Abstract

Fragment-based quantum chemistry methods are a promising route towards massively parallel electronic structure calculations in large systems. Unfortunately, the literature on this topic consists of a bewildering array of different methods, with no clear guiding principles to choose amongst them. Here, we introduce a conceptual framework that unifies many of these ostensibly disparate approaches. The common framework is based upon an approximate supersystem energy formula for a collection of intersecting (i.e., overlapping) fragments. This formula generalizes the traditional many-body expansion to cases where the "bodies" (fragments) share some nuclei in common, and reduces to the traditional many-body expansion for non-overlapping fragments. We illustrate how numerous fragment-based methods fit within this framework. Preliminary applications to molecular and ionic clusters suggest that two-body methods in which dimers are constructed from intersecting fragments may be a route to achieve very high accuracy in fragment-based calculations.