Population Analysis From Variational Chemical Partition of Molecular Position Space

Two years ago, a partition was proposed in the position space of the charge distribution of atoms and molecules, relying on the optimization of the boundaries of space-filling non-overlapping localization regions. This is achieved with the help of global objective functions providing either a measure of the dispersion of numbers of electrons in each localization region or of the dependence of these numbers between them. The output of the method reproduces very well the expected chemical structures in terms of atomic shells, bonds, and lone pairs domains. This method, inspired by the data processing of experimental results, is by construction free of any approximation since it only requires as input the one and two-particle densities of probability. It is based on a strict definition of the concepts of chemical entity, i, e, a set of nuclei and of a given number of localized electrons, for which an electron count followed by a multivariate analysis is carried out. In its first version, the software was extremely inefficient, and its use was therefore limited to atoms and small molecular systems to keep the computational effort within reasonable limits. A significant improvement has been recently achieved, enabling calculations of molecules as large as polyaromatic hydrocarbons. In the present paper, we first discuss the philosophy of previous population schemes, then we describe the features of our new family of population analysis, and then report the results obtained on a rather large sample of molecules and complexes, including polyaromatic hydrocarbons, propellanes, transition metal carbonyl 1:1 complexes and hydrogen-bonded systems as well as reaction mechanisms.