Fate of density functional theory in high-pressure solid hydrogen

TL;DR

This study evaluates the effectiveness of various density functional theory functionals in predicting the properties of high-pressure solid hydrogen, revealing significant discrepancies and the need for more advanced methods.

Contribution

It systematically compares LDA, PBE, and BLYP functionals in modeling high-pressure solid hydrogen, highlighting their limitations and the importance of including proton zero-point energy effects.

Findings

BLYP aligns better with experimental phase transition pressures

Proton zero-point energy significantly affects phase diagram predictions

All three functionals show qualitative agreement on band gaps and phase transitions

Abstract

This paper investigates some of the successes and failures of density functional theory in the study of high-pressure solid hydrogen at low temperature. We calculate the phase diagram, metallization pressure, phonon spectrum, and proton zero-point energy using three popular exchange-correlation functionals: the local density approximation (LDA), the Perdew-Burke-Ernzerhof (PBE) generalized gradient approximation, and the semi-local Becke-Lee-Yang-Parr (BLYP) functional. We focus on the solid molecular P$6_3$/m, C2/c, Cmca-12, and Cmca structures in the pressure range from $100<P<500$ GPa over which phases I, II and III are observed experimentally. At the static level of theory, in which proton zero-point energy is ignored, the LDA, PBE and BLYP functionals give very different structural transition and metallization pressures, with the BLYP phase diagram in better agreement with experiment. Nevertheless, all three functionals provide qualitatively the same information about the band gaps of the four structures and the phase transitions between them. Going beyond the static level, we find that the frequencies of the vibron modes observed above 3000 cm$^{-1}$ depend strongly on the choice of exchange-correlation functional, although the low-frequency part of the phonon spectrum is little affected. The largest and smallest values of the proton zero-point energy, obtained using the BLYP and LDA functionals, respectively, differ by more than 10 meV/proton. Including the proton zero-point energy calculated from the phonon spectrum within the harmonic approximation improves the agreement of the BLYP and PBE phase diagrams with experiment. Taken as a whole, our results demonstrate the inadequacy of mean-field-like density functional calculations of solid molecular hydrogen in phases I, II and III and emphasize the need for more sophisticated methods.