Hard magnets play a crucial role in converting electromechanical energy in electromobility applications and wind turbines. The increased demand for materials containing rare earth metals has, however, reached a critical level. In the search for material substitutes that are free of rare earth metals, ab initio density functional theory is used to calculate magnetic parameters for real and hypothetical crystal phases. “High throughput screening“ material theory and “Data Mining“ information theory are used in simulations to predict the most promising candidates from thousands of magnetic phases.
The web app MagnetPredictor (in German) was developed at the Fraunhofer IWM to demonstrate the benefit of modern Machine Learning techniques for the prediction of magnetic properties. For (hypothetic) intermetallic phases, a quick rough estimate of these properties can be generated for arbitrary chemical compositions.