Discovery and design of Equiatomic Heusler alloys magnetic properties using machine learning with DFT approach

Heusler alloys are recognized for their manifold functionalities, including half-metallic ferromagnetism and several magnetic effects. The study highlights the challenges of the discovery of novel materials in between the huge potential of more than 3.2 million quaternary and Equiatomic Heusler alloys (EQHAs) by touting automated high-throughput computing and machine learning (ML) techniques to enhance the discovery process. The eXtreme Gradient Boosting (XGBoost) model is utilized for the prediction of magnetic properties in the design of new Equiatomic Heusler alloys with improved magnetic efficiency. The proposed methodology involves data collection, preprocessing, feature engineering, model training and testing, metrics performance and its verification it involves a complete dataset to ensure the quality of the data. Stringent evaluation metrics were considered in this study, yielding a high performance with R2 = 0.86, MSE = 0.10, and MAE = 0.26. Bayesian optimization (BO) employs ML to construct a prediction model clarifying the correlation between material design parameters and their attributes, subsequently utilizing decision theory to recommend the most efficient design which is further validated by DFT computation with a very low predictive error reported as the lowest at 1.08 % and the most at 15.01 %. The results highlight the capability of ML in accelerating the discovery of new magnetic materials and show the potential for future research in the broad expanse of Heusler alloys.