Transport of Multivalent Ions under Subnanometer Confinement Revealed by a Machine Learning Potential.

Multivalent ions play a critical role in energy storage, environmental remediation, catalysis, and biomedical research due to their strong interactions with water and charged molecules. However, accurately modeling the transport behavior of multivalent ions within solid-state or biological nanochannels remains a significant challenge. In this study, we develop a machine learning potential trained on data sets derived from molecular dynamics simulations, enabling precise simulation of multivalent ion transport in nanochannels with density functional theory (DFT)-level accuracy. The simulated ion diffusion coefficients at varying salt concentrations show excellent agreement with experimental measurements. Leveraging this potential, we uncover how confinement alters La ion hydration dynamics and the free energy landscapes of ion pairing. In particular, our results reveal that electronic polarization effects reduce the local electric fields generated by ions in nanoconfined multivalent electrolytes, thereby diminishing the tendency for ion association. This work provides a powerful tool for the design of nanofluidic systems in biomimetic applications and energy storage that leverage multivalent electrolytes.