Anisotropy and Magnetism in the LSDA+U Method

TL;DR

This paper analyzes how anisotropy and magnetism interact within LSDA+U functionals, revealing significant differences in energy and magnetic properties depending on the functional used, with implications for electronic structure calculations.

Contribution

It provides a detailed comparison of anisotropy and magnetism effects in various LSDA+U functionals, including new insights into their energetic and magnetic behavior.

Findings

Differences in total energies for integer f-shell configurations across functionals.

AMF functional imposes larger energy penalties on magnetic states than FLL.

Self-consistent LSDA+U enhances magnetic moments and orbital polarization.

Abstract

Consequences of anisotropy (variation of orbital occupation) and magnetism, and their coupling, are analyzed for LSDA+U functionals, both the commonly used ones as well as less commonly applied functionals. After reviewing and extending some earlier observations for an isotropic interaction, the anisotropies are examined more fully and related to use with the local density (LDA) or local spin density (LSDA) approximations. The total energies of all possible integer configurations of an open $f$ shell are presented for three functionals, where some differences are found to be dramatic. Differences in how the commonly used "around mean field" (AMF) and "fully localized limit" (FLL) functionals perform are traced to such differences. The LSDA+U interaction term, applied self-consistently, usually enhances spin magnetic moments and orbital polarization, and the double-counting terms of both functionals provide an opposing, moderating tendency ("suppressing the magnetic moment"). The AMF double counting term gives magnetic states a significantly larger energy penalty than does the FLL counterpart.