(66j) Computational Investigation of Water Glasses Using Machine Learning Potentials

The molecular origins of water's anomalous properties have long been a subject of scientific inquiry. The liquid-liquid phase transition hypothesis posits the existence of distinct low-density and high-density liquid states at low temperatures and moderate pressures, separated by a first-order transition terminating at a liquid-liquid critical point. This hypothesis has gained increasing experimental and computational support and offers a thermodynamically consistent framework to explain many of water’s anomalies. However, experimental challenges associated with avoiding rapid crystallization near the postulated liquid-liquid critical point has focused attention on water's canonical glassy states: low-density and high-density amorphous ice. Recent progress in machine learning potentials have ushered in a new era of exploration for understanding many of water's complex behaviors by combining the accuracy of quantum mechanical calculations with the efficiency required for extensive molecular dynamics simulations.

In this work we employ two Deep Potential machine learning models, trained on the SCAN density functional and the Many-Body Polarizable potential, respectively, to conduct the first investigation of water's glassy phenomenology based on quantum mechanical calculations. Despite being trained only on equilibrium phases, these models accurately capture the structures and transformations of both water glasses, including their interconversion along different thermodynamic paths. Our simulations reveal that isobaric quenching of liquid water at various pressures generates a continuum of intermediate amorphous ices, with fluctuations in glass density reaching a maximum near the critical pressure. This suggests experimental pathways for probing criticality through glasses quenched near the critical pressure, where enhanced density fluctuations may persist in the glassy state. Moreover, the glass transition temperatures of amorphous ices produced at different pressures reveal two distinct branches with opposite pressure dependencies, providing compelling evidence for two distinct amorphous states directly connected to the underlying liquid-liquid phase transition. This work establishes a comprehensive framework for investigating water's complex polyamorphism with quantum mechanical accuracy while further supporting the liquid-liquid phase transition hypothesis.