We have performed computational screening of more than 2,000 initial alloy compositions using density functional theory (DFT), CALPHAD modeling, and an active machine learning (ML) framework. These screening results include significant advances in MM'X-type and Fe2P-type magnetocaloric and permanent magnet materials. Theory suggested Mn-Ni-Fe-Si compositions for phase stability and magnetic distortion, while ML suggested Al doping, leading to promising compositions for the selected applications. For Fe2P-type magnetocaloric materials, theory suggested compositions with Mn/Fe ratios close to one, and substitution of P with Si and B increased the magnetic entropy change. For MM'X-type permanent magnets, various Fe-Ga and Fe,Co-Mn-B compositions were identified as promising materials due to their predicted high magnetization. For Fe2P-type permanent magnets, high anisotropy (1.5 MJ/m³) for (Fe,Co)2(P,Si)) was predicted by substituting Fe with Co and P with Si, with ongoing efforts to avoid or minimize Co while maintaining the magnetic properties.
Experimental synthesis and validation have confirmed the computational predictions. Key results include the successful synthesis of Mn-Ni-Si systems with Fe and Cu substitutions in MM'X-type magnetocaloric materials, achieving desired properties at room temperature. For Fe2P-type magnetocaloric materials, we have synthesized (Mn,Fe)2(P,Si,B)-type compounds, optimizing the Mn/Fe ratio and significantly reducing the thermal hysteresis with boron doping. Discovery of potential hard magnetic phases in MM’X permanent magnets is achieved. Optimum composition balanced with microstructural refinement of Fe2P permanent magnets is also achieved.
