AsianScientist (Sep. 17, 2018) – Scientists in Japan have used machine learning in the interpretation of material spectra. Their methods are published in Scientific Reports.
Spectroscopy techniques are commonly used in materials research because they enable identification of materials from their unique spectral features. However, interpreting material spectra requires considerable expertise. Each spectrum is compared with a database containing numerous reference material properties, but unknown material features that are not present in the database can be problematic, and often have to be interpreted using spectral simulations and theoretical calculations.
In addition, the fact that modern spectroscopy instruments can generate tens of thousands of spectra from a single experiment places considerable strain on conventional human-driven interpretation methods, and a more data-driven approach is thus required.
In this study, researchers at the University of Tokyo’s Institute of Industrial Science in Japan developed a data-driven approach to interpret much larger numbers of spectra. The team used theoretical calculations to construct a spectral database in which each spectrum had a one-to-one correspondence with its atomic structure, where all spectra contained the same parameters.
Applying the method to the interpretation of complex spectra from two core spectroscopy methods—energy-loss near-edge structure (ELNES) and X-ray absorption near-edge structure (XANES)—the researchers were able to obtain information about a material that cannot be determined manually and predict a spectrum from the material’s geometric information alone.
The researchers noted that the proposed machine learning method is not restricted to ELNES and XANES spectra and can be used to analyze any spectral data quickly and accurately without the need for specialist expertise. As a result, they expect that their method to have wide applicability in fields such as semiconductor design, battery development and catalyst analysis.
The article can be found at: Kiyohara et al. (2018) Data-driven Approach for the Prediction and Interpretation of Core-electron Loss Spectroscopy.
———
Source: University of Tokyo; Photo: Shutterstock.
Disclaimer: This article does not necessarily reflect the views of AsianScientist or its staff.
