ADVANCED MATERIALS & PROCESSES | OCTOBER 2023 10 MACHINE LEARNING | AI AI HELPS FIND NEW MAGNETIC MATERIALS Researchers from the DOE’s Ames National Laboratory, Iowa, developed a new machine learning (ML) model for discovering magnet materials that are free of critical elements. The model predicts the Curie temperature of novel material combinations—a vital step in using AI to predict new permanent magnet materials. This model adds to the team’s recently developed capability for discovering thermodynamically stable rare earth materials. High performance magnets containing critical materials such as cobalt and rare earth elements like neodymium and dysprosium are in high demand but have limited availability, thus motivating researchers to find ways to design magnets that require less critical materials. The Ames team used experimental data on Curie temperatures and theoretical modeling to train the ML algorithm. “Finding compounds with high Curie temperature is an important first step in the discovery of materials that can sustain magnetic properties at elevated temperatures,” explains researcher Yaroslav Mudryk. “This aspect is critical for the design of not only permanent magnets but other functional magnetic materials.” According to Mudryk, using an ML method can save time and resources. The team trained their model using experimentally known magnetic materials in order to establish a relationship between several electronic and atomic structure features and Curie temperature. These patterns then give the computer a basis for finding potential candidate materials. To test the model, the team used compounds based on cerium, zirconium, and iron. “The next super magnet must not only be superb in performance, but also rely on abundant domestic components,” says researcher Andriy Palasyuk. The team found that the ML model was successful in predicting the Curie temperature of material candidates, an important first step in creating a faster way to design new magnets for future applications. ameslab.gov. CORROSION-RESISTANT ALLOY DESIGN VIA AI Scientists at the Max Planck Institute for Iron Research, Germany, developed a machine learning (ML) model that enhances predictive accuracy by up to 15% versus existing frameworks for discovering corrosion-resistant alloys. The model’s power stems from fusing both numerical and textual data. “Every Magnet material. Courtesy of Ames Lab. alloy has unique properties concerning corrosion resistance. These properties do not only depend on the alloy composition itself, but on the manufacturing process. Current ML models are only able to benefit from numerical data. However, processing methodologies and experimental testing protocols, which are mostly documented by textual descriptors, are crucial to explain corrosion,” says researcher Kasturi Narasimha Sasidhar. The team used language processing methods similar to ChatGPT along with ML techniques for numerical data and developed a fully automated natural language processing framework. Including textual data in the ML framework enables identification of enhanced alloy compositions resistant to pitting corrosion. With the new ML program, Sasidhar and his team used manually gathered data as textual descriptors. Now, their objective is to automate the data mining process and seamlessly integrate it into the existing framework. Incorporating microscopy images will mark yet another milestone, as the researchers envision the next generation of AI tools that converge textual, numerical, and image-based data. www.mpie.de. Schematic of process-aware deep neural network model. Courtesy of Science Advances.
RkJQdWJsaXNoZXIy MTYyMzk3NQ==