Thermodynamic integration by neural network potentials based on first-principles dynamic calculations

Shogo Fukushima; Eisaku Ushijima; Hiroyuki Kumazoe; Akihide Koura; Fuyuki Shimojo; Kohei Shimamura; Masaaki Misawa; Rajiv K. Kalia; Aiichiro Nakano; Priya Vashishta

doi:10.1103/PhysRevB.100.214108

Thermodynamic integration by neural network potentials based on first-principles dynamic calculations

Shogo Fukushima, Eisaku Ushijima, Hiroyuki Kumazoe, Akihide Koura, Fuyuki Shimojo, Kohei Shimamura, Masaaki Misawa, Rajiv K. Kalia, Aiichiro Nakano, Priya Vashishta

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

8 Citations (Scopus)

Abstract

Simulation-size effect in evaluating the melting temperature of material is studied systematically by combining thermodynamic integration (TI) based on first-principles molecular-dynamics (FPMD) simulations and machine learning. Since the numerical integration to determine the free energies of two different phases as a function of temperature is very time consuming, the FPMD-based TI method has only been applied to small systems, i.e., less than 100 atoms. To accelerate the numerical integration, we here construct an interatomic potential based on the artificial neural-network (ANN) method, which retains the first-principles accuracy at a significantly lower computational cost. The free energies of the solid and liquid phases of rubidium are accurately obtained by the ANN potential, where its weight parameters are optimized to reproduce FPMD results. The ANN results reveal a significant size dependence up to 500 atoms.

Original language	English
Article number	214108
Journal	Physical Review B
Volume	100
Issue number	21
DOIs	https://doi.org/10.1103/PhysRevB.100.214108
Publication status	Published - Dec 9 2019
Externally published	Yes

ASJC Scopus subject areas

Electronic, Optical and Magnetic Materials
Condensed Matter Physics

Access to Document

10.1103/PhysRevB.100.214108

Cite this

@article{a0903a8595a040a087770edbf8391098,

title = "Thermodynamic integration by neural network potentials based on first-principles dynamic calculations",

abstract = "Simulation-size effect in evaluating the melting temperature of material is studied systematically by combining thermodynamic integration (TI) based on first-principles molecular-dynamics (FPMD) simulations and machine learning. Since the numerical integration to determine the free energies of two different phases as a function of temperature is very time consuming, the FPMD-based TI method has only been applied to small systems, i.e., less than 100 atoms. To accelerate the numerical integration, we here construct an interatomic potential based on the artificial neural-network (ANN) method, which retains the first-principles accuracy at a significantly lower computational cost. The free energies of the solid and liquid phases of rubidium are accurately obtained by the ANN potential, where its weight parameters are optimized to reproduce FPMD results. The ANN results reveal a significant size dependence up to 500 atoms.",

author = "Shogo Fukushima and Eisaku Ushijima and Hiroyuki Kumazoe and Akihide Koura and Fuyuki Shimojo and Kohei Shimamura and Masaaki Misawa and Kalia, {Rajiv K.} and Aiichiro Nakano and Priya Vashishta",

note = "Funding Information: This study was supported by JSPS KAKENHI Grant No. 16K05478 and JST CREST Grant No. JPMJCR18I2, Japan. R.K.K., A.N., and P.V. were supported by the US Department of Energy, Office of Science, Basic Energy Sciences, Materials Science and Engineering Division, Grant No. DE-SC0018195. The authors thank the Supercomputer Center, the Institute for Solid State Physics, University of Tokyo for the use of the facilities. The computations in this work were also performed using the facilities of the Research Institute for Information Technology, Kyushu University. Publisher Copyright: {\textcopyright} 2019 American Physical Society.",

year = "2019",

month = dec,

day = "9",

doi = "10.1103/PhysRevB.100.214108",

language = "English",

volume = "100",

journal = "Physical Review B",

issn = "2469-9950",

publisher = "American Physical Society",

number = "21",

}

TY - JOUR

T1 - Thermodynamic integration by neural network potentials based on first-principles dynamic calculations

AU - Fukushima, Shogo

AU - Ushijima, Eisaku

AU - Kumazoe, Hiroyuki

AU - Koura, Akihide

AU - Shimojo, Fuyuki

AU - Shimamura, Kohei

AU - Misawa, Masaaki

AU - Kalia, Rajiv K.

AU - Nakano, Aiichiro

AU - Vashishta, Priya

N1 - Funding Information: This study was supported by JSPS KAKENHI Grant No. 16K05478 and JST CREST Grant No. JPMJCR18I2, Japan. R.K.K., A.N., and P.V. were supported by the US Department of Energy, Office of Science, Basic Energy Sciences, Materials Science and Engineering Division, Grant No. DE-SC0018195. The authors thank the Supercomputer Center, the Institute for Solid State Physics, University of Tokyo for the use of the facilities. The computations in this work were also performed using the facilities of the Research Institute for Information Technology, Kyushu University. Publisher Copyright: © 2019 American Physical Society.

PY - 2019/12/9

Y1 - 2019/12/9

N2 - Simulation-size effect in evaluating the melting temperature of material is studied systematically by combining thermodynamic integration (TI) based on first-principles molecular-dynamics (FPMD) simulations and machine learning. Since the numerical integration to determine the free energies of two different phases as a function of temperature is very time consuming, the FPMD-based TI method has only been applied to small systems, i.e., less than 100 atoms. To accelerate the numerical integration, we here construct an interatomic potential based on the artificial neural-network (ANN) method, which retains the first-principles accuracy at a significantly lower computational cost. The free energies of the solid and liquid phases of rubidium are accurately obtained by the ANN potential, where its weight parameters are optimized to reproduce FPMD results. The ANN results reveal a significant size dependence up to 500 atoms.

AB - Simulation-size effect in evaluating the melting temperature of material is studied systematically by combining thermodynamic integration (TI) based on first-principles molecular-dynamics (FPMD) simulations and machine learning. Since the numerical integration to determine the free energies of two different phases as a function of temperature is very time consuming, the FPMD-based TI method has only been applied to small systems, i.e., less than 100 atoms. To accelerate the numerical integration, we here construct an interatomic potential based on the artificial neural-network (ANN) method, which retains the first-principles accuracy at a significantly lower computational cost. The free energies of the solid and liquid phases of rubidium are accurately obtained by the ANN potential, where its weight parameters are optimized to reproduce FPMD results. The ANN results reveal a significant size dependence up to 500 atoms.

UR - http://www.scopus.com/inward/record.url?scp=85076527389&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85076527389&partnerID=8YFLogxK

U2 - 10.1103/PhysRevB.100.214108

DO - 10.1103/PhysRevB.100.214108

M3 - Article

AN - SCOPUS:85076527389

SN - 2469-9950

VL - 100

JO - Physical Review B

JF - Physical Review B

IS - 21

M1 - 214108

ER -

Thermodynamic integration by neural network potentials based on first-principles dynamic calculations

Abstract

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this