X-ray absorption spectroscopy (XAS) helps researchers determine a material's composition, structure, and electron configuration by analyzing how X-rays interact with samples. The spectral data generated serve as unique fingerprints, revealing how atomic arrangements affect a material's characteristics.
Boron compounds, used widely in emerging technologies, present complex spectra due to atomic modifications and defects. The research team, led by Professor Masato Kotsugi, generated data for various phases and defect types in boron nitride, using both theoretical and experimental methods.
The team applied dimensionality-reduction machine learning, including Principal Component Analysis (PCA), Multidimensional Scaling (MDS), t-distributed Stochastic Neighbor Embedding (t-SNE), and Uniform Manifold Approximation and Projection (UMAP). UMAP performed best, accurately classifying structures and defect types and remaining robust even when experimental noise was present.
"Our findings show that UMAP can be a valuable tool for rapid, scalable, automated, and importantly, objective material identification using complex experimental spectral data," said Professor Kotsugi.
Compared to previous statistical similarity-based methods, this AI-driven process offers higher precision and captures meaningful changes in electronic states. The team plans to deploy the software at the Nano-Terasu synchrotron radiation center, aiming to accelerate progress in fields such as catalysis and energy technology.
"Our method demonstrates the potential of autonomous structural identification, opening up new possibilities for data-driven material design and development of novel materials," Professor Kotsugi said.
Related Links
Tokyo University of Science
Space Technology News - Applications and Research
| Subscribe Free To Our Daily Newsletters |
| Subscribe Free To Our Daily Newsletters |
