Universal Rapid Machine Learning Models for Predicting Unconvoluted and Convoluted X-ray Absorption Spectra

Area of Science:

• Materials Science
• Spectroscopy
• Computational Chemistry

Background:

• X-ray absorption near-edge structure (XANES) is crucial for analyzing local 3D atomic structures in materials and molecules.
• Accurate and rapid computation of XANES spectra from structural data is essential for quantitative analysis.
• Current methods often require element-specific models and extensive data.

Purpose of the Study:

• To develop a universal XANES prediction model that accepts 3D atomic structures as input.
• To demonstrate the model's generalizability across different X-ray absorption spectroscopy (XAS) types and elements.
• To enable rapid, real-time structural validation during XAS experiments.

Main Methods:

• Development of a novel computational model for XANES prediction.
• Input: 3D atomic structures of materials or molecules.
• Output: Predicted XANES or unconvoluted XANES spectra.

Main Results:

• The model demonstrates excellent generalizability across diverse broadening conditions.
• Validated predictive accuracy for hard X-ray XAS (3d/4d transition metals K-edge) and soft X-ray XAS (sulfur K-edge).
• Successfully predicts XANES spectra even with sparse 3D structural data.
• A single unified model can predict XANES for multiple elements, eliminating the need for element-specific models.

Conclusions:

• The developed XANES prediction model offers a powerful, unified approach for analyzing local atomic structures.
• It significantly enhances the efficiency and accessibility of quantitative XANES analysis.
• Empowers researchers with real-time structural validation capabilities during XAS experiments.