Physics-Informed Machine-Learning Prediction of Curie Temperatures and Its Promise for Guiding the Discovery of Functional Magnetic Materials

High-performance permanent magnets with a high Curietemperature,containing less critical materials, are integral to zero-carbon energysolutions. We built a machine-learning model trained over availableexperimentally measured Curie temperature values to predict the T (C) of multicomponent magnetic materials. We chosetwo compositions from a pseudo-binary (Zr1-x Ce (x) )Fe-2 system, namely,(Zr0.16Ce0.84)Fe-2 and (Zr0.94Ce0.06)Fe-2, to experimentally validate theability of our model to predict the Curie temperature of novel compounds.We also provided a detailed discussion on the correlation of the Curietemperature with the de Gennes scaling factor in rare-earth intermetalliccompounds and its breakdown below a certain rare-earth content. Theelectronic structure calculations (density of states and Fermi surface)were performed using the density functional theory on selected compounds(Zr0.16Ce0.84)Fe-2 and (Zr0.94Ce0.06)Fe-2 to understand the electronic originof a strong magnetic exchange. We found that the change in the electronicdensity of states and electron/hole fillings at the Fermi level directlycorrelate with the Curie temperature. Notably, our model was ableto capture these key electronic structure trends, which show thatphysics-informed machine learning can play a crucial role in designingnew high-performance magnets with improved properties for environmentallysustainable applications.

Automated summary

In this study, a machine-learning model was used to predict the Curie temperature of multicomponent magnetic materials, and the accuracy of the predictions was validated through experiments. Electronic structure calculations were performed to investigate the relationship between the electronic density of states and the Curie temperature, providing insights for designing new high-performance magnets.