Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2023 Feb 3;14(1):579.
doi: 10.1038/s41467-023-36329-y.

Learning local equivariant representations for large-scale atomistic dynamics

Albert Musaelian #  1 , Simon Batzner #  2 , Anders Johansson  1 , Lixin Sun  1 , Cameron J Owen  1 , Mordechai Kornbluth  3 , Boris Kozinsky  4   5
Affiliations

Learning local equivariant representations for large-scale atomistic dynamics

Albert Musaelian et al. Nat Commun. .

Abstract

A simultaneously accurate and computationally efficient parametrization of the potential energy surface of molecules and materials is a long-standing goal in the natural sciences. While atom-centered message passing neural networks (MPNNs) have shown remarkable accuracy, their information propagation has limited the accessible length-scales. Local methods, conversely, scale to large simulations but have suffered from inferior accuracy. This work introduces Allegro, a strictly local equivariant deep neural network interatomic potential architecture that simultaneously exhibits excellent accuracy and scalability. Allegro represents a many-body potential using iterated tensor products of learned equivariant representations without atom-centered message passing. Allegro obtains improvements over state-of-the-art methods on QM9 and revMD17. A single tensor product layer outperforms existing deep MPNNs and transformers on QM9. Furthermore, Allegro displays remarkable generalization to out-of-distribution data. Molecular simulations using Allegro recover structural and kinetic properties of an amorphous electrolyte in excellent agreement with ab-initio simulations. Finally, we demonstrate parallelization with a simulation of 100 million atoms.

PubMed Disclaimer

Conflict of interest statement

The authors declare no competing interests.

Figures

Fig. 1
Fig. 1. The Allegro network.
a shows the Allegro model architecture and b details a tensor product layer. Blue and red arrows represent scalar and tensor information, respectively, ⊗ denotes the tensor product, and ⊕ is concatenation.
Fig. 2
Fig. 2. Structural properties of Li3PO4.
Left: radial distribution function, right: angular distribution function of tetrahedral bond angle. All defined as probability density functions. Results from Allegro are shown in red, and those from AIMD are shown in black.
Fig. 3
Fig. 3. Li dynamics in Li3PO4.
Comparison of the Li MSD of AIMD vs. Allegro. Results are averaged over 10 runs of Allegro, shading indicates +/– one standard deviation. Results from Allegro are shown in red, and those from AIMD are shown in blue.
Fig. 4
Fig. 4. Structure of Li3PO4.
The quenched Li3PO4 structure at T = 600 K.
Fig. 5
Fig. 5. Scaling results.
Strong scaling results on a Li3PO4 structure of 421,824 atoms, performed in LAMMPS.
See this image and copyright information in PMC

References

    1. Richards WD, et al. Design and synthesis of the superionic conductor na 10 snp 2 s 12. Nat. Commun. 2016;7:1–8. - PMC - PubMed
    1. Lindorff-Larsen K, Piana S, Dror RO, Shaw DE. How fast-folding proteins fold. Science. 2011;334:517–520. - PubMed
    1. Blank TB, Brown SD, Calhoun AW, Doren DJ. Neural network models of potential energy surfaces. J. Chem. Phys. 1995;103:4129–4137.
    1. Handley CM, Hawe GI, Kell DB, Popelier PL. Optimal construction of a fast and accurate polarisable water potential based on multipole moments trained by machine learning. Phys. Chem. Chem. Phys. 2009;11:6365–6376. - PubMed
    1. Behler J, Parrinello M. Generalized neural-network representation of high-dimensional potential-energy surfaces. Phys. Rev. Lett. 2007;98:146401. - PubMed