Researchers at Ames National Laboratory in the United States have developed a new machine learning model that can predict the Curie temperature of new material combinations (the highest temperature at which the material maintains magnetism) for searching for permanent magnet materials that do not contain key elements.

      Researchers used experimental data and theoretical models of Curie temperature to train machine learning algorithms, and selected (Zr0.16Ce0.84) Fe2 and (Zr0.94Ce0.06) Fe2 components from the compound (Zr1-xCex) Fe2 system based on cerium, zirconium, and iron to test the model. The test results indicate that the model has successfully predicted the Curie temperature of candidate materials, which creates conditions for finding high-performance permanent magnet materials with high Curie temperature and fewer key elements. The relevant research results are published in the journal Chemistry of Materials.