SchNet - A deep learning architecture for molecules and materials

K. T. Schütt, H. E. Sauceda, P. J. Kindermans, A. Tkatchenko, Klaus Muller

Research output: Contribution to journalArticle

101 Citations (Scopus)

Abstract

Deep learning has led to a paradigm shift in artificial intelligence, including web, text, and image search, speech recognition, as well as bioinformatics, with growing impact in chemical physics. Machine learning, in general, and deep learning, in particular, are ideally suitable for representing quantum-mechanical interactions, enabling us to model nonlinear potential-energy surfaces or enhancing the exploration of chemical compound space. Here we present the deep learning architecture SchNet that is specifically designed to model atomistic systems by making use of continuous-filter convolutional layers. We demonstrate the capabilities of SchNet by accurately predicting a range of properties across chemical space for molecules and materials, where our model learns chemically plausible embeddings of atom types across the periodic table. Finally, we employ SchNet to predict potential-energy surfaces and energy-conserving force fields for molecular dynamics simulations of small molecules and perform an exemplary study on the quantum-mechanical properties of C20-fullerene that would have been infeasible with regular ab initio molecular dynamics.

Original languageEnglish
Article number241722
JournalJournal of Chemical Physics
Volume148
Issue number24
DOIs
Publication statusPublished - 2018 Jun 28

Fingerprint

learning
Potential energy surfaces
Molecules
Molecular dynamics
potential energy
molecular dynamics
Fullerenes
molecules
artificial intelligence
chemical compounds
machine learning
Chemical compounds
speech recognition
Bioinformatics
Speech recognition
chemical properties
embedding
Chemical properties
field theory (physics)
fullerenes

ASJC Scopus subject areas

  • Physics and Astronomy(all)
  • Physical and Theoretical Chemistry

Cite this

Schütt, K. T., Sauceda, H. E., Kindermans, P. J., Tkatchenko, A., & Muller, K. (2018). SchNet - A deep learning architecture for molecules and materials. Journal of Chemical Physics, 148(24), [241722]. https://doi.org/10.1063/1.5019779

SchNet - A deep learning architecture for molecules and materials. / Schütt, K. T.; Sauceda, H. E.; Kindermans, P. J.; Tkatchenko, A.; Muller, Klaus.

In: Journal of Chemical Physics, Vol. 148, No. 24, 241722, 28.06.2018.

Research output: Contribution to journalArticle

Schütt, KT, Sauceda, HE, Kindermans, PJ, Tkatchenko, A & Muller, K 2018, 'SchNet - A deep learning architecture for molecules and materials', Journal of Chemical Physics, vol. 148, no. 24, 241722. https://doi.org/10.1063/1.5019779
Schütt KT, Sauceda HE, Kindermans PJ, Tkatchenko A, Muller K. SchNet - A deep learning architecture for molecules and materials. Journal of Chemical Physics. 2018 Jun 28;148(24). 241722. https://doi.org/10.1063/1.5019779
Schütt, K. T. ; Sauceda, H. E. ; Kindermans, P. J. ; Tkatchenko, A. ; Muller, Klaus. / SchNet - A deep learning architecture for molecules and materials. In: Journal of Chemical Physics. 2018 ; Vol. 148, No. 24.
@article{3d1111221eba43a69821b307ba5fbf28,
title = "SchNet - A deep learning architecture for molecules and materials",
abstract = "Deep learning has led to a paradigm shift in artificial intelligence, including web, text, and image search, speech recognition, as well as bioinformatics, with growing impact in chemical physics. Machine learning, in general, and deep learning, in particular, are ideally suitable for representing quantum-mechanical interactions, enabling us to model nonlinear potential-energy surfaces or enhancing the exploration of chemical compound space. Here we present the deep learning architecture SchNet that is specifically designed to model atomistic systems by making use of continuous-filter convolutional layers. We demonstrate the capabilities of SchNet by accurately predicting a range of properties across chemical space for molecules and materials, where our model learns chemically plausible embeddings of atom types across the periodic table. Finally, we employ SchNet to predict potential-energy surfaces and energy-conserving force fields for molecular dynamics simulations of small molecules and perform an exemplary study on the quantum-mechanical properties of C20-fullerene that would have been infeasible with regular ab initio molecular dynamics.",
author = "Sch{\"u}tt, {K. T.} and Sauceda, {H. E.} and Kindermans, {P. J.} and A. Tkatchenko and Klaus Muller",
year = "2018",
month = "6",
day = "28",
doi = "10.1063/1.5019779",
language = "English",
volume = "148",
journal = "Journal of Chemical Physics",
issn = "0021-9606",
publisher = "American Institute of Physics Publising LLC",
number = "24",

}

TY - JOUR

T1 - SchNet - A deep learning architecture for molecules and materials

AU - Schütt, K. T.

AU - Sauceda, H. E.

AU - Kindermans, P. J.

AU - Tkatchenko, A.

AU - Muller, Klaus

PY - 2018/6/28

Y1 - 2018/6/28

N2 - Deep learning has led to a paradigm shift in artificial intelligence, including web, text, and image search, speech recognition, as well as bioinformatics, with growing impact in chemical physics. Machine learning, in general, and deep learning, in particular, are ideally suitable for representing quantum-mechanical interactions, enabling us to model nonlinear potential-energy surfaces or enhancing the exploration of chemical compound space. Here we present the deep learning architecture SchNet that is specifically designed to model atomistic systems by making use of continuous-filter convolutional layers. We demonstrate the capabilities of SchNet by accurately predicting a range of properties across chemical space for molecules and materials, where our model learns chemically plausible embeddings of atom types across the periodic table. Finally, we employ SchNet to predict potential-energy surfaces and energy-conserving force fields for molecular dynamics simulations of small molecules and perform an exemplary study on the quantum-mechanical properties of C20-fullerene that would have been infeasible with regular ab initio molecular dynamics.

AB - Deep learning has led to a paradigm shift in artificial intelligence, including web, text, and image search, speech recognition, as well as bioinformatics, with growing impact in chemical physics. Machine learning, in general, and deep learning, in particular, are ideally suitable for representing quantum-mechanical interactions, enabling us to model nonlinear potential-energy surfaces or enhancing the exploration of chemical compound space. Here we present the deep learning architecture SchNet that is specifically designed to model atomistic systems by making use of continuous-filter convolutional layers. We demonstrate the capabilities of SchNet by accurately predicting a range of properties across chemical space for molecules and materials, where our model learns chemically plausible embeddings of atom types across the periodic table. Finally, we employ SchNet to predict potential-energy surfaces and energy-conserving force fields for molecular dynamics simulations of small molecules and perform an exemplary study on the quantum-mechanical properties of C20-fullerene that would have been infeasible with regular ab initio molecular dynamics.

UR - http://www.scopus.com/inward/record.url?scp=85044731105&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85044731105&partnerID=8YFLogxK

U2 - 10.1063/1.5019779

DO - 10.1063/1.5019779

M3 - Article

C2 - 29960322

AN - SCOPUS:85044731105

VL - 148

JO - Journal of Chemical Physics

JF - Journal of Chemical Physics

SN - 0021-9606

IS - 24

M1 - 241722

ER -