Searching for Mechanically Superior Solid-State Electrolytes in Li-Ion Batteries via Data-Driven Approaches

Area of Science:

• Materials Science
• Electrochemistry
• Computational Chemistry

Background:

• Lithium-ion solid-state electrolytes (SSEs) offer significant potential for next-generation batteries.
• Commercialization is hindered by interfacial instability and dendrite formation, impacting safety and performance.
• Identifying mechanically robust SSEs is crucial for overcoming these limitations.

Purpose of the Study:

• To develop and apply a machine learning (ML) model for screening mechanically superior SSE candidates.
• To predict the elastic properties (shear and bulk moduli) of a large dataset of potential SSEs.
• To demonstrate the efficacy of active learning in improving prediction accuracy and reducing data requirements.

Main Methods:

• A machine learning regression algorithm was employed to screen 17,619 candidate SSE structures.
• Elasticity data from 14,238 structures were used to train ML models predicting shear and bulk moduli.
• Physiochemical and structural properties were utilized as descriptors for the ML models.
• First-principles calculations were performed for validation of ML predictions.
• An active learning strategy was implemented to iteratively refine model predictions and reduce uncertainty.

Main Results:

• The ML surrogate model achieved high prediction accuracy, with R-squared values of 0.819 for shear modulus and 0.863 for bulk modulus.
• Active learning significantly improved prediction accuracy (R-squared from ~0.6-0.8) by incorporating a smaller percentage of new data (32-63%).
• The validated model effectively predicts elastic properties for screening mechanically superior SSEs.

Conclusions:

• The developed ML model and active learning approach can significantly accelerate the identification of SSEs with desired mechanical properties.
• This computational strategy addresses key challenges in SSE development, paving the way for safer and more efficient lithium-ion batteries.
• The findings provide a pathway to overcome commercialization barriers related to interfacial stability and dendrite growth.