High Throughput Virtual Screening of Polymer Electrolyes for Li-Ion Batteries with Molecular Dynamics Simulations

Developing high-power, high-density energy storage like lithium metal batteries is vital for sustainable, cost-effective electricity management in electric vehicles and intermittent renewables, reducing reliance on fossil fuels. Furthermore, lithium metal batteries have been proposed for high energy density applications to decarbonize the light- to heavy-duty vehicle sector, as well as for other applications with stringent energy density requirements. Unfortunately, the conventional carbonate-based liquid electrolytes used in lithium-ion batteries degrade both chemically and electrochemically when in the presence of solid lithium electrodes. Thus, alternative electrolyte systems are being investigated to increase diffusion kinetics and chemical stability to ultimately develop long-lasting batteries for electric vehicles and intermittent short-term renewable energy storage. Polymer electrolytes have been an interest in the field because of their enhanced electrochemical stability, increased safety as alternatives to liquid electrolytes.<br/><br/>We propose to use high-throughput molecular dynamics (MD) simulations to screen a large amount potential polymer electrolyte candidates for Li batteries. Using MD simulations for the screening of new materials accelerates the design and commercialization process and helps predict device-level performance using new candidates. However, the parameters historically used in interatomic potentials are inaccurate for new chemical systems, which limits the throughput of these simulations. Therefore, we use a machine learning pipeline to optimize interatomic potential parameters reproducibly for these species by using quantum chemistry simulations as training data across many unique polymer electrolyte systems. We show the capabilities of classical MD simulations with improved class I force fields are able to be used as a tool for ranking polymers based on their experimental conductivities and are able to simulate a large array of new polymer candidates, such as ether, ester, carbonate derivative as well as more different chemistries containing thiol, nitrile, and chemical groups.