of molecular charge density construction is used to model the geometric and electronic properties of 91 test molecules. In most cases, the results compare remarkably well with results obtained using traditional ab initio methods. The cases where large discrepancies exist between the results obtained by the two methods suggest a need for more suitable atomic charge density fragments not presently available in the library.;Charge density-based descriptors are used in a quantitative structure-property relationships (QSPR) investigation involving the high performance liquid chromatography (HPLC) retention properties of 22 high-energy molecules. The QSPR models generated using these descriptors are shown to be of a higher quality than those models generated in a previous study using the theoretical linear solvation energy relationships (TLSER) methodology.