Spatio-Temporal Characterization of Water Diffusion Anomalies in Saline Solutions Using Machine Learning Force Field

Understanding the dynamics of water in salt solutions is a complex challenge that computational chemists have been addressing. Previously, force fields have encountered difficulties in fully capturing the detailed behaviors of water in the presence of various salts and concentrations, highlighting the necessity for more sophisticated approaches. The emergence of machine learning in computational chemistry, particularly through innovations like the Deep Potential Molecular Dynamics (DPMD), offers a promising alternative that closely aligns with the accuracy of first-principles methods. In this study, we utilized DPMD to explore the effects of salts on water dynamics, examining its performance in relation to ab-initio molecular dynamics, SPC/Fw, AMOEBA, and MB-Pol models. Our focus was on understanding water behavior in salt solutions through the lens of spatio-temporally correlated dynamics. We discovered that the ability of each model to accurately reflect water dynamics in salt solutions is closely tied to its approach to spatio-temporal correlation. This investigation not only highlights the advanced capabilities of MLFFs like DPMD in addressing the complexities of water-salt interactions but also broadens our understanding of the fundamental mechanisms governing these interactions.