Mining Solid-State Electrolytes from Metal-Organic Framework Databases through Large Language Models and...

Area of Science:

• Materials Science
• Electrochemistry
• Artificial Intelligence

Background:

• Metal-organic frameworks (MOFs) show promise as solid-state electrolytes (SSEs) for Li+ ion conduction.
• Development of MOF SSEs is limited by complexity and lack of design guidelines.

Purpose of the Study:

• To leverage AI, specifically LLMs and machine learning, to accelerate the discovery and design of MOF SSEs.
• To establish a new paradigm for materials discovery through AI-assisted mining.

Main Methods:

• Interactive text mining using LLMs to extract MOF SSE data.
• Construction of a specialized dataset of MOF structural and electrochemical properties.
• Representation clustering to identify promising MOF SSE candidates from a large dataset.

Main Results:

• Successfully mined MOF SSEs from over 11,000 candidates using LLMs and clustering.
• Identified NOTT-400 as a promising MOF SSE with high Li+ conductivity (2.23 × 10-4 S cm-1) and wide electrochemical stability (0-4.79 V).
• Validated the AI-driven approach through physicochemical characterization and electrochemical demonstration.

Conclusions:

• AI, particularly LLMs, can significantly accelerate the identification of novel MOF SSEs.
• The AI-driven methodology provides a reliable and efficient approach for materials discovery.
• This work establishes a new paradigm for designing MOF SSEs with desirable properties.