Bypassing the Kohn-Sham equations with machine learning

Felix Brockherde, Leslie Vogt, Li Li, Mark Tuckerman, Kieron Burke, Klaus Robert Müller

Research output: Contribution to journalArticle

Abstract

Last year, at least 30,000 scientific papers used the Kohn-Sham scheme of density functional theory to solve electronic structure problems in a wide variety of scientific fields. Machine learning holds the promise of learning the energy functional via examples, bypassing the need to solve the Kohn-Sham equations. This should yield substantial savings in computer time, allowing larger systems and/or longer time-scales to be tackled, but attempts to machine-learn this functional have been limited by the need to find its derivative. The present work overcomes this difficulty by directly learning the density-potential and energy-density maps for test systems and various molecules. We perform the first molecular dynamics simulation with a machine-learned density functional on malonaldehyde and are able to capture the intramolecular proton transfer process. Learning density models now allows the construction of accurate density functionals for realistic molecular systems.

Original languageEnglish (US)
Article number872
JournalNature Communications
Volume8
Issue number1
DOIs
StatePublished - Dec 1 2017

Fingerprint

machine learning
Proton transfer
Malondialdehyde
Electronic structure
Density functional theory
Learning systems
Molecular dynamics
Learning
learning
Derivatives
Molecules
Computer simulation
Molecular Dynamics Simulation
Protons
functionals
flux density
molecular dynamics
density functional theory
electronic structure
protons

ASJC Scopus subject areas

  • Chemistry(all)
  • Biochemistry, Genetics and Molecular Biology(all)
  • Physics and Astronomy(all)

Cite this

Bypassing the Kohn-Sham equations with machine learning. / Brockherde, Felix; Vogt, Leslie; Li, Li; Tuckerman, Mark; Burke, Kieron; Müller, Klaus Robert.

In: Nature Communications, Vol. 8, No. 1, 872, 01.12.2017.

Research output: Contribution to journalArticle

Brockherde, Felix ; Vogt, Leslie ; Li, Li ; Tuckerman, Mark ; Burke, Kieron ; Müller, Klaus Robert. / Bypassing the Kohn-Sham equations with machine learning. In: Nature Communications. 2017 ; Vol. 8, No. 1.
@article{cdab4673530c42d399c692be2e750b9d,
title = "Bypassing the Kohn-Sham equations with machine learning",
abstract = "Last year, at least 30,000 scientific papers used the Kohn-Sham scheme of density functional theory to solve electronic structure problems in a wide variety of scientific fields. Machine learning holds the promise of learning the energy functional via examples, bypassing the need to solve the Kohn-Sham equations. This should yield substantial savings in computer time, allowing larger systems and/or longer time-scales to be tackled, but attempts to machine-learn this functional have been limited by the need to find its derivative. The present work overcomes this difficulty by directly learning the density-potential and energy-density maps for test systems and various molecules. We perform the first molecular dynamics simulation with a machine-learned density functional on malonaldehyde and are able to capture the intramolecular proton transfer process. Learning density models now allows the construction of accurate density functionals for realistic molecular systems.",
author = "Felix Brockherde and Leslie Vogt and Li Li and Mark Tuckerman and Kieron Burke and M{\"u}ller, {Klaus Robert}",
year = "2017",
month = "12",
day = "1",
doi = "10.1038/s41467-017-00839-3",
language = "English (US)",
volume = "8",
journal = "Nature Communications",
issn = "2041-1723",
publisher = "Nature Publishing Group",
number = "1",

}

TY - JOUR

T1 - Bypassing the Kohn-Sham equations with machine learning

AU - Brockherde, Felix

AU - Vogt, Leslie

AU - Li, Li

AU - Tuckerman, Mark

AU - Burke, Kieron

AU - Müller, Klaus Robert

PY - 2017/12/1

Y1 - 2017/12/1

N2 - Last year, at least 30,000 scientific papers used the Kohn-Sham scheme of density functional theory to solve electronic structure problems in a wide variety of scientific fields. Machine learning holds the promise of learning the energy functional via examples, bypassing the need to solve the Kohn-Sham equations. This should yield substantial savings in computer time, allowing larger systems and/or longer time-scales to be tackled, but attempts to machine-learn this functional have been limited by the need to find its derivative. The present work overcomes this difficulty by directly learning the density-potential and energy-density maps for test systems and various molecules. We perform the first molecular dynamics simulation with a machine-learned density functional on malonaldehyde and are able to capture the intramolecular proton transfer process. Learning density models now allows the construction of accurate density functionals for realistic molecular systems.

AB - Last year, at least 30,000 scientific papers used the Kohn-Sham scheme of density functional theory to solve electronic structure problems in a wide variety of scientific fields. Machine learning holds the promise of learning the energy functional via examples, bypassing the need to solve the Kohn-Sham equations. This should yield substantial savings in computer time, allowing larger systems and/or longer time-scales to be tackled, but attempts to machine-learn this functional have been limited by the need to find its derivative. The present work overcomes this difficulty by directly learning the density-potential and energy-density maps for test systems and various molecules. We perform the first molecular dynamics simulation with a machine-learned density functional on malonaldehyde and are able to capture the intramolecular proton transfer process. Learning density models now allows the construction of accurate density functionals for realistic molecular systems.

UR - http://www.scopus.com/inward/record.url?scp=85031128428&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85031128428&partnerID=8YFLogxK

U2 - 10.1038/s41467-017-00839-3

DO - 10.1038/s41467-017-00839-3

M3 - Article

VL - 8

JO - Nature Communications

JF - Nature Communications

SN - 2041-1723

IS - 1

M1 - 872

ER -