Representing spherical tensors with scalar-based machine-learning models
The Journal of Chemical Physics · 28.10.2025 · 163 (16)
DOI: 10.1063/5.0284802 ↗Kurz gefasst
Fachartikel in der Zeitschrift The Journal of Chemical Physics, online erschienen am 23.10.2025. Verfasst von M Domina und 3 weiteren Autor:innen. Beteiligte Einrichtung: Laboratory of Computational Science and Modeling, Institut des Matériaux, École Polytechnique Fédérale de Lausanne, 1015 Lausanne, Switzerland. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 56 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: HORIZON EUROPE European Research Council, National Center of Competence in Research Materials' Revolution: Computational Design and Discovery of Novel Materials.
Abstract (Originalsprache)
Rotational symmetry plays a central role in physics, providing an elegant framework to describe how the properties of 3D objects—from atoms to the macroscopic scale—transform under the action of rigid rotations. Equivariant models of 3D point clouds are able to approximate structure–property relations in a way that is fully consistent with the structure of the rotation group by combining intermediate representations that are themselves spherical tensors. The symmetry constraints, however, make this approach computationally demanding and cumbersome to implement, which motivates increasingly popular unconstrained architectures that learn approximate symmetries as part of the training process. In this work, we explore a third route to tackle this learning problem, where equivariant functions are expressed as the product of a scalar function of the point cloud coordinates and a small basis of tensors with the appropriate symmetry. In particular, we show that it is always possible to separate the learning of an equivariant property into learnable scalars and fixed geometric terms built as the maximal coupling of interatomic vectors. We also propose approximations of the general expressions that, while lacking universal approximation properties, are fast, simple to implement, and accurate in practical settings.
Autor:innen und Institutionen
- Laboratory of Computational Science and Modeling, Institut des Matériaux, École Polytechnique Fédérale de Lausanne, 1015 Lausanne, Switzerland.
ORCID-Profile: M Domina, F Bigi, P Pegolo, M Ceriotti
Förderung
- HORIZON EUROPE European Research Council
- National Center of Competence in Research Materials' Revolution: Computational Design and Discovery of Novel Materials
Zitieren
APA
Domina, M., Bigi, F., Pegolo, P., & Ceriotti, M. (2025). Representing spherical tensors with scalar-based machine-learning models. The Journal of Chemical Physics, 163(16). https://doi.org/10.1063/5.0284802
Vancouver
Domina M, Bigi F, Pegolo P, Ceriotti M. Representing spherical tensors with scalar-based machine-learning models. The Journal of Chemical Physics. 2025;163(16). doi:10.1063/5.0284802
BibTeX
@article{Domina2025,
author = {Domina, M and Bigi, F and Pegolo, P and Ceriotti, M},
title = {Representing spherical tensors with scalar-based machine-learning models},
journal = {The Journal of Chemical Physics},
year = {2025},
volume = {163},
number = {16},
doi = {10.1063/5.0284802}
}
EUVIMED