A Caltech Library Service

Density of states prediction for materials discovery via contrastive learning from probabilistic embeddings

Kong, Shufeng and Ricci, Francesco and Guevarra, Dan and Neaton, Jeffrey B. and Gomes, Carla P. and Gregoire, John M. (2022) Density of states prediction for materials discovery via contrastive learning from probabilistic embeddings. Nature Communications, 13 . Art. No. 949. ISSN 2041-1723. doi:10.1038/s41467-022-28543-x.

[img] PDF - Published Version
Creative Commons Attribution.

[img] PDF - Accepted Version
Creative Commons Attribution Non-commercial No Derivatives.

[img] PDF - Supplemental Material
Creative Commons Attribution.

[img] PDF (Peer Review File) - Supplemental Material
Creative Commons Attribution.


Use this Persistent URL to link to this item:


Machine learning for materials discovery has largely focused on predicting an individual scalar rather than multiple related properties, where spectral properties are an important example. Fundamental spectral properties include the phonon density of states (phDOS) and the electronic density of states (eDOS), which individually or collectively are the origins of a breadth of materials observables and functions. Building upon the success of graph attention networks for encoding crystalline materials, we introduce a probabilistic embedding generator specifically tailored to the prediction of spectral properties. Coupled with supervised contrastive learning, our materials-to-spectrum (Mat2Spec) model outperforms state-of-the-art methods for predicting ab initio phDOS and eDOS for crystalline materials. We demonstrate Mat2Spec’s ability to identify eDOS gaps below the Fermi energy, validating predictions with ab initio calculations and thereby discovering candidate thermoelectrics and transparent conductors. Mat2Spec is an exemplar framework for predicting spectral properties of materials via strategically incorporated machine learning techniques.

Item Type:Article
Related URLs:
URLURL TypeDescription Paper ItemUDiscoverIt Materials Discovery
Kong, Shufeng0000-0003-4264-3330
Guevarra, Dan0000-0002-9592-3195
Neaton, Jeffrey B.0000-0001-7585-6135
Gomes, Carla P.0000-0002-4441-7225
Gregoire, John M.0000-0002-2863-5265
Additional Information:© The Author(s) 2022. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit Received 08 October 2021; Accepted 26 January 2022; Published 17 February 2022. This work was funded by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Award DE-SC0020383 (design of prediction task, development of use case, validation of predicted materials, model evaluation; grant received by J.N., C.G., and J.G.) and by the Toyota Research Institute through the Accelerated Materials Design and Discovery program (development of machine learning models; grant received by C.G. and J.G.). Data availability: The input data as well as the predicted phDOS and eDOS data generated in this study have been deposited in the CaltechData database under accession code 8975 and, and are available at Code availability: Source code for Mat2Spec62 is available from and from Author Contributions: These authors contributed equally: Shufeng Kong, Francesco Ricci. S.K. designed and implemented Mat2Spec with guidance from C.G., F.R., D.G., J.N., and J.G. designed the use case. F.R. performed DFT calculations and interpreted results. S.K., F.R., D.G., and J.G. wrote the manuscript with contributions from all authors. J.N., C.G., and J.G. conceived the project and supervised the work. The authors declare no competing interests. Peer review information: Nature Communications thanks Pierre-Paul De Breuck, Logan Ward, and the other, anonymous, reviewer for their contribution to the peer review of this work. Peer reviewer reports are available.
Funding AgencyGrant Number
Department of Energy (DOE)DE-SC0020383
Toyota Research InstituteUNSPECIFIED
Subject Keywords:Computational methods; Computer science; Thermoelectrics
Record Number:CaltechAUTHORS:20220222-762467100
Persistent URL:
Official Citation:Kong, S., Ricci, F., Guevarra, D. et al. Density of states prediction for materials discovery via contrastive learning from probabilistic embeddings. Nat Commun 13, 949 (2022).
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:113517
Deposited By: Tony Diaz
Deposited On:22 Feb 2022 18:00
Last Modified:22 Feb 2022 18:00

Repository Staff Only: item control page