Simon Batzner^1*, Albert Musaelian^1*, Lixin Sun^1*, Mario Geiger^2*, Jonathan P. Mailoa^3*, Mordechai Kornbluth^3*, Nicola Molinari^1*, Tess E. Smidt^4,5*, Boris Kozinsky^1,3*

1 John A. Paulson School of Engineering and Applied Sciences, Harvard University, Cambridge, MA 02138, USA

2 Research Laboratory of Electronics, Massachusetts Institute of Technology, Cambridge, MA 02139, USA

3 Robert Bosch Research and Technology Center, Cambridge, MA 02139, USA

4 Computational Research Division and Center for Advanced Mathematics for Energy Research Applications, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA

5 Department of Electrical Engineering and Computer Science and Research Laboratory of Electronics, Massachusetts Institute of Technology, Cambridge, MA 02139, USA

* Corresponding authors emails: batzner@g.harvard.edu, albym@seas.harvard.edu, lixinsun@g.harvard.edu, geiger.mario@gmail.com, jpmailoa@gmail.com, Mordechai.Kornbluth@us.bosch.com, nmolinari@seas.harvard.edu, tsmidt@mit.edu, bkoz@seas.harvard.edu

DOI10.24435/materialscloud:s0-5n [version v1]

Publication date: Mar 30, 2022

How to cite this record

Simon Batzner, Albert Musaelian, Lixin Sun, Mario Geiger, Jonathan P. Mailoa, Mordechai Kornbluth, Nicola Molinari, Tess E. Smidt, Boris Kozinsky, E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials, Materials Cloud Archive 2022.45 (2022), https://doi.org/10.24435/materialscloud:s0-5n

Description

This work presents Neural Equivariant Interatomic Potentials (NequIP), an E(3)-equivariant neural network approach for learning interatomic potentials from ab-initio calculations for molecular dynamics simulations. While most contemporary symmetry-aware models use invariant convolutions and only act on scalars, NequIP employs E(3)-equivariant convolutions for interactions of geometric tensors, resulting in a more information-rich and faithful representation of atomic environments. The method achieves state-of-the-art accuracy on a challenging and diverse set of molecules and materials while exhibiting remarkable data efficiency. NequIP outperforms existing models with up to three orders of magnitude fewer training data, challenging the widely held belief that deep neural networks require massive training sets. The high data efficiency of the method allows for the construction of accurate potentials using high-order quantum chemical level of theory as reference and enables high-fidelity molecular dynamics simulations over long time scales.

Materials Cloud sections using this data

Files

File name	Size	Description
lipo-quench.xyz MD5md5:d436cfb166276304c8acf9ffe8f36738	524.9 MiB	Li4P2O7 quench data set
lips.xyz MD5md5:16f91b4aca0b5759026f8bd62790e5c6	212.6 MiB	Li6.75P3S11 data set
fcu.xyz MD5md5:eb0224ff48b6cf3592ec96de0a1548b2	36.6 MiB	Formate on Cu dataset
README.md MD5md5:aa22b3ba42b42ed12fc05dea78b6ba01	558 Bytes	README file.

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

machine learning molecular dynamics density-functional theory