A DFT-driven multifidelity framework for constructing efficient energy models for atomic-scale simulations

Research output: Contribution to journalArticle

Standard

A DFT-driven multifidelity framework for constructing efficient energy models for atomic-scale simulations. / Messina, Luca; Quaglino, Alessio; Goryaeva, Alexandra; Marinica, Mihai-Cosmin; Domain, Christophe; Castin, Nicolas; Bonny, Giovanni; Krause, Rolf.

In: Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms, Vol. 483, 15.11.2020, p. 15-21.

Research output: Contribution to journalArticle

Harvard

Messina, L, Quaglino, A, Goryaeva, A, Marinica, M-C, Domain, C, Castin, N, Bonny, G & Krause, R 2020, 'A DFT-driven multifidelity framework for constructing efficient energy models for atomic-scale simulations', Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms, vol. 483, pp. 15-21. https://doi.org/10.1016/j.nimb.2020.09.011

APA

Messina, L., Quaglino, A., Goryaeva, A., Marinica, M-C., Domain, C., Castin, N., ... Krause, R. (2020). A DFT-driven multifidelity framework for constructing efficient energy models for atomic-scale simulations. Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms, 483, 15-21. https://doi.org/10.1016/j.nimb.2020.09.011

Vancouver

Messina L, Quaglino A, Goryaeva A, Marinica M-C, Domain C, Castin N et al. A DFT-driven multifidelity framework for constructing efficient energy models for atomic-scale simulations. Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms. 2020 Nov 15;483:15-21. https://doi.org/10.1016/j.nimb.2020.09.011

Author

Messina, Luca ; Quaglino, Alessio ; Goryaeva, Alexandra ; Marinica, Mihai-Cosmin ; Domain, Christophe ; Castin, Nicolas ; Bonny, Giovanni ; Krause, Rolf. / A DFT-driven multifidelity framework for constructing efficient energy models for atomic-scale simulations. In: Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms. 2020 ; Vol. 483. pp. 15-21.

Bibtex - Download

@article{b089fc308c1047dab2b2c0a06f13ae5c,
title = "A DFT-driven multifidelity framework for constructing efficient energy models for atomic-scale simulations",
abstract = "The reliability of atomistic simulations depends on the quality of the underlying energy models providing the source of physical information, for instance for the calculation of migration barriers in atomistic Kinetic Monte Carlo simulations. Accurate (high-fidelity) methods are often available, but since they are usually computationally expensive, they must be replaced by less accurate (low-fidelity) models that introduce some degrees of approximation. Machine-learning techniques such as artificial neural networks can be employed to work aroundthis limitation and extract the needed parameters from large databases of high-fidelity data. However, the latter are often computationally expensive to produce. This work introduces an alternative method based on the multifidelity approach. Correlations between high-fidelity and low-fidelity predictions are exploited to make an educated guess of the high-fidelity value based only on quick low-fidelity estimations, to be used for instance as an efficient and reliable source of physical data for atomistic simulations. With respect to neural networks, thisapproach requires less training data because of the lower amount of fitting parameters involved. The method is tested on the prediction of ab initio formation and migration energies of vacancy diffusion in iron-copper alloys,and compared with the neural networks trained on the same database.",
keywords = "Machine Learning, Kinetic Monte Carlo, Atomistic simulation",
author = "Luca Messina and Alessio Quaglino and Alexandra Goryaeva and Mihai-Cosmin Marinica and Christophe Domain and Nicolas Castin and Giovanni Bonny and Rolf Krause",
note = "Score=10",
year = "2020",
month = "11",
day = "15",
doi = "10.1016/j.nimb.2020.09.011",
language = "English",
volume = "483",
pages = "15--21",
journal = "Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms",
issn = "0168-583X",
publisher = "Elsevier",

}

RIS - Download

TY - JOUR

T1 - A DFT-driven multifidelity framework for constructing efficient energy models for atomic-scale simulations

AU - Messina, Luca

AU - Quaglino, Alessio

AU - Goryaeva, Alexandra

AU - Marinica, Mihai-Cosmin

AU - Domain, Christophe

AU - Castin, Nicolas

AU - Bonny, Giovanni

AU - Krause, Rolf

N1 - Score=10

PY - 2020/11/15

Y1 - 2020/11/15

N2 - The reliability of atomistic simulations depends on the quality of the underlying energy models providing the source of physical information, for instance for the calculation of migration barriers in atomistic Kinetic Monte Carlo simulations. Accurate (high-fidelity) methods are often available, but since they are usually computationally expensive, they must be replaced by less accurate (low-fidelity) models that introduce some degrees of approximation. Machine-learning techniques such as artificial neural networks can be employed to work aroundthis limitation and extract the needed parameters from large databases of high-fidelity data. However, the latter are often computationally expensive to produce. This work introduces an alternative method based on the multifidelity approach. Correlations between high-fidelity and low-fidelity predictions are exploited to make an educated guess of the high-fidelity value based only on quick low-fidelity estimations, to be used for instance as an efficient and reliable source of physical data for atomistic simulations. With respect to neural networks, thisapproach requires less training data because of the lower amount of fitting parameters involved. The method is tested on the prediction of ab initio formation and migration energies of vacancy diffusion in iron-copper alloys,and compared with the neural networks trained on the same database.

AB - The reliability of atomistic simulations depends on the quality of the underlying energy models providing the source of physical information, for instance for the calculation of migration barriers in atomistic Kinetic Monte Carlo simulations. Accurate (high-fidelity) methods are often available, but since they are usually computationally expensive, they must be replaced by less accurate (low-fidelity) models that introduce some degrees of approximation. Machine-learning techniques such as artificial neural networks can be employed to work aroundthis limitation and extract the needed parameters from large databases of high-fidelity data. However, the latter are often computationally expensive to produce. This work introduces an alternative method based on the multifidelity approach. Correlations between high-fidelity and low-fidelity predictions are exploited to make an educated guess of the high-fidelity value based only on quick low-fidelity estimations, to be used for instance as an efficient and reliable source of physical data for atomistic simulations. With respect to neural networks, thisapproach requires less training data because of the lower amount of fitting parameters involved. The method is tested on the prediction of ab initio formation and migration energies of vacancy diffusion in iron-copper alloys,and compared with the neural networks trained on the same database.

KW - Machine Learning

KW - Kinetic Monte Carlo

KW - Atomistic simulation

UR - https://ecm.sckcen.be/OTCS/llisapi.dll/overview/42750883

U2 - 10.1016/j.nimb.2020.09.011

DO - 10.1016/j.nimb.2020.09.011

M3 - Article

VL - 483

SP - 15

EP - 21

JO - Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms

JF - Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms

SN - 0168-583X

ER -

ID: 7000565