Stress measurement based on magnetic Barkhausen noise for thin films

Non-destructive testing methods based on magnetic Barkhausen noise (MBN) are expected to be applied to thin films. Recently, MBN analysis using machine learning has been performed on bulk materials and used as an in-situ stress and material evaluation method. However, for thin films, it is difficult to measure due to the weak MBN signal intensity, and few examples of its use as a stress and material evaluation method have been reported. This study acquired and analyzed MBN of Fe–Co polycrystalline thin films under bending stress. After pre-processing the acquired MBNs, two representative machine learning algorithms were used to learn the relationship between MBN and stress. By quantitatively comparing the prediction accuracy of each machine learning method, the characteristics of the MBN-based stress evaluation method were discussed from two perspectives: the need for domain knowledge and its applicability to unknown data. This study provides insight into machine learning-assisted MBN analysis as a stress evaluation method for thin films. The extension of MBN-based stress evaluation methods to thin films could be applied to the non-destructive stress evaluation of micro- and nanostructures, where stress state is critical, and could improve process development efficiency.