Machine-learned acceleration for molecular dynamics in CASTEP

Tamás Stenczel; Zakariya El-Machachi; Guoda Liepuoniute; Joe Morrow; Albert Bartók; Matt Probert; Gábor Csányi; Volker Deringer

doi:10.1063/5.0155621

Machine-learned acceleration for molecular dynamics in CASTEP

Tamás Stenczel, Zakariya El-Machachi, Guoda Liepuoniute, Joe Morrow, Albert Bartók, Matt Probert, Gábor Csányi^*, Volker Deringer^*

^*Corresponding author for this work

Physics

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

Abstract

Machine learning (ML) methods are of rapidly growing interest for materials modeling, and yet, the use of ML interatomic potentials for new systems is often more demanding than that of established density-functional theory (DFT) packages. Here, we describe computational methodology to combine the CASTEP first-principles simulation software with the on-the-fly fitting and evaluation of ML interatomic potential models. Our approach is based on regular checking against DFT reference data, which provides a direct measure of the accuracy of the evolving ML model. We discuss the general framework and the specific solutions implemented, and we present an example application to high-temperature molecular-dynamics simulations of carbon nanostructures. The code is freely available for academic research.

Original language	English
Article number	044803
Number of pages	12
Journal	Journal of Chemical Physics
Volume	159
DOIs	https://doi.org/10.1063/5.0155621
Publication status	Published - 27 Jul 2023

Bibliographical note

Access to Document

10.1063/5.0155621Licence: CC BY

Machine-learned acceleration for molecular dynamics in CASTEP
© Author(s) 2023
Final published version, 7.56 MBLicence: CC BY

CASTEP-USER: Predictive materials modelling for experimental scientists
Hasnip, P. J., Byrne, P. J. P. & Probert, M.
EPSRC
1/06/22 → 31/05/25
Project: Research project (funded) › Research

Cite this

@article{9e7fc2855ca345428f36e61fc47bb56a,

title = "Machine-learned acceleration for molecular dynamics in CASTEP",

abstract = "Machine learning (ML) methods are of rapidly growing interest for materials modeling, and yet, the use of ML interatomic potentials for new systems is often more demanding than that of established density-functional theory (DFT) packages. Here, we describe computational methodology to combine the CASTEP first-principles simulation software with the on-the-fly fitting and evaluation of ML interatomic potential models. Our approach is based on regular checking against DFT reference data, which provides a direct measure of the accuracy of the evolving ML model. We discuss the general framework and the specific solutions implemented, and we present an example application to high-temperature molecular-dynamics simulations of carbon nanostructures. The code is freely available for academic research.",

author = "Tam{\'a}s Stenczel and Zakariya El-Machachi and Guoda Liepuoniute and Joe Morrow and Albert Bart{\'o}k and Matt Probert and G{\'a}bor Cs{\'a}nyi and Volker Deringer",

note = "{\textcopyright} Author(s) 2023",

year = "2023",

month = jul,

day = "27",

doi = "10.1063/5.0155621",

language = "English",

volume = "159",

journal = "Journal of Chemical Physics",

issn = "0021-9606",

publisher = "American Institute of Physics",

}

TY - JOUR

T1 - Machine-learned acceleration for molecular dynamics in CASTEP

AU - Stenczel, Tamás

AU - El-Machachi, Zakariya

AU - Liepuoniute, Guoda

AU - Morrow, Joe

AU - Bartók, Albert

AU - Probert, Matt

AU - Csányi, Gábor

AU - Deringer, Volker

PY - 2023/7/27

Y1 - 2023/7/27

N2 - Machine learning (ML) methods are of rapidly growing interest for materials modeling, and yet, the use of ML interatomic potentials for new systems is often more demanding than that of established density-functional theory (DFT) packages. Here, we describe computational methodology to combine the CASTEP first-principles simulation software with the on-the-fly fitting and evaluation of ML interatomic potential models. Our approach is based on regular checking against DFT reference data, which provides a direct measure of the accuracy of the evolving ML model. We discuss the general framework and the specific solutions implemented, and we present an example application to high-temperature molecular-dynamics simulations of carbon nanostructures. The code is freely available for academic research.

AB - Machine learning (ML) methods are of rapidly growing interest for materials modeling, and yet, the use of ML interatomic potentials for new systems is often more demanding than that of established density-functional theory (DFT) packages. Here, we describe computational methodology to combine the CASTEP first-principles simulation software with the on-the-fly fitting and evaluation of ML interatomic potential models. Our approach is based on regular checking against DFT reference data, which provides a direct measure of the accuracy of the evolving ML model. We discuss the general framework and the specific solutions implemented, and we present an example application to high-temperature molecular-dynamics simulations of carbon nanostructures. The code is freely available for academic research.

U2 - 10.1063/5.0155621

DO - 10.1063/5.0155621

M3 - Article

SN - 0021-9606

VL - 159

JO - Journal of Chemical Physics

JF - Journal of Chemical Physics

M1 - 044803

ER -

Machine-learned acceleration for molecular dynamics in CASTEP

Abstract

Bibliographical note

Access to Document

Projects

CASTEP-USER: Predictive materials modelling for experimental scientists

Cite this