Correlation energies beyond the random-phase approximation: ISTLS applied to spherical atoms and ions

TL;DR

This paper evaluates the ISTLS correlation energy functional's accuracy for spherical atoms and ions, demonstrating it surpasses the dRPA and shows promise for high-precision DFT calculations.

Contribution

The study tests the ISTLS functional on atomic systems, showing it provides highly accurate correlation energies beyond the dRPA with promising potential for benchmark DFT.

Findings

Achieves within 2mHa/e- for neutral atoms (except Be)

Within 4mHa/e- for ions

Performs better than the dRPA

Abstract

The inhomogeneous Singwi, Tosi, Land and Sjolander (ISTLS) correlation energy functional of Dobson, Wang and Gould [PRB {\bf 66} 081108(R) (2008)] has proved to be excellent at predicting correlation energies in semi-homogeneous systems, showing promise as a robust `next step' fifth-rung functional by using dynamic correlation to go beyond the limitations of the direct random-phase approximation (dRPA), but with similar numerical scaling with system size. In this work we test the functional on fourteen spherically symmetric, neutral and charged atomic systems and find it gives excellent results (within 2mHa/$e^-$ except Be) for the absolute correlation energies of the neutral atoms tested, and good results for the ions (within 4mHa/$e^-$). In all cases it performs better than the dRPA. When combined with the previous successes, these new results point to the ISTLS functional being a prime contender for high-accuracy, benchmark DFT correlation energy calculations.