Breaking through the trade-off between saturation magnetization and coercivity: a data-driven strategy

Abstract

Remarkable trade-off paradoxes widely exist among different magnetic properties of permanent magnetic alloys, especially between saturation magnetization and coercivity, largely impeding the improvement of comprehensive properties. Taking Sm-Co-based alloys as an example, this study proposed a new data-driven material design strategy to dissolve the saturation magnetization-coercivity trade-off and enhance the comprehensive magnetic properties. The machine learning approach and multi-objective optimization method were applied to establish a model for composition design and microstructure regulation to simultaneously maximize saturation magnetization and coercivity. It was found that the electronegativity of the doping element is a key feature that affects both the saturation magnetization and coercivity, and the Pareto front with appropriate alloy composition and grain size was obtained. The materials with best comprehensive magnetic properties in the optimal set were selected for experimental preparation, and the results fully verified the model predictions. The machine learning model and multi-objective optimization method established in this study break through the trade-off between saturation magnetization and coercivity of Sm-Co-based alloys, and the strategy for synergistic improvement of the mutually exclusive properties is appliable to a variety of multi-objective materials design issues.