Kevin K. Huguenin-Dumittan¹, Philip Loche¹, Haoran Ni¹, Michele Ceriotti^1*

1 Laboratory of Computational Science and Modelling (COSMO), École Polytechnique Fédérale de Lausanne (EPFL), CH-1015 Lausanne, Vaud, Switzerland

* Corresponding authors emails: michele.ceriotti@epfl.ch

DOI10.24435/materialscloud:23-99 [version v1]

Publication date: Oct 03, 2023

How to cite this record

Kevin K. Huguenin-Dumittan, Philip Loche, Haoran Ni, Michele Ceriotti, Physics-inspired equivariant descriptors of non-bonded interactions, Materials Cloud Archive 2023.151 (2023), https://doi.org/10.24435/materialscloud:23-99

Description

One essential ingredient in many machine learning (ML) based methods for atomistic modeling of materials and molecules is the use of locality. While allowing better system-size scaling, this systematically neglects long-range (LR) effects, such as electrostatics or dispersion interaction. We present an extension of the long distance equivariant (LODE) framework that can handle diverse LR interactions in a consistent way, and seamlessly integrates with preexisting methods by building new sets of atom centered features. We provide a direct physical interpretation of these using the multipole expansion, which allows for simpler and more efficient implementations. The framework is applied to simple toy systems as proof of concept, and a heterogeneous set of molecular dimers to push the method to its limits. By generalizing LODE to arbitrary asymptotic behaviors, we provide a coherent approach to treat arbitrary two- and many-body non-bonded interactions in the data-driven modeling of matter.

Materials Cloud sections using this data

Files

File name	Size	Description
point_charges_Training_set_p1.xyz MD5md5:9e76b4a8972a2e829335163fe31c7d7c	6.8 MiB	Dataset containing the structures and energies of "NaCl" atoms interacting via a pure Coulomb potential
point_charges_Training_set_p6.xyz MD5md5:57c54e0ea0b01be89e0a0622f0f2ff96	6.8 MiB	Dataset containing the structures and energies of atoms interacting via a pure dispersion (1/r^6) potential
bio_dimers.xyz MD5md5:cdf5f3e139bd56a2c31014c8e50e8b93	66.3 MiB	Dataset containing structures and DFT energies & forces of dimers taken from sidechain-sidechain fragments in biomolecules
bio_dimers_monomers.xyz MD5md5:3d4b87ec5e1a79b75c41c48a78f78552	4.9 MiB	The structures and DFT energies & forces of the individual monomers contained in the "bio_dimers.xyz" dataset
bio_dimers_control.in MD5md5:47e973400e61374c26fb2cedbf12838d	16.7 KiB	The FHI-AIMS input file to compute the energies of the "bio_dimers.xyz" and "bio_dimers_monomers.xyz" structures
xenon.xyz MD5md5:9371ed17727e39635fa0c11e27425ace	40.9 KiB	Dataset containing structures and DFT energies & forces of Xenon dimers and trimers
xenon_control.in MD5md5:4ddc62a1c4d83bf68518010daf5903bd	2.6 KiB	The FHI-AIMS input file to compute the energies and forces of the structures in "xenon.xyz"

License

Files and data are licensed under the terms of the following license: Creative Commons Attribution 4.0 International.
Metadata, except for email addresses, are licensed under the Creative Commons Attribution Share-Alike 4.0 International license.

External references

Keywords

ERC machine learning long-range interactions electrostatics dispersion