Accurate Computational Modeling of Aqueous Phase Chemistry Across Phases and Interfaces

Understanding the physical and chemical behavior of aqueous systems under varying temperature and pressure conditions is a critical scientific challenge. This study presents an innovative approach to simulating the phase diagram of water with unparalleled accuracy. By integrating MB-pol(2023)—a data-driven many-body potential built on first-principles calculations, ensuring chemical precision across all phases—with state-of-the-art enhanced-sampling algorithms, our method captures the quantum nature of molecular motion and thermodynamic equilibria with remarkable fidelity. It provides a detailed and realistic depiction of water's behavior across phases. However, while highly accurate, the model is limited in its ability to simulate bond formation and breaking events.

To address this, we employ an equivariant graph neural network potential (e3NN) trained using the density-corrected SCAN method—the first DFT-based model capable of accurately describing water’s properties and delivering a precise estimation of the autoionization constant. This approach allows us to investigate proton transfer dynamics in aqueous environments, particularly self-ion behavior in bulk water and interfacial regions. Our research explores how distinct molecular interactions influence proton transfer rates and mechanisms in different environments, with special attention to water-wire formations and their impact. Additionally, we examine ion surface propensities to assess the acidic or basic nature of water interfaces.

By combining cutting-edge computational methods, our study provides a comprehensive understanding of water and ion dynamics across diverse phases and interfaces, offering critical insights into water science and interfacial chemistry.