Ab-initio Structure and Thermodynamics of the RPBE-D3 Water/Vapor Interface by Neural-Network Molecular Dynamics

TL;DR

This paper employs neural network-enhanced molecular dynamics based on DFT to accurately study the structure and thermodynamics of the water/vapor interface, achieving unprecedented precision in sampling properties along the coexistence line.

Contribution

It introduces a neural network approach to efficiently simulate water's interfacial properties using ab initio calculations, enabling large-scale and long-time simulations.

Findings

Precise structural characterization of the water/vapor interface.

Thermodynamic properties consistent with experimental data.

Enhanced sampling of water properties along the coexistence line.

Abstract

Aided by a neural network representation of the density functional theory (DFT) potential energy landscape of water in the RPBE approximation corrected for dispersion, we calculate several structural and thermodynamic properties of its liquid/vapor interface. The neural network speed allows us to bridge the size and time scale gaps required to sample the properties of water along its liquid/vapor coexistence line with unprecedented precision.