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

Title: Atomic-position independent descriptor for machine learning of material properties

Abstract

The high-throughput screening of periodic inorganic solids using machine learning methods requires atomic positions to encode structural and compositional details into appropriate material descriptors. These atomic positions are not available a priori for new materials, which severely limits exploration of novel materials. In this work, we overcome this limitation by using only crystallographic symmetry information in the structural description of materials. We show that for materials with identical structural symmetry, machine learning is trivial, and accuracies similar to that of density functional theory calculations can be achieved by using only atomic numbers in the material description. For machine learning of formation energies of bulk crystalline solids, this simple material descriptor is able to achieve prediction mean absolute errors of only 0.07 eV/at on a test dataset consisting of more than 85 000 diverse materials. Finally, this atomic-position independent material descriptor presents a new route of materials discovery wherein millions of materials can be screened by training a machine learning model over a drastically reduced subspace of materials.

Authors:
 [1];  [2]
  1. Stanford Univ., CA (United States)
  2. SLAC National Accelerator Lab., Menlo Park, CA (United States)
Publication Date:
Research Org.:
SLAC National Accelerator Lab., Menlo Park, CA (United States)
Sponsoring Org.:
USDOE
OSTI Identifier:
1493407
Grant/Contract Number:  
AC02-76SF00515
Resource Type:
Journal Article: Accepted Manuscript
Journal Name:
Physical Review B
Additional Journal Information:
Journal Volume: 98; Journal Issue: 21; Journal ID: ISSN 2469-9950
Publisher:
American Physical Society (APS)
Country of Publication:
United States
Language:
English
Subject:
75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND SUPERFLUIDITY; 36 MATERIALS SCIENCE

Citation Formats

Jain, Ankit, and Bligaard, Thomas. Atomic-position independent descriptor for machine learning of material properties. United States: N. p., 2018. Web. doi:10.1103/physrevb.98.214112.
Jain, Ankit, & Bligaard, Thomas. Atomic-position independent descriptor for machine learning of material properties. United States. doi:10.1103/physrevb.98.214112.
Jain, Ankit, and Bligaard, Thomas. Thu . "Atomic-position independent descriptor for machine learning of material properties". United States. doi:10.1103/physrevb.98.214112. https://www.osti.gov/servlets/purl/1493407.
@article{osti_1493407,
title = {Atomic-position independent descriptor for machine learning of material properties},
author = {Jain, Ankit and Bligaard, Thomas},
abstractNote = {The high-throughput screening of periodic inorganic solids using machine learning methods requires atomic positions to encode structural and compositional details into appropriate material descriptors. These atomic positions are not available a priori for new materials, which severely limits exploration of novel materials. In this work, we overcome this limitation by using only crystallographic symmetry information in the structural description of materials. We show that for materials with identical structural symmetry, machine learning is trivial, and accuracies similar to that of density functional theory calculations can be achieved by using only atomic numbers in the material description. For machine learning of formation energies of bulk crystalline solids, this simple material descriptor is able to achieve prediction mean absolute errors of only 0.07 eV/at on a test dataset consisting of more than 85 000 diverse materials. Finally, this atomic-position independent material descriptor presents a new route of materials discovery wherein millions of materials can be screened by training a machine learning model over a drastically reduced subspace of materials.},
doi = {10.1103/physrevb.98.214112},
journal = {Physical Review B},
issn = {2469-9950},
number = 21,
volume = 98,
place = {United States},
year = {2018},
month = {12}
}

Journal Article:
Free Publicly Available Full Text
Publisher's Version of Record

Citation Metrics:
Cited by: 5 works
Citation information provided by
Web of Science

Save / Share:

Works referenced in this record:

Big Data of Materials Science: Critical Role of the Descriptor
journal, March 2015


The Open Quantum Materials Database (OQMD): assessing the accuracy of DFT formation energies
journal, December 2015


How Chemical Composition Alone Can Predict Vibrational Free Energies and Entropies of Solids
journal, July 2017


Accelerated Materials Design of Lithium Superionic Conductors Based on First-Principles Calculations and Machine Learning Algorithms
journal, April 2013

  • Fujimura, Koji; Seko, Atsuto; Koyama, Yukinori
  • Advanced Energy Materials, Vol. 3, Issue 8
  • DOI: 10.1002/aenm.201300060

Including crystal structure attributes in machine learning models of formation energies via Voronoi tessellations
journal, July 2017


Crystal structure representations for machine learning models of formation energies
journal, April 2015

  • Faber, Felix; Lindmaa, Alexander; von Lilienfeld, O. Anatole
  • International Journal of Quantum Chemistry, Vol. 115, Issue 16
  • DOI: 10.1002/qua.24917

Solid-liquid coexistence in small systems: A statistical method to calculate melting temperatures
journal, September 2013

  • Hong, Qi-Jun; van de Walle, Axel
  • The Journal of Chemical Physics, Vol. 139, Issue 9
  • DOI: 10.1063/1.4819792

Crystal Graph Convolutional Neural Networks for an Accurate and Interpretable Prediction of Material Properties
journal, April 2018


Finding Unprecedentedly Low-Thermal-Conductivity Half-Heusler Semiconductors via High-Throughput Materials Modeling
journal, February 2014


The optimal one dimensional periodic table: a modified Pettifor chemical scale from data mining
journal, September 2016