Deconvolving Lithium Diffusion Mechanisms for Solid Polymer Electrolyte Systems Through Graph Neural Network-Enabled Geometric Analysis of Local Structure in Amorphous Polyethylene Oxide Simulations

Solid polymer electrolytes (SPEs) have emerged as promising alternatives to conventional liquid electrolytes in lithium battery systems due to their low density, mechanical compliance, and low flammability but are challenged by lower ionic conductivity. Molecular dynamics (MD) simulations can be used to accelerate the design of novel SPEs by allowing for mechanistic understanding of lithium ion diffusion at the atomic scale. It has been determined that lithium diffusion in the most promising polymer candidate, polyethylene oxide (PEO), is limited by lithium’s dependence on the polymer’s segmental motion. However, the deconvolved contributions of segmental motion and ion hopping to lithium’s overall conductivity across a range of polymer systems have not been fully explained. Also, the complex interplay between amorphous polymer structure and lithium diffusion mechanisms is not yet understood.<br/><br/>In this work, we find evidence that local order in the amorphous polymer electrolytes can help elucidate and explain lithium diffusion mechanisms across a range of polymer chemistries. A combination of local polymer geometry analysis and graph neural network-based atomic environment featurization is used to identify and characterize sites that can be occupied by lithium ions. Additionally, by extracting site-to-site hopping events, the contributions of hopping and segmental motion on lithium diffusion are decoupled and quantified. This analysis was performed on in-house MD simulations run for hundreds of nanoseconds at a range of salt concentrations. By using the developed high-throughput simulation and analysis workflow, lithium salt concentration effects on lithium diffusion were able to be methodically compared, allowing for better understanding of polymer design criteria for novel solid polymer electrolyte systems.