A Graph Neural Network-Based Approach to XANES Data Analysis

TL;DR

This paper introduces a physics-informed graph neural network and transformer approach for analyzing XANES data to determine three-dimensional structures of materials, reducing the need for expert structural parameter summarization.

Contribution

It develops a novel machine learning framework combining physics-informed models with optimization to analyze XANES data without requiring structural parameters.

Findings

Effective prediction of 3D structures from XANES data

Enhanced analysis efficiency through physics-informed models

Potential for online structure analysis at beamlines

Abstract

X-ray absorption spectroscopy (XAS) is an indispensable tool to characterize the atomic-scale three-dimensional local structure of the system, in which XANES is the most important energy region to reflect the three-dimensional structure. However quantitative analysis of three-dimensional structure from XANES requires users to have a deep understanding and accurate judgment of structural information and summarize several structural parameters, which is often difficult to achieve. In this work, We construct \textbf{physics-informed Graph neural network} and \textbf{Transformer} models for calculating XANES from the input three-dimensional structure; we improve the efficiency of the model based on the physical meaning of XAS; then we combine the model and optimization algorithm to fit the three-dimensional structure of given system. This method does not require users to summarize the structural parameters, has wide application range. It can be applied to the three-dimensional structure analysis of solid materials and has positive significance for the study of structure-function relationship in the fields of energy and catalysis. In addition, this method is expected to be developed into an online three-dimensional structure analysis method for XAS related beamlines.