Just how good is DFT?

Abstract

Density functional theory (DFT) has become a ‘‘household tool,’’ so to speak, for the average chemist. We now take it for granted: we expect DFT to provide a reasonably good description of almost any molecule we can make—or rather, we can think of—and justifiably so. For organic molecules, this is a perfectly valid expectation. However, for transition metal complexes and transition states, the performance of DFT, while generally very good, is less certain, while van der Waals complexes are largely beyond the range of applicability of current DFT methods. Given DFT’s broad and increasing popularity among bioinorganic chemists, a need has been felt for a critical evaluation of the performance of DFT, from a bioinorganic perspective. This series of four Commentaries is intended to address that need. Modern DFT began in 1964, with a famous paper by Hohenberg and Kohn [1] which contained the following two rigorous theoretical results. First, the groundstate electron density uniquely determines the electronic wavefunction and hence all ground-state properties of an electronic system. Second, the energy of an electron distribution is a function—more technically, a functional—of the electron density and this functional is a minimum for ground-state densities. The Hohenberg–Kohn theorem assures us that the density functional is a universal quantity, but it does not specify its form. These results reduce the problem