Discovery of high Fe content amorphous alloys with desired soft magnetic properties by incremental machine learning

Abstract

The challenge of increasing Fe content in Fe-based amorphous alloys while maintaining their glass forming ability and thereby achieving high magnetic flux density is an important issue in the field of soft magnetic amorphous alloys. Despite extensive efforts in designing such alloys through high-throughput computational methods, identifying and preparing amorphous alloys with Fe content exceeding 85 at% remains challenging, primarily due to the lack of sustained optimization for high Fe-content compositions. In this study, an updated incremental machine learning approach is employed for the first time to address this issue. Initial models were developed, followed by the designation of a series of high Fe content alloys. Then, models were iteratively refined and optimized based on experimental results, utilizing a k-nearest neighbors classifier with 95.4% accuracy and a gaussian process regressor with a coefficient of determination of 0.94. As a result, a series new Fe-Si-B-P-C amorphous alloys with Fe content higher than 85 at% are successfully prepared. Among these alloys, the Fe_85.5Si₂B_8.5P₂C₂ amorphous alloy stands out with the highest Fe content in Fe-Si-B-P-C alloys compared to previous studies, achieving a high magnetic flux density of 1.68 T. The incremental machine learning model allowed for precise component adjustment, achieving a balance between glass forming ability and high Fe content, thereby offering a more accurate method for designing novel Fe-based amorphous alloys.