Static strain aging is an important concept in metalurgy that refers to the hardening of a material that has undergone plastic deformation and then is aged for a certain period of time. A theory proposed in the late 1940s by Cottrell and Bilby explains this phenomenon as being caused by the pinning of dislocations by impurities (e.g., carbon atoms in solid solution) that migrate to the vicinity of the line defect. In the course of this PhD work, the atomistic mechanism behind the static strain aging phenomenon in bcc iron has been studied by means of computer simulations. Given the fact that diffusion in the solid state proceeds slowly, thus preventing the use of molecular dynamics at low temperatures (when the effect of the dislocation stress field on carbon diffusion is more pronounced), we have preferentially employed a method coupling molecular statics with atomistic kinetic Monte Carlo. Three major points have been addressed by this thesis: (i) the effect of the stress field of an edge or screw dislocation on a carbon atom diffusing nearby; (ii) the diffusion of a carbon atom in the tight channel found in the dislocation core (pipe diffusion); and (iii) the equilibrium carbon distribution in a Cottrell atmosphere. The main effect of the dislocation stress field outside the dislocation core consists of biasing carbon diffusion, such that some transitions become more likely than others. This effect is expected to drive the early stages of Cottrell atmosphere formation, when the mutual interaction between carbon atoms is negligible. Right in the dislocation core, as expected, carbon was seen to diffuse faster than in the bulk. Carbon concentration in the neighborhood of an edge or a screw dislocation was modeled by an approach based in statistical physics using the binding energies calculated by molecular statics, revealing a good agreement with experimental data obtained by atom probe techniques.

Ag thin film growth on ZnO substrates has been investigated theoretically using multi-timescale simulation methods. The models are based on an atomistic approach where the interactions between atoms are treated classically using a mixture of fixed and variable charge potential energy functions. After some preliminary tests it was found that existing fixed charge potential functions were unreliable for surface growth simulations. This resulted in the development of a ReaxFF variable charge potential fitted to Ag/ZnO surface interactions. Ab initio models of simple crystal structures and surface configurations were used for potential fitting and testing. The dynamic interaction of the Ag atoms with the ZnO surface was first investigated using single point depositions, via molecular dynamics, whereby the Ag impacted various points on an irreducible symmetry zone of the ZnO surface at a range of energies. This enabled the determination of the relative numbers of atoms that could penetrate, reflect or bond to the surface as a function of incident energy. The results showed that at an energy of up to 10 eV, most atoms deposited adsorbed on top of the surface layer. The second part of the dynamic interaction involved a multi-timescale technique whereby molecular dynamics (MD) was used in the initial stages followed by an adaptive kinetic Monte Carlo (AKMC) approach to model the diffusion over the surface between impacts. An impact energy of 3 eV was chosen for this investigation. Ag was grown on various ZnO surfaces including perfect polar, O-deficient and surfaces with step edges. Initial growth suggests that Ag prefers to be spread out across a perfect surface until large clusters are forced to form. After further first layer growth, subsequent Ag atoms begin to deposit on the existing Ag clusters and are unlikely to join the first layer. Ag island formation (as mentioned within the literature) can then occur via this growth mechanism. O-deficient regions of ZnO surfaces result in unfavourable Ag adsorption sites and cause cluster formation to occur away from O-vacancies. In contrast, ZnO step edges attract deposited Ag atoms and result in the migration of surface Ag atoms to under-coordinated O atoms in the step edge. Various improvements have been made to the existing methodology in which transitions are determined. A new method for determining defects within a system, by considering the coordination number of atoms, is shown to increase the number of transitions found during single ended search methods such as the relaxation and translation (RAT) algorithm. A super-basin approach based on the mean rate method is also introduced as a method of accelerating a simulation when small energy barriers dominate. This method effectively combines states connected by small energy barriers into a single large basin and calculates the mean time to escape such basin. To accelerate growth simulations further and allow larger systems to be considered, a lattice based adaptive kinetic Monte Carlo (LatAKMC) method is developed. As off-lattice AKMC and MD results suggest Ag resides in highly symmetric adsorption sites and that low energy deposition events lead to no penetrating Ag atoms or surface deformation, the on-lattice based approach is used to grow Ag on larger perfect polar ZnO surfaces. Results from the LatAKMC approach agree with off-lattice AKMC findings and predict Ag island formation. Critical island sizes of Ag on ZnO are also approximated using a mean rate approach. Single Ag atoms are placed above an existing Ag cluster and all transition states are treated as belonging to a single large super-basin . Results indicate that small Ag clusters on the perfect ZnO surface grow in the surface plane until a critical island size of around 500 atoms is reached. Once a critical island size is reached, multiple Ag ad-atoms will deposit on the island before existing Ag atoms join the cluster layer and hence islands will grow upwards. A marked difference is seen for second layer critical island sizes; second layer Ag islands are predicted to be two orders of magnitude smaller (< 7 atoms). This analysis suggests that Ag on ZnO (0001) may exhibit Stranski-Krastanov (layer plus island) growth.

