E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials

Batzner, Simon; Musaelian, Albert; Sun, Lixin; Geiger, Mario; Mailoa, Jonathan P.; Kornbluth, Mordechai; Molinari, Nicola; Smidt, Tess E.; Kozinsky, Boris

doi:10.1038/s41467-022-29939-5

Title: E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials

Abstract

Abstract 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.

Authors:

; Musaelian, Albert; Sun, Lixin; Geiger, Mario; Mailoa, Jonathan P.;

; Molinari, Nicola;

;

Publication Date:: Wed May 04 00:00:00 EDT 2022

Research Org.:: Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)

Sponsoring Org.:: USDOE Office of Science (SC), Advanced Scientific Computing Research (ASCR); USDOE Office of Science (SC), Basic Energy Sciences (BES); National Science Foundation (NSF); US Department of the Navy, Office of Naval Research (ONR); USDOE Advanced Research Projects Agency - Energy (ARPA-E)

OSTI Identifier:: 1866303

Alternate Identifier(s):: OSTI ID: 1880812

Grant/Contract Number:: SC0021110; SC001257; AR0000775; AC02-05CH11231; SC0022199; SC0012573; AC05-00OR22725; DMR-2011754; N00014-20-1-2418; DMR20009; DMR20013

Resource Type:: Published Article

Journal Name:: Nature Communications

Additional Journal Information:: Journal Name: Nature Communications Journal Volume: 13 Journal Issue: 1; Journal ID: ISSN 2041-1723

Publisher:: Nature Publishing Group

Country of Publication:: United Kingdom

Language:: English

Subject:: 37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CHEMISTRY; Atomistic models; Computational chemistry; Computational methods; Computer science; Molecular dynamics

Citation Formats


                    Batzner, Simon, Musaelian, Albert, Sun, Lixin, Geiger, Mario, Mailoa, Jonathan P., Kornbluth, Mordechai, Molinari, Nicola, Smidt, Tess E., and Kozinsky, Boris. E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials.  United Kingdom: N. p., 2022. 
Web.  doi:10.1038/s41467-022-29939-5.

Copy to clipboard


                    Batzner, Simon, Musaelian, Albert, Sun, Lixin, Geiger, Mario, Mailoa, Jonathan P., Kornbluth, Mordechai, Molinari, Nicola, Smidt, Tess E., & Kozinsky, Boris. E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials.  United Kingdom.  https://doi.org/10.1038/s41467-022-29939-5

Copy to clipboard


                    Batzner, Simon, Musaelian, Albert, Sun, Lixin, Geiger, Mario, Mailoa, Jonathan P., Kornbluth, Mordechai, Molinari, Nicola, Smidt, Tess E., and Kozinsky, Boris. Wed .  
"E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials".  United Kingdom.  https://doi.org/10.1038/s41467-022-29939-5.

Copy to clipboard


                    
@article{osti_1866303,

  title        = {E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials},

  author       = {Batzner, Simon and Musaelian, Albert and Sun, Lixin and Geiger, Mario and Mailoa, Jonathan P. and Kornbluth, Mordechai and Molinari, Nicola and Smidt, Tess E. and Kozinsky, Boris},

  abstractNote = {Abstract 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.},

  doi          = {10.1038/s41467-022-29939-5},

  journal      = {Nature Communications},

  number       = 1,

  volume       = 13,

  place        = {United Kingdom},

  year         = {Wed May 04 00:00:00 EDT 2022},

  month        = {Wed May 04 00:00:00 EDT 2022}

}

Copy to clipboard

Journal Article:

Free Publicly Available Full Text

Publisher's Version of Record
https://doi.org/10.1038/s41467-022-29939-5

Other availability

Search WorldCat to find libraries that may hold this journal

Save / Share:

Export Metadata

Save to My Library

Works referenced in this record:

Isotope effects in liquid water via deep potential molecular dynamics
journal, October 2019

Ko, Hsin-Yu; Zhang, Linfeng; Santra, Biswajit
Molecular Physics, Vol. 117, Issue 22
DOI: 10.1080/00268976.2019.1652366

Atomic cluster expansion for accurate and transferable interatomic potentials
journal, January 2019

Drautz, Ralf
Physical Review B, Vol. 99, Issue 1
DOI: 10.1103/PhysRevB.99.014104

ANI-1: an extensible neural network potential with DFT accuracy at force field computational cost
journal, January 2017

