Predicting Electrolyte Stability in Lithium and Post-Lithium Metal Batteries: The Failure of Linear Single...

Area of Science:

• Battery Technology
• Materials Science
• Computational Chemistry

Background:

• A stable solid electrolyte interphase (SEI) is critical for lithium metal battery (LMB) safety and performance.
• Conventionally, lowest unoccupied molecular orbital (LUMO) energy predicts electrolyte reductive decomposition, but its accuracy is questioned.

Purpose of the Study:

• To critically evaluate the predictive power of LUMO energy for electrolyte reductive stability.
• To develop advanced computational models for accurate electrolyte design.

Main Methods:

• Density functional theory (DFT) calculations to determine free energy barriers for 200 ion-solvent complexes.
• Development and training of machine learning (ML) models using multiple electronic and structural features.
• Comparison of ML model performance against traditional LUMO energy-based predictions.

Main Results:

• LUMO energies do not linearly correlate with reductive decomposition kinetics or electrolyte stability.
• Electrolyte stability exhibits a nonlinear relationship with LUMO levels across various chemistries.
• Machine learning models significantly outperform LUMO-based linear approaches in predicting reductive stability.

Conclusions:

• Relying solely on LUMO energy for electrolyte screening is insufficient due to nonlinear relationships.
• Multifeature, nonlinear machine learning models offer a more accurate computational framework for electrolyte design.
• This work accelerates the development of stable electrolytes for next-generation batteries.