The random phase approximation applied to ice

TL;DR

This paper evaluates the random phase approximation (RPA) for modeling ice phases, showing it provides a balanced and accurate description of different ice structures, comparable to diffusion Monte Carlo methods.

Contribution

The study demonstrates that RPA offers a more accurate and balanced description of ice phases than standard density functionals, approaching the accuracy of diffusion Monte Carlo.

Findings

RPA yields accurate relative energies of ice phases.

RPA provides balanced volume predictions for different ice structures.

RPA approaches the accuracy of diffusion Monte Carlo.

Abstract

Standard density functionals without van der Waals interactions yield an unsatisfactory description of ice phases, specifically, high density phases occurring under pressure are too unstable compared to the common low density phase I$_h$ observed at ambient conditions. Although the description is improved by using functionals that include van der Waals interactions, the errors in relative volumes remain sizable. Here we assess the random phase approximation (RPA) for the correlation energy and compare our results to experimental data as well as diffusion Monte Carlo data for ice. The RPA yields a very balanced description for all considered phases, approaching the accuracy of diffusion Monte Carlo in relative energies and volumes. This opens a route towards a concise description of molecular water phases on surfaces and in cavities.