Accelerating discovery and design of high-performance solid-state electrolytes: a machine learning approach.

Area of Science:

• Materials Science
• Electrochemistry
• Computational Materials Science

Background:

• Solid-state batteries (SSBs) offer superior performance over liquid electrolyte batteries but face development hurdles.
• Traditional solid-state electrolyte (SSE) screening is slow, costly, and biased, limiting exploration of potential lithium-ion conductors.
• Understanding ion transport mechanisms in crystalline lattices is crucial for designing advanced SSEs.

Purpose of the Study:

• To develop a machine learning (ML) approach for accelerated discovery of high-performance SSEs.
• To identify key physiochemical features governing ion mobility in NASICON compounds.
• To design and validate novel doped SSEs with improved ionic conductivity for lithium-ion batteries.

Main Methods:

• Utilized logistic regression-based machine learning to quantify features influencing ion mobility in NASICON structures.
• Employed ML-identified dopant features to design novel doped SSEs.
• Validated material properties using density functional theory (DFT) calculations.

Main Results:

• Identified two novel doped SSEs, Li₂Mg_0.5Ge_1.5(PO₄)₃ and Li_1.667Y_0.667Ge_1.333(PO₄)₃, with high ionic conductivity.
• Li₂Mg_0.5Ge_1.5(PO₄)₃ exhibits the lowest reported migration barrier (0.261 eV), outperforming LAGP (0.37 eV).
• Li_1.667Y_0.667Ge_1.333(PO₄)₃ shows the second-lowest migration barrier (0.365 eV).

Conclusions:

• The ML-driven approach significantly reduces the time and resources needed for discovering materials with targeted properties.
• This methodology enables efficient exploration of unexplored material compositions for SSEs.
• The adaptable ML framework can be applied to accelerate materials discovery in other fields like catalysis and structural materials.