A NIMS research team has developed a new experimental method capable of rapidly evaluating numerous material compositions by measuring anomalous Hall resistivity 30 times faster than conventional methods. By analyzing the vast amount of data obtained using machine learning and experimentally validating the predictions, the team succeeded in developing a new magnetic sensor material capable of detecting magnetism with much higher sensitivity. This research was published in npj Computational Materials on September 3, 2025.
The anomalous Hall effect is a phenomenon in which a voltage is generated in a magnetic material when an electric current flows through it, appearing in the direction perpendicular to both the current and the material’s magnetization (that is, from the north to the south magnetic pole). By leveraging this property, changes in magnetization can be sensitively detected as electrical signals, making the effect promising for applications such as read heads in next-generation hard disk drives and high-performance magnetic sensors.
Challenges in material discovery
Alloys containing three or more elements that exhibit a large anomalous Hall effect are considered particularly promising as novel materials. However, the number of possible combinations of elements and their composition ratios is enormous—a so-called combinatorial explosion—and because evaluating each sample is time-consuming, this vast search space has remained largely unexplored.
As a result, there is a strong need to establish both experimental methods capable of efficiently evaluating the anomalous Hall effect in thin-film samples and techniques for effectively identifying promising materials within this vast search space.
Breakthroughs in high-throughput evaluation
The research team successfully developed a novel experimental method for high-throughput evaluation of compositional gradient thin films, in which the composition varies continuously across a single sample. This method enables the evaluation of each composition in approximately 0.2 hours (12 minutes)—a roughly 30-fold increase in speed compared to conventional techniques.
Using this approach, the team systematically investigated the anomalous Hall effect in binary thin films composed of iron (Fe) and a single heavy element. Based on the results, a machine learning model was constructed to predict ternary materials—comprising Fe and two heavy elements—likely to exhibit enhanced anomalous Hall effects.
Guided by these predictions, the Fe–Ir–Pt (iron–iridium–platinum) system was identified as a promising candidate, and a novel material within this system exhibited an anomalous Hall resistivity of 6.5 µ?·cm, surpassing the previous maximum value of 5.25 µ?·cm observed in Fe–X binary systems.
Future directions and applications
This study demonstrated that combining a combinatorial experimental approach—which enables the simultaneous evaluation of numerous compositions—with machine learning is effective for efficiently searching for and discovering new thin-film materials exhibiting large anomalous Hall effects.
Going forward, the team plans to expand the materials search space to discover new materials with even larger anomalous Hall effects. In addition, this approach is expected to advance data-driven materials development using the extensive measurement data generated through combinatorial experiments and facilitate the development of faster, automated and autonomous materials discovery systems.
More information:
Ryo Toyama et al, High-throughput materials exploration system for the anomalous Hall effect using combinatorial experiments and machine learning, npj Computational Materials (2025). DOI: 10.1038/s41524-025-01757-5
Provided by
National Institute for Materials Science
The content is provided for information purposes only.