Metal thin films deposited on weakly-interacting substrates constitute an essential element of numerous microelectronic, catalytic, and optical devices. However, the natural tendency of metal atoms to agglomerate, upon condensation on a weakly-interacting surface, in dispersed three-dimensional (3D) islands affects negatively the performance of the above-mentioned devices. The aim of this thesis is to investigate one of the mechanisms governing silver (Ag) 3D island growth on weakly-interacting substrates, i.e. the nucleation of a new layer on the island top. Kinetic Monte Carlo (KMC) simulations are employed to calculate the top island-layer critical radius Rc required for nucleating a new layer in the out-of-plane direction. Single-island simulations are performed for growth temperatures T in the range 250 to 500 K and ratios of the pairwise adatom/substrate atom bond strength EB,sub to the corresponding adatom/adatom value EB,film in the range 0.5 to 0.75. We find that for T values below 250 K the islands exhibit a 2D morphology for all EB,sub/EB,film ratios. In contrast, for T values above 300 K there exists a range of relatively small EB,sub/EB,film values, where 2D morphology dominates. To calculate Rc for each island layer as the island shape evolves, a subroutine is developed and implemented in an existing KMC algorithm. Rc values are computed for 3D island growth at EB,sub/EB,film = 0.5 in the T range 300−500 K and the results show that Rc decreases monotonously from 17.3 to 6.0 Å and saturates approximately at 375 K. This trend is opposite to the typical behavior of islands grown under homoepitaxial conditions, for which the enhancement of downward inter-layer diffusion caused by an increase of T leads to lower atomic densities on the top, i.e. to a lower nucleation probability, and thus to an increase of Rc. This work contributes to the understanding of the physical processes that control thin-film morphological evolution; which is paramount for controlling and manipulating film growth for specific applications.

abstract: In material science, microstructure plays a key role in determining properties, which further determine utility of the material. However, effectively measuring microstructure evolution in real time remains an challenge. To date, a wide range of advanced experimental techniques have been developed and applied to characterize material microstructure and structural evolution on different length and time scales. Most of these methods can only resolve 2D structural features within a narrow range of length scale and for a single or a series of snapshots. The currently available 3D microstructure characterization techniques are usually destructive and require slicing and polishing the samples each time a picture is taken. Simulation methods, on the other hand, are cheap, sample-free and versatile without the special necessity of taking care of the physical limitations, such as extreme temperature or pressure, which are prominent issues for experimental methods. Yet the majority of simulation methods are limited to specific circumstances, for example, first principle computation can only handle several thousands of atoms, molecular dynamics can only efficiently simulate a few seconds of evolution of a system with several millions particles, and finite element method can only be used in continuous medium, etc. Such limitations make these individual methods far from satisfaction to simulate macroscopic processes that a material sample undergoes up to experimental level accuracy. Therefore, it is highly desirable to develop a framework that integrate different simulation schemes from various scales to model complicated microstructure evolution and corresponding properties. Guided by such an objective, we have made our efforts towards incorporating a collection of simulation methods, including finite element method (FEM), cellular automata (CA), kinetic Monte Carlo (kMC), stochastic reconstruction method, Discrete Element Method (DEM), etc, to generate an integrated computational material engineering platform (ICMEP), which could enable us to effectively model microstructure evolution and use the simulated microstructure to do subsequent performance analysis. In this thesis, we will introduce some cases of building coupled modeling schemes and present the preliminary results in solid-state sintering. For example, we use coupled DEM and kinetic Monte Carlo method to simulate solid state sintering, and use coupled FEM and cellular automata method to model microstrucutre evolution during selective laser sintering of titanium alloy. Current results indicate that joining models from different length and time scales is fruitful in terms of understanding and describing microstructure evolution of a macroscopic physical process from various perspectives. / Dissertation/Thesis / Doctoral Dissertation Materials Science and Engineering 2018

