The northeast materials database for magnetic materials.

Area of Science:

• Materials Science
• Condensed Matter Physics
• Computational Materials Science

Background:

• High-performance magnetic materials are crucial for advanced technologies.
• Existing data-driven approaches are hindered by incomplete and inaccurate material databases.
• Developing comprehensive, experiment-based databases is essential for accelerating materials discovery.

Purpose of the Study:

• To create a comprehensive, experiment-based magnetic materials database using Large Language Models (LLMs).
• To enable accurate classification and prediction of magnetic properties for materials discovery.
• To identify novel magnetic materials with high operating temperature ranges.

Main Methods:

• Utilized Large Language Models (LLMs) for automated data extraction from experimental sources.
• Developed the Northeast Materials Database (NEMAD) with 67,573 entries, including composition, magnetic, and structural data.
• Trained machine learning models for material classification (ferromagnetic, antiferromagnetic, non-magnetic) and transition temperature prediction.

Main Results:

• The NEMAD database provides a rich resource for magnetic materials research.
• Machine learning classification achieved 90% accuracy.
• Regression models accurately predicted Curie (Néel) temperatures with R² values of 0.87 (0.83).
• Identified 25 ferromagnetic and 13 antiferromagnetic material candidates with high predicted transition temperatures.

Conclusions:

• LLMs can effectively automate data extraction for creating comprehensive materials databases.
• Machine learning models trained on NEMAD can accelerate the discovery of novel magnetic materials.
• This approach demonstrates a feasible pathway for discovering high-operating-temperature magnetic materials.