Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • harvard1
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Introducing ab initio-based neural networks for transition-rate prediction in kinetic Monte Carlo simulations
KTH, School of Engineering Sciences (SCI), Physics, Reactor Physics.ORCID iD: 0000-0003-0562-9070
KTH, School of Engineering Sciences (SCI), Physics, Reactor Physics.ORCID iD: 0000-0002-2381-3309
(English)Manuscript (preprint) (Other academic)
Abstract [en]

This work presents an innovative approach to kinetic Monte Carlo (KMC) simulations, in which atomic transition rates are predicted by an artificial neural network trained on ab initio migration barriers. The method is applied to the parameterization of a hybrid atomistic-object KMC model to simulate copper precipitation during thermal aging in iron. The stability and mobility of copper clusters containing one vacancy is analyzed by means of independent atomistic KMC simulations driven by the same neural network, with the aim of parameterizing the object KMC part of the model. Copper clusters are found to be more stable and mobile with respect to previous studies, and can cover longer diffusion paths, reaching up to a few lattice units. The mean free path increases with cluster size up to around 100 copper atoms. In addition, the emission of the vacancy often occurs concurrently with the emission of one or more copper atoms, because of strong vacancy-copper correlations and kinetic coupling. In the hybrid KMC simulations, the density of copper clusters is overestimated because of the excessively high solution energy predicted by the ab initio method. Nevertheless, this work proves the capability of neural networks to transfer detailed ab initio thermodynamic and kinetic properties to the KMC model, and sets the ground for reliable microstructure evolution simulations in a wide range of alloys.

Keyword [en]
thermal aging, computer simulations, ab initio, kinetic Monte Carlo, neural networks
National Category
Condensed Matter Physics
Research subject
Physics
Identifiers
URN: urn:nbn:se:kth:diva-177523OAI: oai:DiVA.org:kth-177523DiVA: diva2:873144
Funder
Göran Gustafsson Foundation for promotion of scientific research at Uppala University and Royal Institute of TechnologyVattenfall ABEU, FP7, Seventh Framework Programme
Note

QS 2015

Available from: 2015-11-23 Created: 2015-11-23 Last updated: 2015-11-23Bibliographically approved
In thesis
1. Multiscale modeling of atomic transport phenomena in ferritic steels
Open this publication in new window or tab >>Multiscale modeling of atomic transport phenomena in ferritic steels
2015 (English)Doctoral thesis, comprehensive summary (Other academic)
Abstract [en]

Defect-driven transport of impurities plays a key role in the microstructure evolution of alloys, and has a great impact on the mechanical properties at the macroscopic scale. This phenomenon is greatly enhanced in irradiated materials because of the large amount of radiation-induced crystal defects (vacancies and interstitials). For instance, the formation of nanosized solute clusters in neutron-irradiated reactor pressure vessel (RPV) ferritic steels has been shown to hinder dislocation motion and induce hardening and embrittlement. In Swedish RPV steels, this mechanical-property degradation is enhanced by the high content of manganese and nickel impurities. It has been suggested that the formation of Mn-Ni-rich clusters (which contain also Cu, Si, and P) might be the outcome of a dynamic process, where crystal defects act both as nucleation sites and solute carriers. Solute transport by point defects is therefore a crucial mechanism to understand the origin and the dynamics of the clustering process.

The first part of this work aims at modeling solute transport by point defects in dilute iron alloys, to identify the intrinsic diffusion mechanisms for a wide range of impurities. Transport and diffusion coefficients are obtained by combining accurate ab initio calculations of defect transition rates with an exact mean-field model. The results show that solute drag by single vacancies is a common phenomenon occurring at RPV temperature (about 300 °C) for all impurities found in the solute clusters, and that transport of phosphorus and manganese atoms is dominated by interstitial-type defects. These transport tendencies confirm that point defects can indeed carry impurities towards nucleated solute clusters. Moreover, the obtained flux-coupling tendencies can also explain the observed radiation-induced solute enrichment on grain boundaries and dislocations.

In the second part of this work, the acquired knowledge about solute-transport mechanisms is transferred to kinetic Monte Carlo (KMC) models, with the aim of simulating the RPV microstructure evolution. Firstly, the needed parameters in terms of solute-defect cluster stability and mobility are calculated by means of dedicated KMC simulations. Secondly, an innovative approach to the prediction of transition rates in complex multicomponent alloys is introduced. This approach relies on a neural network based on ab initio-computed migration barriers. Finally, the evolution of the Swedish RPV steels is simulated in a "gray-alloy" fashion, where impurities are introduced indirectly as a modification of the defect-cluster mobilities. The latter simulations are compared to the experimental characterization of the Swedish RPV surveillance samples, and confirm the possibility that solute clusters might form on small interstitial clusters.

In conclusion, this work identifies from a solid theoretical perspective the atomic-transport phenomena underlying the formation of embrittling nanofeatures in RPV steels. In addition, it prepares the ground for the development of predictive KMC tools that can simulate the microstructure evolution of a wide variety of irradiated alloys. This is of great interest not only for reactor pressure vessels, but also for many other materials in extreme environments.

Place, publisher, year, edition, pages
Stockholm: KTH Royal Institute of Technology, 2015. xvii, 90 p.
Series
TRITA-FYS, ISSN 0280-316X ; 2015:80
Keyword
diffusion, impurities, iron, metals, kinetic Monte Carlo, ab initio, mean field, defects, embrittlement, reactor pressure vessel, neural networks
National Category
Condensed Matter Physics
Research subject
Physics
Identifiers
urn:nbn:se:kth:diva-177525 (URN)978-91-7595-764-7 (ISBN)
Public defence
2015-12-11, Svedberg Hall, Room FD5, Albanova Universtitetscentrum, Roslagstullsbacken 21, Stockholm, 13:00 (English)
Opponent
Supervisors
Funder
Göran Gustafsson Foundation for promotion of scientific research at Uppala University and Royal Institute of TechnologyVattenfall ABEU, FP7, Seventh Framework Programme
Note

QC 20151123

Available from: 2015-11-23 Created: 2015-11-23 Last updated: 2015-11-24Bibliographically approved

Open Access in DiVA

No full text

Authority records BETA

Messina, LucaOlsson, Pär

Search in DiVA

By author/editor
Messina, LucaOlsson, Pär
By organisation
Reactor Physics
Condensed Matter Physics

Search outside of DiVA

GoogleGoogle Scholar

urn-nbn

Altmetric score

urn-nbn
Total: 268 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • harvard1
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf