A Deep Neural Network for the Rapid Prediction of X-ray Absorption Spectra

Area of Science:

• Physical sciences
• Biological sciences
• Materials science

Background:

• X-ray spectroscopy provides detailed information on electronic and geometric structures.
• Advanced theoretical calculations are often needed but are resource-intensive.
• Complex systems like catalysts and batteries present interpretation challenges.

Purpose of the Study:

• To develop a rapid and accessible method for interpreting X-ray absorption spectra.
• To utilize supervised machine learning for analyzing complex spectral data.
• To overcome the limitations of traditional computational methods in X-ray spectroscopy.

Main Methods:

• Development of a deep neural network (DNN) for spectral analysis.
• Input to the DNN is geometric information of the absorption site.
• DNN estimates Fe K-edge X-ray absorption near-edge structure (XANES) spectra.

Main Results:

• Spectra estimation in under a second.
• Prediction of peak positions with sub-eV accuracy.
• Prediction of peak intensities with high precision, significantly smaller than spectral variations.

Conclusions:

• The DNN shows promise for structural refinement of complex iron compounds.
• Machine learning offers a powerful tool for accelerating X-ray spectroscopy data interpretation.
• Future work should focus on expanding the DNN's capabilities and applications.