Laplacian-based generalized gradient approximations for the exchange energy

TL;DR

This paper introduces a new class of exchange energy functionals in density functional theory that incorporate the Laplacian of the density, aiming to improve the modeling of the exchange hole and potential near nuclei.

Contribution

It develops Laplacian-inclusive GGA models that satisfy known constraints and exhibit finite potentials at nuclei, extending traditional GGA approaches.

Findings

Accurately reproduces exchange energies of small atoms.

Maintains finite potential at the atomic nucleus.

Obeys known physical constraints like the Lieb-Oxford bound.

Abstract

It is well known that in the gradient expansion approximation to density functional theory (DFT) the gradient and Laplacian of the density make interchangeable contributions to the exchange correlation (XC) energy. This is an arbitrary "gauge" freedom for building DFT models, normally used to eliminate the Laplacian from the generalized gradient approximation (GGA) level of DFT development. We explore the implications of keeping the Laplacian at this level of DFT, to develop a model that fits the known behavior of the XC hole, which can only be described as a system average in conventional GGA. We generate a family of exchange models that obey the same constraints as conventional GGA's, but which in addition have a finite-valued potential at the atomic nucleus unlike GGA's. These are tested against exact densities and exchange potentials for small atoms, and for constraints chosen to reproduce the SOGGA and the APBE variants of the GGA. The model reliably reproduces exchange energies of closed shell atoms, once constraints such the local Lieb-Oxford bound, whose effects depend upon choice of energy-density gauge, are recast in invariant form.