skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Machine-Learning X-Ray Absorption Spectra to Quantitative Accuracy

Abstract

Simulations of excited state properties, such as spectral functions, are often computationally expensive and therefore not suitable for high-throughput modeling. As a proof of principle, here we demonstrate that graph-based neural networks can be used to predict the x-ray absorption near-edge structure spectra of molecules to quantitative accuracy. Specifically, the predicted spectra reproduce nearly all prominent peaks, with 90% of the predicted peak locations within 1 eV of the ground truth. Besides its own utility in spectral analysis and structure inference, our method can be combined with structure search algorithms to enable high-throughput spectrum sampling of the vast material configuration space, which opens up new pathways to material design and discovery.

Authors:
ORCiD logo [1];  [2]; ORCiD logo [3];  [4]
  1. Columbia Univ., New York, NY (United States)
  2. Brookhaven National Lab. (BNL), Upton, NY (United States). Nuclear Science and Technology Dept.
  3. Brookhaven National Lab. (BNL), Upton, NY (United States). Center for Functional Nanomaterials
  4. Brookhaven National Lab. (BNL), Upton, NY (United States). Computational Science Initiative
Publication Date:
Research Org.:
Brookhaven National Lab. (BNL), Upton, NY (United States)
Sponsoring Org.:
USDOE Office of Science (SC), Advanced Scientific Computing Research (SC-21); USDOE
OSTI Identifier:
1615592
Alternate Identifier(s):
OSTI ID: 1615112
Report Number(s):
BNL-213850-2020-JAAM
Journal ID: ISSN 0031-9007; PRLTAO
Grant/Contract Number:  
SC0012704; FG02-97ER25308
Resource Type:
Journal Article: Accepted Manuscript
Journal Name:
Physical Review Letters
Additional Journal Information:
Journal Volume: 124; Journal Issue: 15; Journal ID: ISSN 0031-9007
Publisher:
American Physical Society (APS)
Country of Publication:
United States
Language:
English
Subject:
97 MATHEMATICS AND COMPUTING; artificial neural networks; first-principle calculations; machine learning; x-ray absorption spectroscopy

Citation Formats

Carbone, Matthew R., Topsakal, Mehmet, Lu, Deyu, and Yoo, Shinjae. Machine-Learning X-Ray Absorption Spectra to Quantitative Accuracy. United States: N. p., 2020. Web. doi:10.1103/PhysRevLett.124.156401.
Carbone, Matthew R., Topsakal, Mehmet, Lu, Deyu, & Yoo, Shinjae. Machine-Learning X-Ray Absorption Spectra to Quantitative Accuracy. United States. doi:10.1103/PhysRevLett.124.156401.
Carbone, Matthew R., Topsakal, Mehmet, Lu, Deyu, and Yoo, Shinjae. Thu . "Machine-Learning X-Ray Absorption Spectra to Quantitative Accuracy". United States. doi:10.1103/PhysRevLett.124.156401.
@article{osti_1615592,
title = {Machine-Learning X-Ray Absorption Spectra to Quantitative Accuracy},
author = {Carbone, Matthew R. and Topsakal, Mehmet and Lu, Deyu and Yoo, Shinjae},
abstractNote = {Simulations of excited state properties, such as spectral functions, are often computationally expensive and therefore not suitable for high-throughput modeling. As a proof of principle, here we demonstrate that graph-based neural networks can be used to predict the x-ray absorption near-edge structure spectra of molecules to quantitative accuracy. Specifically, the predicted spectra reproduce nearly all prominent peaks, with 90% of the predicted peak locations within 1 eV of the ground truth. Besides its own utility in spectral analysis and structure inference, our method can be combined with structure search algorithms to enable high-throughput spectrum sampling of the vast material configuration space, which opens up new pathways to material design and discovery.},
doi = {10.1103/PhysRevLett.124.156401},
journal = {Physical Review Letters},
issn = {0031-9007},
number = 15,
volume = 124,
place = {United States},
year = {2020},
month = {4}
}

Journal Article:
Free Publicly Available Full Text
This content will become publicly available on April 16, 2021
Publisher's Version of Record

Save / Share: