Deep potential: A general representation of a many-body potential energy surface

Jiequn Han; Linfeng Zhang; Roberto Car; E. Weinan

doi:10.4208/CICP.OA-2017-0213

Deep potential: A general representation of a many-body potential energy surface

Jiequn Han, Linfeng Zhang, Roberto Car, E. Weinan

Research output: Contribution to journal › Article › peer-review

182 Scopus citations

Abstract

We present a simple, yet general, deep neural network representation of the potential energy surface for atomic and molecular systems. It is “first-principle” based, in the sense that no ad hoc approximations or empirical fitting functions are required. When tested on a wide variety of examples, it reproduces the original model within chemical accuracy. This brings us one step closer to carrying out molecular simulations with quantum mechanics accuracy at empirical potential computational cost.

Original language	English (US)
Pages (from-to)	629-639
Number of pages	11
Journal	Communications in Computational Physics
Volume	23
Issue number	3
DOIs	https://doi.org/10.4208/CICP.OA-2017-0213
State	Published - 2018

All Science Journal Classification (ASJC) codes

Physics and Astronomy (miscellaneous)

Keywords

Deep learning
Molecular simulation
Potential energy surface

Access to Document

10.4208/CICP.OA-2017-0213

Cite this

@article{c908b0cd00fd4ca6aa6edf741c8bf648,

title = "Deep potential: A general representation of a many-body potential energy surface",

abstract = "We present a simple, yet general, deep neural network representation of the potential energy surface for atomic and molecular systems. It is “first-principle” based, in the sense that no ad hoc approximations or empirical fitting functions are required. When tested on a wide variety of examples, it reproduces the original model within chemical accuracy. This brings us one step closer to carrying out molecular simulations with quantum mechanics accuracy at empirical potential computational cost.",

keywords = "Deep learning, Molecular simulation, Potential energy surface",

author = "Jiequn Han and Linfeng Zhang and Roberto Car and E. Weinan",

note = "Publisher Copyright: {\textcopyright} 2018 Global-Science Press",

year = "2018",

doi = "10.4208/CICP.OA-2017-0213",

language = "English (US)",

volume = "23",

pages = "629--639",

journal = "Communications in Computational Physics",

issn = "1815-2406",

publisher = "Global Science Press",

number = "3",

}

TY - JOUR

T1 - Deep potential

T2 - A general representation of a many-body potential energy surface

AU - Han, Jiequn

AU - Zhang, Linfeng

AU - Car, Roberto

AU - Weinan, E.

PY - 2018

Y1 - 2018

N2 - We present a simple, yet general, deep neural network representation of the potential energy surface for atomic and molecular systems. It is “first-principle” based, in the sense that no ad hoc approximations or empirical fitting functions are required. When tested on a wide variety of examples, it reproduces the original model within chemical accuracy. This brings us one step closer to carrying out molecular simulations with quantum mechanics accuracy at empirical potential computational cost.

AB - We present a simple, yet general, deep neural network representation of the potential energy surface for atomic and molecular systems. It is “first-principle” based, in the sense that no ad hoc approximations or empirical fitting functions are required. When tested on a wide variety of examples, it reproduces the original model within chemical accuracy. This brings us one step closer to carrying out molecular simulations with quantum mechanics accuracy at empirical potential computational cost.

KW - Deep learning

KW - Molecular simulation

KW - Potential energy surface

UR - http://www.scopus.com/inward/record.url?scp=85093578702&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85093578702&partnerID=8YFLogxK

U2 - 10.4208/CICP.OA-2017-0213

DO - 10.4208/CICP.OA-2017-0213

M3 - Article

AN - SCOPUS:85093578702

SN - 1815-2406

VL - 23

SP - 629

EP - 639

JO - Communications in Computational Physics

JF - Communications in Computational Physics

IS - 3

ER -

Deep potential: A general representation of a many-body potential energy surface

Abstract

All Science Journal Classification (ASJC) codes

Keywords

Access to Document

Other files and links

Fingerprint

Cite this