Universal machine learning interatomic potentials are ready for phonons

Abstract There has been an ongoing race for the past several years to develop the best universal machine learning interatomic potential. This progress has led to increasingly accurate models for predicting energy, forces, and stresses, combining innovative architectures with big data. Here, we bench...

Full description

Saved in:
Bibliographic Details
Main Authors: Antoine Loew, Dewen Sun, Hai-Chen Wang, Silvana Botti, Miguel A. L. Marques
Format: Article
Language:English
Published: Nature Portfolio 2025-06-01
Series:npj Computational Materials
Online Access:https://doi.org/10.1038/s41524-025-01650-1
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849691314137858048
author Antoine Loew
Dewen Sun
Hai-Chen Wang
Silvana Botti
Miguel A. L. Marques
author_facet Antoine Loew
Dewen Sun
Hai-Chen Wang
Silvana Botti
Miguel A. L. Marques
author_sort Antoine Loew
collection DOAJ
description Abstract There has been an ongoing race for the past several years to develop the best universal machine learning interatomic potential. This progress has led to increasingly accurate models for predicting energy, forces, and stresses, combining innovative architectures with big data. Here, we benchmark these models on their ability to predict harmonic phonon properties, which are critical for understanding the vibrational and thermal behavior of materials. Using around 10 000 ab initio phonon calculations, we evaluate model performance across various phonon-related parameters to test the universal applicability of these models. The results reveal that some models achieve high accuracy in predicting harmonic phonon properties. However, others still exhibit substantial inaccuracies, even if they excel in the prediction of the energy and the forces for materials close to dynamical equilibrium. These findings highlight the importance of considering phonon-related properties in the development of universal machine learning interatomic potentials.
format Article
id doaj-art-182b0b5e787c4e0695981b0a452db577
institution DOAJ
issn 2057-3960
language English
publishDate 2025-06-01
publisher Nature Portfolio
record_format Article
series npj Computational Materials
spelling doaj-art-182b0b5e787c4e0695981b0a452db5772025-08-20T03:21:03ZengNature Portfolionpj Computational Materials2057-39602025-06-011111810.1038/s41524-025-01650-1Universal machine learning interatomic potentials are ready for phononsAntoine Loew0Dewen Sun1Hai-Chen Wang2Silvana Botti3Miguel A. L. Marques4Research Center Future Energy Materials and Systems of the University Alliance Ruhr and Interdisciplinary Centre for Advanced Materials Simulation, Ruhr University Bochum, Universitätsstraße 150Research Center Future Energy Materials and Systems of the University Alliance Ruhr and Interdisciplinary Centre for Advanced Materials Simulation, Ruhr University Bochum, Universitätsstraße 150Research Center Future Energy Materials and Systems of the University Alliance Ruhr and Interdisciplinary Centre for Advanced Materials Simulation, Ruhr University Bochum, Universitätsstraße 150Research Center Future Energy Materials and Systems of the University Alliance Ruhr and Interdisciplinary Centre for Advanced Materials Simulation, Ruhr University Bochum, Universitätsstraße 150Research Center Future Energy Materials and Systems of the University Alliance Ruhr and Interdisciplinary Centre for Advanced Materials Simulation, Ruhr University Bochum, Universitätsstraße 150Abstract There has been an ongoing race for the past several years to develop the best universal machine learning interatomic potential. This progress has led to increasingly accurate models for predicting energy, forces, and stresses, combining innovative architectures with big data. Here, we benchmark these models on their ability to predict harmonic phonon properties, which are critical for understanding the vibrational and thermal behavior of materials. Using around 10 000 ab initio phonon calculations, we evaluate model performance across various phonon-related parameters to test the universal applicability of these models. The results reveal that some models achieve high accuracy in predicting harmonic phonon properties. However, others still exhibit substantial inaccuracies, even if they excel in the prediction of the energy and the forces for materials close to dynamical equilibrium. These findings highlight the importance of considering phonon-related properties in the development of universal machine learning interatomic potentials.https://doi.org/10.1038/s41524-025-01650-1
spellingShingle Antoine Loew
Dewen Sun
Hai-Chen Wang
Silvana Botti
Miguel A. L. Marques
Universal machine learning interatomic potentials are ready for phonons
npj Computational Materials
title Universal machine learning interatomic potentials are ready for phonons
title_full Universal machine learning interatomic potentials are ready for phonons
title_fullStr Universal machine learning interatomic potentials are ready for phonons
title_full_unstemmed Universal machine learning interatomic potentials are ready for phonons
title_short Universal machine learning interatomic potentials are ready for phonons
title_sort universal machine learning interatomic potentials are ready for phonons
url https://doi.org/10.1038/s41524-025-01650-1
work_keys_str_mv AT antoineloew universalmachinelearninginteratomicpotentialsarereadyforphonons
AT dewensun universalmachinelearninginteratomicpotentialsarereadyforphonons
AT haichenwang universalmachinelearninginteratomicpotentialsarereadyforphonons
AT silvanabotti universalmachinelearninginteratomicpotentialsarereadyforphonons
AT miguelalmarques universalmachinelearninginteratomicpotentialsarereadyforphonons