Smith, J. S.; Isayev, O.; Roitberg, A. E.
Chemical Science, Vol. 8, Issue 4
DOI: 10.1039/C6SC05720A

NewtonNet: A Newtonian message passing network for deep learning of interatomic potentials and forces
journal, January 2022

Haghighatlari, Mojtaba; Li, Jie; Guan, Xingyi
Digital Discovery
DOI: 10.1039/D2DD00008C

Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set
journal, July 1996

Kresse, G.; Furthmüller, J.
Computational Materials Science, Vol. 6, Issue 1, p. 15-50
DOI: 10.1016/0927-0256(96)00008-0

Towards exact molecular dynamics simulations with machine-learned force fields
journal, September 2018

Chmiela, Stefan; Sauceda, Huziel E.; Müller, Klaus-Robert
Nature Communications, Vol. 9, Issue 1
DOI: 10.1038/s41467-018-06169-2

FCHL revisited: Faster and more accurate quantum machine learning
journal, January 2020

Christensen, Anders S.; Bratholm, Lars A.; Faber, Felix A.
The Journal of Chemical Physics, Vol. 152, Issue 4
DOI: 10.1063/1.5126701

Ab initio thermodynamics of liquid and solid water
journal, January 2019

Cheng, Bingqing; Engel, Edgar A.; Behler, Jörg
Proceedings of the National Academy of Sciences, Vol. 116, Issue 4
DOI: 10.1073/pnas.1815117116

Linear Atomic Cluster Expansion Force Fields for Organic Molecules: Beyond RMSE
journal, November 2021

Kovács, Dávid Péter; Oord, Cas van der; Kucera, Jiri
Journal of Chemical Theory and Computation, Vol. 17, Issue 12
DOI: 10.1021/acs.jctc.1c00647

Understanding the Composition and Activity of Electrocatalytic Nanoalloys in Aqueous Solvents: A Combination of DFT and Accurate Neural Network Potentials
journal, April 2014

Artrith, Nongnuch; Kolpak, Alexie M.
Nano Letters, Vol. 14, Issue 5
DOI: 10.1021/nl5005674

Design and synthesis of the superionic conductor Na10SnP2S12
journal, March 2016

Richards, William D.; Tsujimura, Tomoyuki; Miara, Lincoln J.
Nature Communications, Vol. 7, Issue 1
DOI: 10.1038/ncomms11009

On-the-fly active learning of interpretable Bayesian force fields for atomistic rare events
journal, March 2020

Vandermause, Jonathan; Torrisi, Steven B.; Batzner, Simon
npj Computational Materials, Vol. 6, Issue 1
DOI: 10.1038/s41524-020-0283-z

A fast neural network approach for direct covariant forces prediction in complex multi-element extended systems
journal, September 2019

Mailoa, Jonathan P.; Kornbluth, Mordechai; Batzner, Simon
Nature Machine Intelligence, Vol. 1, Issue 10
DOI: 10.1038/s42256-019-0098-0

The atomic simulation environment—a Python library for working with atoms
journal, June 2017

Hjorth Larsen, Ask; Jørgen Mortensen, Jens; Blomqvist, Jakob
Journal of Physics: Condensed Matter, Vol. 29, Issue 27
DOI: 10.1088/1361-648X/aa680e

Multiple Bonding Configurations of Adsorbed Formate on Ag{111}
journal, January 1996

Sim, W. S.; Gardner, P.; King, D. A.
The Journal of Physical Chemistry, Vol. 100, Issue 30
DOI: 10.1021/jp9600726

How Fast-Folding Proteins Fold
journal, October 2011

Lindorff-Larsen, K.; Piana, S.; Dror, R. O.
Science, Vol. 334, Issue 6055
DOI: 10.1126/science.1208351

Generalized Neural-Network Representation of High-Dimensional Potential-Energy Surfaces
journal, April 2007

Behler, Jörg; Parrinello, Michele
Physical Review Letters, Vol. 98, Issue 14
DOI: 10.1103/PhysRevLett.98.146401

Spectral neighbor analysis method for automated generation of quantum-accurate interatomic potentials
journal, March 2015

Thompson, A. P.; Swiler, L. P.; Trott, C. R.
Journal of Computational Physics, Vol. 285
DOI: 10.1016/j.jcp.2014.12.018

Deep Residual Learning for Image Recognition
conference, June 2016

He, Kaiming; Zhang, Xiangyu; Ren, Shaoqing
2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
DOI: 10.1109/CVPR.2016.90