Dans le cadre de la gestion des déchets nucléaires, les coques et embouts des gaines de combustibles sont coupées, compactées et empilés dans les colis CSD-C (Colis Standards de Déchets Compactés). Actuellement entreposés à la Hague, ces déchets seront stockés en milieu géologique profond dans des alvéoles en béton. Ces déchets MA-VL contiennent des RN d’intérêts dont le carbone-14 qui provient de l’activation neutronique de l’azote-14 et de l’oxygène-17 présents dans le Zircaloy. L’objectif de cette thèse, qui est menée en collaboration avec EDF et AREVA, est de répondre à la question du devenir et du comportement du carbone-14 dans ZrO₂ et Zr-métal. Cette thèse consistera à utiliser la simulation numérique, au moyen de la Théorie de la Fonctionnelle de la Densité (DFT), afin demodéliser le comportement du carbone-14 dans les CSD-C. Les simulations sont effectuées avec une approche multi-échelles : 1) Échelle atomique, optimisation de modèles atomiques permettant de représenter les différentes phases de ZrO₂ ; identification des différents sites d’insertion possibles pour l’atome de carbone en sites interstitiels et en substitutions à un atome d’oxygène ou de zirconium ; modélisation des différentes chemins d’énergie minimum associées à la migration d’un atome de carbone d’un site à un autre par mécanismes interstitiels et lacunaires. 2) Échelle macroscopique, détermination des coefficients de diffusion par la méthode Monte-Carlo dans ZrO₂ monoclinique pure et Zr pure. / As part of the nuclear waste management, hulls and ends of fuel claddings are cut, compacted and put in CSD-C containers (compacted standard waste containers). Currently stored at La Hague, the waste will be stored in deep geological environment. The MA-VL waste contains RN including carbon-14, which comes from the neutron activation of nitrogen-14 and oxygen-17 present in the oxide Zircaloy. The objective of this thesis, which is taken in collaboration with EDF and AREVA, is to answer the question of the carbon-14 behavior in ZrO₂ and Zr metal. This thesis is to perform numerical simulations, using the Theory of Density Functional (DFT) to model the behavior of carbon-14 in the CSD-C. The simulations are performed with multi-scale approach: 1) At atomic scale, optimization of atomic models to represent the different phases of ZrO₂; identification of the insertion sites for the carbon atom in interstitial sites and substitutions of an oxygen or zirconium atom; modeling different minimum energy path for the migration of a carbon atom from one site to another by interstitial and vacancye mechanisms. 2) At macroscopic scale, determination of diffusion coefficients in pure bulk monoclinic ZrO₂ and pure bulk Zr by the Monte Carlo method.

This thesis investigates dierent types of Monte Carlo estimators for use in computationof chemical system, mainly to be used in calculating surface growthand evolution of SiC. Monte Carlo methods are a class of algorithms using randomsampling to numerical solve problems and are used in many cases. Threedierent types of Monte Carlo methods are studied, a simple Monte Carlo estimatorand two types of Markov chain Monte Carlo Metropolis algorithm MonteCarlo and kinetic Monte Carlo. The mathematical background is given for allmethods and they are tested both on smaller system, with known results tocheck their mathematical and chemical soundness and on larger surface systemas an example on how they could be used

