Potentiel interatomique | EPFL Graph Search

Le potentiel interatomique est un modèle d'énergie potentielle servant à décrire l'interaction entre atomes et, par extension, entre molécules. Il permet d'accéder à nombre de quantités optiques, thermodynamiques, mécaniques et de transport de la matière. Atomes Les atomes à température ambiante peuvent s'associer spontanément pour former des molécules stables. Cela implique une force \mathbf{F} attractive à longue distance et répulsive à faible distance, et par suite nulle à la valeur d'équilibre de la liaison. Cette force de liaison est conservative, et on peut donc lui associer une énergie potentielle E, qu'on appelle souvent « potentiel » par abus de langage, relié à la force par \mathbf{F}=-\nabla E. L'énergie d'un système baisse lorsqu'il est plus stable. Par convention, on fixe l'énergie égale à zéro lorsque les atomes sont très éloignés (« à l'infini »). Le potentiel est souvent exprimé sous forme réduite par la constante d

An Atomistic Simulation Method for Oxygen Impurities in Aluminium Based on Variable Charge Molecular Dynamics

Andreas Elsener

Impurities are known to have a significant impact on materials properties. In particular, the presence of impurities can change mechanical properties and stabilize the microstructure by reducing grain growth and recrystallization processes. In the past atomistic simulations, in particular molecular dynamics, have contributed a lot to the understanding of deformation mechanisms in nanocrystalline metals. Especially, insights into details of dislocation/GB interactions have been gained. Additionally, simulations on coupled grain boundary migration in bicrystalline samples have played an important role in understanding stress assisted grain growth in nanocrystalline metals. However, all these simulations have been performed on monatomic systems. This means that the role of impurities has not yet been considered in these simulations. Some of the important difference between experiments and simulations could be removed by introducing dilute O distributions to computational Al samples. For this purpose, a simulation method is required which can deal with the oxidation of Al atoms around the O impurities and which can simulate the ionic and metallic nature of bonding simultaneously. Here, a method is suggested which fulfills the requirements by modifying the variable charge method of Streitz and Mintmire [Phys. Rev. B 50, 11996 (1994)]. A local chemical potential approach which optimizes the charge on only those atoms expected to be ionic is developed. Additionally the EAM potential for the non-electrostatic interaction given in the publication of Streitz and Mintmire is substituted by empirical potentials which treat the Al-Al interactions with a well-known Al potential. The Al-O and O-O interactions in these EAM potentials are fitted for the simulation of dilute O impurities in Al and additionally for the simulation of corundum in case of a second potential. The aforementioned developments are applied to study the effect of O impurities on the deformation of nc Al. A fully three-dimensional nc sample containing 15 grains of 12 nm grain-size is deformed with a dilute O content in the triple junction lines. The impact of O impurities on dislocation propagation is also considered in a sample which consists of a single grain extracted from a large molecular dynamics simulation. This simulation geometry contains a perfect dislocation, which is pinned at the surrounding grain boundary network. Therefore it allows the study of pinning strength and after unpinning the propagation behavior of the dislocation under various O impurity distributions. Additionally, the role of O impurities on coupled GB migration is investigated by studying two bicrystalline samples with different O distributions in both of them. In the discussion, the emphasis is on the simulation method. Especially, the significance of the samples, simulation conditions and the performance of the method in the different applications are discussed. Additionally, the different developments (local chemical potential approach and EAM potentials) are critically analyzed and improvements for future simulations of impurities are suggested.

EPFL2010

Neural Network Potentials for Age Hardening Aluminum Alloys

Daniel John Gilles Marchand

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.In this work we will consider new interatomic potentials for the 2xxx (Al-Cu-Mg), 6xxx (Al-Mg-Si), and 7xxx (Al-Zn-Mg-Cu) age-hardenable aluminum alloys. First a family of neural-network potentials (NNPs) for the Al-Cu system are presented as a simple example of a machine-learning potential that can achieve near-first-principles accuracy for many different metallurgically-important aspects of 2xxx alloys. 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 captures the subtle entropically-induced transition between

\theta'

and

\theta

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-4at%Cu at T=300K in agreement with experiments.Then, a family of NNPs for the Al-Mg-Si system is developed to enable quantitative studies of 6xxx alloys. The NNP is trained on metallurgically-important quantities computed by first principles density function theory (DFT) leading to high fidelity predictions of intermetallic compounds, elastic constants, dilute solid-solution energetics, precipitate/matrix interfaces, Al stacking fault energies, antisite defect energies, and other quantities. The Generalized Stacking Fault Energy surfaces (GSFE) for the three prevalent

\beta "

precipitate compositions in peak-aged Al-6xxx are then computed with the NNP, and are validated by DFT computations at key points. A preliminary examination of early-stage clustering kinetics and energetics in Al-6xxx is then made, showing the formation of low-energy Mg-Si structures and the trapping of vacancies in these clusters.Finally we demonstrate a family of NNPs to produce what we believe to be the first Al-Cu-Mg (2xxx) and Al-Cu-Mg-Zn (7xxx) potentials of sufficient accuracy to be used for atomistic simulations. These NNP studies show that the NNP has significant transferability to defects and properties outside the structures used to train 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.

EPFL2022

Atomistic modeling of the solid-liquid interface of metals and alloys

Edoardo Baldi

Solidification is a phase transformation of utmost importance in material science, for it largely controls materials' microstructure on which a wide range of mechanical properties depends. Almost every human artifact undergoes a transformation that leads to a solid phase, be it via well-established manufacturing processes as casting or forging, or more recent technological advances such as 3D printing. This thesis aims to study some fundamental aspects of solidification, focussing in particular on metals and alloys. Despite being a phenomenon investigated for a long time with both experiments and theoretical models, solidification still involves a very complex set of phenomena that requires a multi-scale approach to provide useful insights. For example, properties relating to the thermodynamics and kinetics of solid-liquid interfaces play a crucial role in micro-scale modeling of solidification, yet are particularly challenging to assess with experimental techniques. Our main instrument of investigation has been a set of well-established computer simulation techniques at the atomic scale, and we applied and extended some of these methods, with the primary goal of improving the reliability of some results related to the properties of solid-liquid interfaces and being able to study systems whose level of complexity comes closer to that of interest to some real manufacturing processes. The approach of atomistic simulations involves several technical and theoretical problems. A first issue is related to the size of the systems that it is necessary to simulate to obtain results of some relevance despite an inherent statistical error that is often substantial. Moreover, some theoretical subtleties, such as the arbitrariness in the definition of the interface, inevitably emerge from an approach at the atomic scale, and they lead to problems whose solution is anything but univocal; often, in fact, different formulations of these problems provide different results. The first contribution of this thesis focused on extending a computational method that serves to calculate a fundamental quantity known as interface free energy and, in particular, to decrease its computational cost by reducing the number of particles of the simulated systems. The second contribution addressed the study of crystal-melt interface properties of a particular metallic binary alloy. The idea behind this part of the work was to combine different techniques of atomistic simulations whose outcomes make it possible to obtain an exhaustive description of both the thermodynamics and the dynamics of the interfaces. We have based this approach on a precise definition of dividing surface, and we have derived all our results in a consistent way allowing us to eliminate some arbitrary choices that similar kinds of simulations usually entail. A part of the results obtained confirmed the reliability of our approach, showing satisfactory agreement with some other established results. However, challenges remain associated with the accuracy of the interatomic potential, the presence of significant finite-size effects, and the difficulty in converging to satisfactory statistical accuracy the thermodynamic and dynamical properties of solid-liquid interfaces.

EPFL2020