Accelerated Discovery of Novel Garnet-Type Solid-State Electrolyte Candidates via Machine Learning

Area of Science:

• Materials Science
• Electrochemistry
• Computational Materials Science

Background:

• All-solid-state batteries (ASSBs) offer higher energy density and stability compared to conventional lithium-ion batteries (LIBs).
• Solid-state electrolytes (SSEs) are critical for the structural integrity and performance of ASSBs.

Purpose of the Study:

• To develop a machine-learning-based surrogate model for efficiently searching ideal garnet-type solid-state electrolyte candidates.
• To identify novel SSE materials with superior mechanical and ionic conductivity properties for next-generation batteries.

Main Methods:

• A machine learning surrogate model was developed using a database of elasticity and descriptors from prior studies.
• The model predicted elastic properties for 5329 potential garnet structures derived from Li7La3Zr2O12 by substituting 73 elements.
• First-principles calculations and ab initio molecular dynamics simulations were used for validation and confirmation of ionic conductivity and diffusion behavior.

Main Results:

• The active learning process effectively reduced prediction uncertainty in material property prediction.
• 10 new tetragonal-phase garnet SSEs were identified, exhibiting superior mechanical stability and high ionic conductivity.
• The developed model and database serve as a foundation for discovering next-generation SSE materials.

Conclusions:

• Machine learning accelerates the discovery of advanced solid-state electrolytes for all-solid-state batteries.
• The identified garnet SSEs demonstrate significant potential for improving battery safety and performance.
• This computational approach provides a powerful tool for materials design in energy storage applications.