Abstract

Models for the interaction energy inevitably rely on some means of partitioning the molecular density and density responses into contributions from atom-like domains. These partitioning methods are often arbitrary and unsatisfactory, leading to either unphysical atomic properties, or to poorly convergent expansions. In this talk I will demonstrate that the iterated stockholder atoms (ISA) procedure of Lillestolen and Wheatley (Chem. Comm. 2008), in particular, a recent robust basis-space version of this approach (Misquitta et al. 2014) alleviates these problems thereby providing us with a unique and numerically robust means of obtaining atomic-like properties which are physically compelling, and with which intermolecular interaction models can be constructed with accuracy and ease. The resulting potentials are systematically improvable, that is, detail can be systematically added or removed, they are minimally anisotropic in a rigorous sense, and may be good candidates for parameter transferability. Additionally, the ISA partitioning provides us with a means for accounting for intramolecular flexibility effects in a natural way. I will describe how we have used these ideas to develop a general methodology for accurate potential development and will illustrate this with the pyridine system.