The formation of Ge nanowires in V-grooves has been studied experimentally as well as theoretically. As substrate oxide covered Si V-grooves were used formed by anisotropic etching of (001)Si wafers and subsequent oxidation of their surface. Implantation of 1E17 Ge+ cm^-2 at 70 keV was carried out into the oxide layer covering the V-grooves. Ion irradiation induces shape changes of the V-grooves, which are captured in a novel continuum model of surface evolution. It describes theoretically the effects of sputtering, redeposition of sputtered atoms, and swelling. Thereby, the time evolution of the target surface is determined by a nonlinear integro-differential equation, which was solved numerically for the V-groove geometry. A very good agreement is achieved for the predicted surface shape and the shape observed in XTEM images. Surprisingly, the model predicts material (Si, O, Ge) transport into the V-groove bottom which also suggests an Ge accumulation there proven by STEM-EDX investigations. In this Ge rich bottom region, subsequent annealing in N2 atmosphere results in the formation of a nanowire by coalescence of Ge precipitates shown by XTEM images. The process of phase separation during the nanowire growth was studied by means of kinetic 3D lattice Monte-Carlo simulations. These simulations also indicate the disintegration of continuous wires into droplets mediated by thermal fluctuations. Energy considerations have identified a fragmentation threshold and a lower boundary for the droplet radii which were confirmed by the Monte Carlo simulation. The here given results indicate the possibility of achieving nanowires being several nanometers wide by further growth optimizations as well as chains of equally spaced clusters with nearly uniform diameter.

Multiple studies on the microstructures of advanced high-strength steels are presented here that seek to add to the already substantial body of knowledge on martensite in steel. These studies seek to gain additional insight into the role that the martensite transformation has on the observed mechanical properties of modern steels. Crystallographic Reconstruction of Parent Austenite Twin Boundaries in a Lath Martensitic Steel The study of post-transformation microstructures and their properties can be greatly enhanced by studying their dependence on the grain boundary content of parent microstructures. Recent work has extended the crystallographic reconstruction of parent austenite in steels to include the reconstruction of special boundaries, such as annealing twins. These reconstructions present unique challenges, as twinned austenite grains share a subset of possible daughter variant orientations. This gives rise to regions of ambiguity in a reconstruction. A technique for the reconstruction of twin boundaries is presented here that is capable of reconstructing 60 degree twins, even in the case where twin regions are comprised entirely of variants that are common between the twin and the parent. This technique is demonstrated in the reconstruction of lath martensitic steels. The reconstruction method utilizes a delayed decision-making approach, where a chosen orientation relationship is used to define all possible groupings of daughter grains into possible parents before divisive decisions are made. These overlapping, inclusive groupings (called clusters) are compared to each other individually using their calculated parent austenite orientations and the topographical nature of the overlapping region. These comparisons are used to uncover possible locations of twin boundaries present in the parent austenite. This technique can be applied to future studies on the dependence of post-transformation microstructures on the special grain boundary content of parent microstructures. Coupling Kinetic Monte Carlo and Implicit Finite Element Methods for Predicting the Strain Path Sensitivity of the Mechanically Induced Martensite Transformation The kinetic Monte Carlo method is coupled with a finite-element solver to simulate the nucleation of martensite inside the retained austenite regions of a TRIP (transformation induced plasticity) assisted steel. Nucleation kinetics are expressed as a function of load path and kinematic coupling between retained austenite regions. The model for martensite nucleation incorporates known elements of the kinetics and crystallography of martensite. The dependence of martensite transformation on load path is simulated and compared to published experimental results. The differences in transformation rates of retained austenite are shown to depend on load path through the Magee effect. The effects of average nearest neighbor distance between austenite grains is shown to affect the rate at which martensite nucleates differently depending on load path. Ductility and Strain Localization of Advanced High-Strength Steel in the Presence of a Sheared Edge The localization of strain in the microstructures of DP 980 and TBF 980 is quantified and compared. Of particular interest is the difference in final elongation observed for both materials in the presence of a sheared edge. Scanning electron micrographs of etched microstructures near the sheared edge are gathered for both materials at varying amounts of macroscopic strain. These micrographs are used to generate strain maps using digital image correlation. A two point statistical measure for strain localization is developed that utilizes strain map data to quantify the degree to which strain localizes around the hard phase of both materials. The DP steel exhibits higher strain localization around the martensite phase. Reasons for differences in strain localization and shear banding between the two materials are suggested, and the role played by the mechanically induced martensite transformation is speculated.

No description available.

No description available.