Abstract
The altered solvation structures and dynamical properties of water molecules at the metal/water interfaces will affect the elementary step of an electrochemical process. Simulating the interfacial structure and dynamics with a realistic representation will provide us with a solid foundation to make a connection with experimental studies. To surmount the accuracy-efficiency tradeoff and provide dynamical insights, we use state-of-the-art machine learning molecular dynamics (MLMD) to study the water exchange dynamics, which are fundamental to adsorption/desorption and electrochemical reaction steps. We reproduce interfacial structures consistent with ab initio molecular dynamics (AIMD) results and obtain diffusion and reorientation dynamics in agreement with the experiment. We show that the hydrogen bonds at the interface become stronger than those in bulk water, which makes the diffusion, reorientation, and hydrogen-bond dynamics slower. Our findings reveal that the spatial correlation of desorption events, driven by the breaking and making of hydrogen bonds, accelerates water exchange dynamics. These dynamics occur on timescales of several hundred picoseconds at 337 K and 302 K. We take a solid step forward toward studying the in situ interface water dynamics and attribute the fast water exchange dynamics to the spatial correlation of the desorption events.