Understanding the heterogeneous nucleation of ice on silver iodide using deep potential molecular dynamics.

Ice nucleation is one of the most unique and widespread phase transitions on Earth. Due to the relatively low phase transition energy barrier to overcome and the ubiquitous existence of foreign substrates, ice nucleation primarily occurs heterogeneously in nature. Despite extensive studies, our understanding of the molecular-scale heterogeneous nucleation process under the influence of silver iodide (AgI) substrate interactions, one of the most efficient ice nucleating agents, remains limited. Using a deep neural network potential, we perform molecular dynamics simulations with ab initio accuracy to investigate the heterogeneous nucleation of ice on AgI. By analyzing the free energy surface of water molecules at the AgI-water interface, we systematically elucidate the mechanism behind the formation of an ice-like hexagonal layer on AgI. The reconstruction of the metastable, disordered hydrogen bond network into this ice-like hexagonal layer facilitates ice nucleation and contributes to the asynchronous crystallization manner. Furthermore, we find that the influence of the AgI substrate propagates through the highly dynamical and collaborative hydrogen bond network, leading to a pre-ordered region at the ice-water interface that reduces the ice growth rate to approximately one-third compared to ice homogeneous nucleation conditions. These findings provide new insights into the early stages of ice heterogeneous nucleation on the AgI surface and expand our understanding of the role substrates play in this process.