Novel iron-cobalt phosphides and their magnetic properties

Computational methods often spearhead discovery by establishing the energy landscape of a compositional space. The formation energy relative to the minimum energy surface (i.e. convex hull) quantifies the thermodynamic stability and potential synthesizability. We utilize adaptive machine learning feedback to screen over 850,000 structures in the Fe-Co-P ternary space. Our motivation is to find rare-earth free permanent magnets with applications in green energy and advanced technology. Further adaptive genetic algorithm searches and first-principles methods aid in the identification of sixteen new structures below the convex hull. Five of them possess high magnetic polarization (Js>1T). The structures with desirable magnetic properties center on (Fe,Co)2P formulae. This supports the conventional wisdom of the scientific community, whose focuses have included attempting the mixture of the two known end compounds - Fe2P and Co2P. Our work should provide guidance for future synthesis of Fe-Co-P line compounds.