Self-Diffusion of Point Defects in Hexagonal Ice using Machine Learning Interatomic Potentials

Hexagonal ice is the most common form of ice found in the biosphere and plays a crucial role in sustaining life on Earth. The most common point defects in ice can be categorized into two main types: electronic defects, related to the hydrogen-bond network, and structural defects, associated with the inclusion or removal of water molecules, such as interstitials and vacancies. First-principles calculations indicate that vacancy defects occur at higher concentrations than interstitial ones, while experimental evidence shows that interstitials preferentially diffuse within the hexagonal rings rather than between them. In this work, we investigate the concentration and diffusion mechanisms of point defects in crystalline ice. Density functional theory is employed to evaluate the impact of nonlocal interactions on defect formation and stability. Additionally, a machine learning interatomic potential, combined with harmonic transition state theory, is used to determine the diffusion coefficients of various defects. Our results reveal that interstitial molecules diffuse through hexagonal rings, suggesting a new diffusion mechanism for this defect type.