Daniel Marchand^1*, Abhinav Jain^1*, Albert Glensk^1*, W. A. Curtin^1*

1 Institute of Mechanical Engineering, École Polytechnique Fédérale de Lausanne, CH-1015, Vaud, Switzerland

* Corresponding authors emails: daniel.marchand@epfl.ch, Abhinav.Jain@epfl.ch, Albert.Glensk@epfl.ch, william.curtin@epfl.ch

DOI10.24435/materialscloud:6k-je [version v3]

Publication date: Oct 14, 2020

How to cite this record

Daniel Marchand, Abhinav Jain, Albert Glensk, W. A. Curtin, Machine learning for metallurgy: a neural network potential for Al-Cu, Materials Cloud Archive 2020.119 (2020), doi: 10.24435/materialscloud:6k-je.

Description

High-strength metal alloys achieve their performance via careful control of precipitates and solutes. The nucleation, growth, and kinetics of precipitation, and the resulting mechanical properties, are inherently atomic-scale phenomena, particularly during early-stage nucleation and growth. Atomistic modeling using interatomic potentials is a desirable tool for understanding the detailed phenomena involved in precipitation and strengthening, which requires length and time scales far larger than those accessible by first-principles methods. Current interatomic potentials for alloys are not, however, sufficiently accurate for such studies. Here, a family of neural-network potentials (NNPs) for the Al-Cu system is presented as a first example of a machine-learning potential that can achieve near-first-principles accuracy for many different metallurgically-important aspects of this alloy. High fidelity predictions of intermetallic compounds, elastic constants, dilute solid-solution energetics, precipitate/matrix interfaces, generalized stacking fault energies, and surfaces for slip in matrix and precipitates, antisite defect energies, and other quantities, are shown. The NNPs also capture the subtle entropically-induced transition between θ' and θ at temperatures around 600K. Many comparisons are made with the state-of-the-art Angular-Dependent Potential for Al-Cu, demonstrating the significant quantitative benefit of a machine-learning approach. A preliminary kinetic Monte Carlo study shows the NNP to predict the emergence of GP zones in Al 4%Cu at T=300K in agreement with experiments. The NNP shows some significant transferability to defects and properties outside the structures used to develop the NNP but also shows some errors, highlighting that the use of any interatomic potential requires careful validation in application to specific metallurgical problems of interest.

Materials Cloud sections using this data

Files

File name	Size	Description
AllResultsSummary.tar.gz MD5md5:85caa1e8a850ef614aa756024b2344ad	2.1 MiB	Summary results in json format for all potentials tested in this work + DFT
AlCu-n2p2_inputdata.tar.gz MD5md5:c5d92750d28eebe0c6100b9bb708d9ad	4.6 MiB	All the training data used to generate the NNP: each structure has a uuid linking it to the AiiDA DB
n2p2_potentials.tar.gz MD5md5:dfa21b3648e64f7ef86ded95f2e4753a	2.0 MiB	All NNP potentials used in this study
README.txt MD5md5:33c605baa8ca6bb83ae934b961276128	4.5 KiB	Description of all Files and links to codes
aiida_repository.tar.gz MD5md5:11db54839d0f1dd0cc26662faa4ea1a0	40.7 GiB	Repository dump of AiiDA
dmarchand_migrated.psql.tar.gz MD5md5:29417bb86d6b0c861d395f22013e8911	93.1 MiB	Postgres dump of AiiDA database

License

Files and data are licensed under the terms of the following license: Creative Commons Attribution 4.0 International.
Metadata, except for email addresses, are licensed under the Creative Commons Attribution Share-Alike 4.0 International license.

External references

Keywords

Aluminum Machine Learning Metallurgy MARVEL/DD2