A Stable Thin-Film Lithium Electrolyte: Lithium Phosphorus Oxynitride
journal, January 1997

Yu, Xiaohua; Bates, J. B.; Jellison, G. E.
Journal of The Electrochemical Society, Vol. 144, Issue 2, p. 524-532
DOI: 10.1149/1.1837443

Accurate and scalable graph neural network force field and molecular dynamics with direct force architecture
journal, May 2021

Park, Cheol Woo; Kornbluth, Mordechai; Vandermause, Jonathan
npj Computational Materials, Vol. 7, Issue 1
DOI: 10.1038/s41524-021-00543-3

Moment Tensor Potentials: A Class of Systematically Improvable Interatomic Potentials
journal, January 2016

Shapeev, Alexander V.
Multiscale Modeling & Simulation, Vol. 14, Issue 3
DOI: 10.1137/15M1054183

Quantum-chemical insights from deep tensor neural networks
journal, January 2017

Schütt, Kristof T.; Arbabzadah, Farhad; Chmiela, Stefan
Nature Communications, Vol. 8, Issue 1
DOI: 10.1038/ncomms13890

SpookyNet: Learning force fields with electronic degrees of freedom and nonlocal effects
journal, December 2021

Unke, Oliver T.; Chmiela, Stefan; Gastegger, Michael
Nature Communications, Vol. 12, Issue 1
DOI: 10.1038/s41467-021-27504-0

Why Is Formate Synthesis Insensitive to Copper Surface Structures?
journal, December 2005

Wang, Guichang; Morikawa, Yoshitada; Matsumoto, Taketoshi
The Journal of Physical Chemistry B, Vol. 110, Issue 1
DOI: 10.1021/jp055689e

PhysNet: A Neural Network for Predicting Energies, Forces, Dipole Moments, and Partial Charges
journal, April 2019

Unke, Oliver T.; Meuwly, Markus
Journal of Chemical Theory and Computation, Vol. 15, Issue 6
DOI: 10.1021/acs.jctc.9b00181

Machine learning of accurate energy-conserving molecular force fields
journal, May 2017

Chmiela, Stefan; Tkatchenko, Alexandre; Sauceda, Huziel E.
Science Advances, Vol. 3, Issue 5
DOI: 10.1126/sciadv.1603015

Deep Potential Molecular Dynamics: A Scalable Model with the Accuracy of Quantum Mechanics
journal, April 2018

Zhang, Linfeng; Han, Jiequn; Wang, Han
Physical Review Letters, Vol. 120, Issue 14
DOI: 10.1103/PhysRevLett.120.143001

Plasma Synthesis of Spherical Crystalline and Amorphous Electrolyte Nanopowders for Solid-State Batteries
journal, February 2020

Westover, Andrew S.; Kercher, Andrew K.; Kornbluth, Mordechai
ACS Applied Materials & Interfaces, Vol. 12, Issue 10
DOI: 10.1021/acsami.9b20812

Study of Li atom diffusion in amorphous Li3PO4 with neural network potential
journal, December 2017

Li, Wenwen; Ando, Yasunobu; Minamitani, Emi
The Journal of Chemical Physics, Vol. 147, Issue 21
DOI: 10.1063/1.4997242

First Principles Molecular Dynamics Study of Ziegler−Natta Heterogeneous Catalysis
journal, March 1998

Boero, Mauro; Parrinello, Michele; Terakura, Kiyoyuki
Journal of the American Chemical Society, Vol. 120, Issue 12
DOI: 10.1021/ja972367i

cp2k: atomistic simulations of condensed matter systems
journal, June 2013

Hutter, Jürg; Iannuzzi, Marcella; Schiffmann, Florian
Wiley Interdisciplinary Reviews: Computational Molecular Science, Vol. 4, Issue 1
DOI: 10.1002/wcms.1159

Extending the Applicability of the ANI Deep Learning Molecular Potential to Sulfur and Halogens
journal, June 2020

Devereux, Christian; Smith, Justin S.; Davis, Kate K.
Journal of Chemical Theory and Computation, Vol. 16, Issue 7
DOI: 10.1021/acs.jctc.0c00121

Similar Records in DOE PAGES and OSTI.GOV collections:

Learning local equivariant representations for large-scale atomistic dynamics

Journal Article Musaelian, Albert ; Batzner, Simon ; Johansson, Anders ; ... - Nature Communications

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-artmore »« less
https://doi.org/10.1038/s41467-023-36329-y
Data efficiency and extrapolation trends in neural network interatomic potentials

