A cautionary comment on the use of orthogonal localized molecular orbitals for the quantitative analysis of through-space and through-bond orbital interactions

Abstract

The quantitative dissection of orbital interactions into through-space (TS) and through–bond (TB) contributions using orthogonal localized MOs (LMOs), as originally proposed by Heilbronner and Schmelzer, is critically discussed, using as a test case, cyclohexa-1,4-diene. It was found that the (HF/STO-3G) TS interaction energies for this molecule, obtained using three different orbital localization procedures (Foster–Boys, the Weinhold natural bond orbitals and the Weinhold natural localized orbitals) are unsatisfactory in the sense that the calculations give a relatively small splitting between the π₊ and π_– orbitals and an inverted sequence of π* orbitals resulting from supposedly pure TS interactions. This result contradicts the Hoffmann conceptual model of orbital interactions. The source of the problem is traced to the presence of the ‘orthogonalization’ tails associated with the LMOs. An alternative strategy for dissecting orbital interactions, termed the ‘cluster’ approach is discussed using an ethene dimer model as an example.