Accelerating the discovery of novel magnetic materials using machine learning-guided adaptive feedback.

Area of Science:

• Materials Science
• Condensed Matter Physics
• Computational Materials Science

Background:

• Rare earth (RE) elements are crucial for current magnetic materials used in energy and information devices.
• Developing RE-free magnets is a significant scientific challenge for performance and cost-effectiveness.

Purpose of the Study:

• To discover and synthesize novel rare earth-free magnetic materials.
• To develop an efficient framework for materials design and discovery.

Main Methods:

• Integrated machine learning (ML), adaptive genetic algorithms, and first-principles calculations.
• Experimental synthesis and magnetic property characterization.
• Utilized a feedback framework for iterative design and discovery.

Main Results:

• Successfully synthesized a new RE-free magnetic compound: Fe3CoB2.
• Fe3CoB2 exhibits high magnetic anisotropy (K1 = 1.2 MJ/m3) and saturation magnetic polarization (Js = 1.39 T).
• The material's properties are suitable for permanent magnet applications.

Conclusions:

• The ML-guided approach accelerates the discovery of functional materials.
• Fe3CoB2 represents a viable candidate for RE-free permanent magnets.
• The methodology can be extended to discover other novel functional materials.