Journal Article Vita, Joshua A. ; Schwalbe-Koda, Daniel - Machine Learning: Science and Technology

Abstract Recently, key architectural advances have been proposed for neural network interatomic potentials (NNIPs), such as incorporating message-passing networks, equivariance, or many-body expansion terms. Although modern NNIP models exhibit small differences in test accuracy, this metric is still considered the main target when developing new NNIP architectures. In this work, we show how architectural and optimization choices influence the generalization of NNIPs, revealing trends in molecular dynamics (MD) stability, data efficiency, and loss landscapes. Using the 3BPA dataset, we uncover trends in NNIP errors and robustness to noise, showing these metrics are insufficient to predict MD stability in the high-accuracymore »« less
https://doi.org/10.1088/2632-2153/acf115
Combining protein sequences and structures with transformers and equivariant graph neural networks to predict protein function

Journal Article Boadu, Frimpong ; Cao, Hongyuan ; Cheng, Jianlin - Bioinformatics

Abstract Motivation Millions of protein sequences have been generated by numerous genome and transcriptome sequencing projects. However, experimentally determining the function of the proteins is still a time consuming, low-throughput, and expensive process, leading to a large protein sequence-function gap. Therefore, it is important to develop computational methods to accurately predict protein function to fill the gap. Even though many methods have been developed to use protein sequences as input to predict function, much fewer methods leverage protein structures in protein function prediction because there was lack of accurate protein structures for most proteins until recently. Results We developed TransFun—amore »« less
https://doi.org/10.1093/bioinformatics/btad208
Lorentz group equivariant autoencoders

Journal Article Hao, Zichun ; Kansal, Raghav ; Duarte, Javier ; ... - European Physical Journal. C, Particles and Fields (Online)

Absmore »« less
https://doi.org/10.1140/epjc/s10052-023-11633-5
Hierarchical, rotation‐equivariant neural networks to select structural models of protein complexes

Journal Article Eismann, Stephan ; Townshend, Raphael J. L. ; Thomas, Nathaniel ; ... - Proteins

Abstract Predicting the structure of multi‐protein complexes is a grand challenge in biochemistry, with major implications for basic science and drug discovery. Computational structure prediction methods generally leverage predefined structural features to distinguish accurate structural models from less accurate ones. This raises the question of whether it is possible to learn characteristics of accurate models directly from atomic coordinates of protein complexes, with no prior assumptions. Here we introduce a machine learning method that learns directly from the 3D positions of all atoms to identify accurate models of protein complexes, without using any precomputed physics‐inspired or statistical terms. Our neuralmore »« less
https://doi.org/10.1002/prot.26033

Similar Records

Title: E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials

Abstract

Citation Formats

Isotope effects in liquid water via deep potential molecular dynamics journal, October 2019

Atomic cluster expansion for accurate and transferable interatomic potentials journal, January 2019

ANI-1: an extensible neural network potential with DFT accuracy at force field computational cost journal, January 2017

NewtonNet: A Newtonian message passing network for deep learning of interatomic potentials and forces journal, January 2022

Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set journal, July 1996

Towards exact molecular dynamics simulations with machine-learned force fields journal, September 2018

FCHL revisited: Faster and more accurate quantum machine learning journal, January 2020

Ab initio thermodynamics of liquid and solid water journal, January 2019

Linear Atomic Cluster Expansion Force Fields for Organic Molecules: Beyond RMSE journal, November 2021

Understanding the Composition and Activity of Electrocatalytic Nanoalloys in Aqueous Solvents: A Combination of DFT and Accurate Neural Network Potentials journal, April 2014

Design and synthesis of the superionic conductor Na10SnP2S12 journal, March 2016

On-the-fly active learning of interpretable Bayesian force fields for atomistic rare events journal, March 2020

A fast neural network approach for direct covariant forces prediction in complex multi-element extended systems journal, September 2019

The atomic simulation environment—a Python library for working with atoms journal, June 2017

Multiple Bonding Configurations of Adsorbed Formate on Ag{111} journal, January 1996

How Fast-Folding Proteins Fold journal, October 2011

Generalized Neural-Network Representation of High-Dimensional Potential-Energy Surfaces journal, April 2007

Spectral neighbor analysis method for automated generation of quantum-accurate interatomic potentials journal, March 2015

Deep Residual Learning for Image Recognition conference, June 2016

A Stable Thin-Film Lithium Electrolyte: Lithium Phosphorus Oxynitride journal, January 1997

