Understanding The Interplay between Anion Dynamics and Proton Conduction in Solid-Acid Compounds Using Machine Learning Molecular Dynamics

Solid acid materials play a pivotal role as electrolytes for intermediate temperature hydrogen fuel cells. However, our current understanding of the relevant atomic motions that govern proton conduction remains incomplete, necessitating further investigation in different solid acid compounds. It is imperative to explore the atomic-scale correlation between anion and proton dynamics to design novel solid acid compounds with high proton conductivity.<br/><br/>In the superprotonic phase, solid acid proton conductors exhibit intricate behaviors, characterized by local proton hops in the O-H...O bond and anion reorientation. These phenomena comprise a two-step process necessary for long-range proton motion. Existing computational studies treat anion and proton dynamics as independent processes, while we find strong correlations between anion and proton dynamics in both CsH₂PO₄ (CDP) and CsHSO₄ (CHS) from machine learning molecular dynamics over nanosecond timescales. Achieving a nanosecond timescale is crucial to comprehensively study the diffusive regime. We observe that not all anion reorientations contribute to long-range proton motion, underscoring its multifaceted role in superprotonic behavior. Additionally, the contribution of anion reorientations to long-range proton motion exhibits distinctive characteristics in CDP and CHS. Furthermore, we confirm a significant O-H reorientation preceding the long-range proton motion, substantiating a previously hypothesized but unverified process. Our approach leverages machine learning interatomic force fields (MLFFs) developed through uncertainty-aware active learning [1] and equivariant neural networks [2]. By combining ab-initio precision with simulations of thousands of atoms over nanosecond timescales, our MLFFs support our findings on the correlations of anion dynamics and the long-range proton transfer with sufficient statistics.<br/><br/>This work bridges crucial knowledge gaps in the superprotonic behavior of solid acid proton conductors and has the potential to inform the design of advanced solid acid compounds for renewable energy technologies.<br/><br/>[1] Xie, Y., Vandermause, J., Ramakers, S. et al. Uncertainty-aware molecular dynamics from Bayesian active learning for phase transformations and thermal transport in SiC. npj Comput Mater 9, 36 (2023).<br/>[2] A. Musaelian, S. Batzner, A. Johansson, L. Sun, C. Owen, M. Kornbluth, and B. Kozinsky, Learning local equivariant representations for large-scale atomistic dynamics Nat. Commun., 14, 579 (2023)