Machine Learning Assisted Prediction of Donor Numbers: Guiding Rational Fabrication of Polymer Electrolytes for Lithium-ion Batteries.

Abstract

Polymer electrolytes are of interest in high-energy-density batteries. However, how the intrinsic electron-donating capability of polymer segments involved in coordination affects lithium-ion dissociation/transmission and rationally guides the design and fabrication of electrolytes is a highly exploratory topic. This study proposes a workable method that integrates machine learning with density functional theory to predict donor numbers (DN) for polymer building units. Using this approach, polymer chains with optimized DN are designed, effectively modulating the solvation structure of lithium-ion. Molecular dynamics simulations confirm the positive impact of polymer chains on rapid transport of lithium ions. Experimental validation of the proposed zwitterionic polymer electrolyte (ZPE) showcases satisfactory parameters: ion conductivity (0.59 mS cm-1), ion migration numbers (0.82), and activation energy (0.016 eV). Electrochemical analysis on Li|ZPE|Li symmetric batteries demonstrate sustained plating/stripping performance exceeding 3000 hours at a current density of 0.2 mA cm-2. Assembled NCM|ZPE|Li batteries exhibit stable cycling over 1400 cycles at 4.3 V, with a capacity retention ratio of 92.3%. Moreover, even under ultra-high voltages of 4.5 V and 4.7 V, NCM|ZPE|Li batteries display stable cycling performances. This approach offers a paradigmatic strategy for polymer molecule design, advancing sustainable battery technologies.