Accurate and scalable graph neural network force field and molecular dynamics with direct force architecture journal, May 2021

Moment Tensor Potentials: A Class of Systematically Improvable Interatomic Potentials journal, January 2016

Quantum-chemical insights from deep tensor neural networks journal, January 2017

SpookyNet: Learning force fields with electronic degrees of freedom and nonlocal effects journal, December 2021

Why Is Formate Synthesis Insensitive to Copper Surface Structures? journal, December 2005

PhysNet: A Neural Network for Predicting Energies, Forces, Dipole Moments, and Partial Charges journal, April 2019

Machine learning of accurate energy-conserving molecular force fields journal, May 2017

Deep Potential Molecular Dynamics: A Scalable Model with the Accuracy of Quantum Mechanics journal, April 2018

Plasma Synthesis of Spherical Crystalline and Amorphous Electrolyte Nanopowders for Solid-State Batteries journal, February 2020

Study of Li atom diffusion in amorphous Li3PO4 with neural network potential journal, December 2017

First Principles Molecular Dynamics Study of Ziegler−Natta Heterogeneous Catalysis journal, March 1998

cp2k: atomistic simulations of condensed matter systems journal, June 2013

Extending the Applicability of the ANI Deep Learning Molecular Potential to Sulfur and Halogens journal, June 2020

Isotope effects in liquid water via deep potential molecular dynamics
journal, October 2019

Atomic cluster expansion for accurate and transferable interatomic potentials
journal, January 2019

ANI-1: an extensible neural network potential with DFT accuracy at force field computational cost
journal, January 2017

NewtonNet: A Newtonian message passing network for deep learning of interatomic potentials and forces
journal, January 2022

Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set
journal, July 1996

Towards exact molecular dynamics simulations with machine-learned force fields
journal, September 2018

FCHL revisited: Faster and more accurate quantum machine learning
journal, January 2020

Ab initio thermodynamics of liquid and solid water
journal, January 2019

Linear Atomic Cluster Expansion Force Fields for Organic Molecules: Beyond RMSE
journal, November 2021

Understanding the Composition and Activity of Electrocatalytic Nanoalloys in Aqueous Solvents: A Combination of DFT and Accurate Neural Network Potentials
journal, April 2014

Design and synthesis of the superionic conductor Na10SnP2S12
journal, March 2016

On-the-fly active learning of interpretable Bayesian force fields for atomistic rare events
journal, March 2020

A fast neural network approach for direct covariant forces prediction in complex multi-element extended systems
journal, September 2019

The atomic simulation environment—a Python library for working with atoms
journal, June 2017

Multiple Bonding Configurations of Adsorbed Formate on Ag{111}
journal, January 1996

How Fast-Folding Proteins Fold
journal, October 2011

Generalized Neural-Network Representation of High-Dimensional Potential-Energy Surfaces
journal, April 2007

Spectral neighbor analysis method for automated generation of quantum-accurate interatomic potentials
journal, March 2015

Deep Residual Learning for Image Recognition
conference, June 2016

A Stable Thin-Film Lithium Electrolyte: Lithium Phosphorus Oxynitride
journal, January 1997

Accurate and scalable graph neural network force field and molecular dynamics with direct force architecture
journal, May 2021

Moment Tensor Potentials: A Class of Systematically Improvable Interatomic Potentials
journal, January 2016

Quantum-chemical insights from deep tensor neural networks
journal, January 2017

SpookyNet: Learning force fields with electronic degrees of freedom and nonlocal effects
journal, December 2021

Why Is Formate Synthesis Insensitive to Copper Surface Structures?
journal, December 2005

PhysNet: A Neural Network for Predicting Energies, Forces, Dipole Moments, and Partial Charges
journal, April 2019

Machine learning of accurate energy-conserving molecular force fields
journal, May 2017

Deep Potential Molecular Dynamics: A Scalable Model with the Accuracy of Quantum Mechanics
journal, April 2018

Plasma Synthesis of Spherical Crystalline and Amorphous Electrolyte Nanopowders for Solid-State Batteries
journal, February 2020

Study of Li atom diffusion in amorphous Li3PO4 with neural network potential
journal, December 2017

First Principles Molecular Dynamics Study of Ziegler−Natta Heterogeneous Catalysis
journal, March 1998

cp2k: atomistic simulations of condensed matter systems
journal, June 2013

Extending the Applicability of the ANI Deep Learning Molecular Potential to Sulfur and Halogens
journal, June 2020