Integrating advanced inference methods and AI into materials science opens up new possibilities for designing and optimizing soft magnetic alloys which play a crucial role in efficient energy conversion [1]. FeNi-based alloys, which are low-cost, abundant, and free from rare earth elements, have attracted significant interest due to their high saturation magnetization, permeability, and low coercivity. The properties of these alloys can be influenced by various factors, including doping, defects, and inhomogeneities. 

We investigate the influence of the role of different inhomogeneities in FeNi-based alloys with the goal of tailoring the hysteresis behavior of the material. We use a combination of advanced latent inference methods [2,3] and machine learning to automatically detect and characterize defects from magneto-optical Kerr effect video data. We find that we can resolve inhomogeneities already in weakly fluctuating regimes without switching the magnet, where defects playing a role in hysteresis become obvious.  

[1] Gutfleisch, et al. Adv. Mater 23, 821 (2011).

[2] Horenko, et al. Commun. Appl. Math. Comput. Sci. 16, 2 (2021).

[3] Rodrigues, et al., iScience 24, 3 (2021).

We acknowledge financial support from the German Research Foundation (Project-ID 320163632 and CRC/TRR 270, Project-ID 405553726).