Loading...
Please wait, while we are loading the content...
Développement de techniques numériques pour l'estimation, la modélisation et la prédiction de propriétés thermodynamiques et structurales de systèmes métalliques à fort ordonnancement chimique
| Content Provider | Semantic Scholar |
|---|---|
| Author | Harvey, J. |
| Copyright Year | 2012 |
| Abstract | Dans ce travail, la possibilite de calculer et d’evaluer avec precision l’energie de Gibbs d’equilibres de phases complexes, pour lesquels le comportement energetique des solutions solides metalliques est decrit par des modeles thermodynamiques decrivant simultanement les ordonnancements chimiques a courte et longue distance ainsi que des solutions metalliques liquides decrites par un modele decrivant l’ordonnancement a courte distance, a ete etudiee. Pour ce faire, les modeles thermodynamiques appeles « cluster site approximation » (CSA) et « cluster variation method » (CVM) ont ete implementes a l’interieur d’une toute nouvelle technique de minimisation de l’energie libre de Gibbs de systemes multicomposants et multiphasiques pour decrire le comportement thermodynamique de solutions solides metalliques presentant un fort ordonnancement chimique. Le modele quasichimique modifie avec approximation de paires (MQMPA) a egalement ete implemente dans le nouvel algorithme de minimisation pour decrire le comportement thermodynamique des solutions liquides. Ce nouvel algorithme de minimisation sous contraintes a ete construit selon une approche de programmation quadratique sequentielle de type Newton exacte (c’est-a-dire l’usage de derivees secondes exactes pour evaluer la matrice Hessienne de la fonction objective) avec recherche lineaire de direction de descente suffisante. L’implementation de cette nouvelle technique de minimisation d’energie de Gibbs sous contraintes a ete necessaire suite a la rencontre de difficultes quant a l’identification de l’etat d’equilibre pour des conditions donnees lors de l’utilisation du logiciel FactSage lorsque les modeles mentionnes ci-haut sont simultanement presents pour decrire le comportement thermodynamique des phases presentes a l’equilibre (ex. : Solide_CSA+Liquide_MQMPA, Solide1_CSA+Solide2_CSA). Une fois le developpement de la technique de minimisation de l’energie de Gibbs valide pour plusieurs systemes binaires et ternaires trouves dans la litterature, il a ete possible de parametrer ces modeles de solutions solides pour des systemes d’interet industriel tel les systemes Cu-Zr et Al-Zr a partir de donnees thermodynamiques coherentes generees a partir d’un simulateur Monte Carlo, completement bâti dans le cadre de ce travail, utilisant le potentiel interatomique appele « modified embedded atom model in the second nearest neighbour pair formalism » (MEAM-2NN) pour decrire l’energie interne des structures etudiees. Ce potentiel interatomique a du etre parametre pour les interactions mixtes du systeme Al-Zr dans le cadre de ce travail. Le chemin d’integration thermodynamique a aussi ete implemente dans ce simulateur de facon a pouvoir evaluer l’energie de Gibbs absolue des structures considerees tant solides que liquides, ce qui permet de deduire l’entropie. L’implementation de cette technique a alors permis pour la premiere fois de calculer de maniere theorique toutes les contributions de melange (contributions enthalpiques et entropiques) d’une phase binaire liquide (Cu-Zr et Al-Zr) et solide (solution CFC Al-Zr) a partir du potentiel MEAM-2NN. Les proprietes thermodynamiques et structurales obtenues suite a de nombreuses simulations Monte Carlo couplees a des simulations de dynamique moleculaire classique ont ensuite ete utilisees pour parametrer les modeles thermodynamiques mentionnes precedemment pour la phase liquide du systeme Cu-Zr ainsi que la phase liquide et la solution solide cubique a faces centrees pour le systeme Al-Zr. L’etude de ces systemes a egalement permis d’introduire de nouveaux parametres d’exces fonction de la configuration de la solution solide decrite par les modeles CVM ou CSA pour ameliorer la precision de ces modeles base sur des evidences experimentales trouvees dans la litterature. Un meilleur parametrage du modele quasichimique modifie pour la phase metallique liquide a egalement ete propose en tenant compte des fractions de paires calculees par simulations Monte Carlo et des differents effets de relaxation volumique independants de la configuration modelises a partir de parametres reguliers. Suite a la generation de donnees thermodynamiques coherentes et au parametrage subsequent des phases solides et liquide du systeme Al-Zr, un diagramme de phases respectant completement ces donnees coherentes dans l’intervalle de composition a ete presente. Des simulations Monte Carlo ont egalement ete faites pour comprendre et definir le comportement thermodynamique de la phase amorphe du systeme Al-Zr. Finalement, les modeles thermodynamiques ainsi parametres pour les solutions solides, liquides et amorphes ont pu servir a definir les conditions, selon des criteres thermodynamiques et volumetriques permettant de definir une gamme de composition, ou les verres metalliques massifs peuvent etre produits. ---------- In this work, the possibility to calculate and evaluate with a high degree of precision the Gibbs energy of complex multiphase equilibria for which chemical ordering is explicitly and simultaneously considered in the thermodynamic description of solid (short range order and long range order) and liquid (short range order) metallic phases is studied. The cluster site approximation (CSA) and the cluster variation method (CVM) are implemented in a new minimization technique of the Gibbs energy of multicomponent and multiphase systems to describe the thermodynamic behaviour of metallic solid solutions showing strong chemical ordering. The modified quasichemical model in the pair approximation (MQMPA) is also implemented in the new minimization algorithm presented in this work to describe the thermodynamic behaviour of metallic liquid solutions. The constrained minimization technique implemented in this work consists of a sequential quadratic programming technique based on an exact Newton’s method (i.e. the use of exact second derivatives in the determination of the Hessian of the objective function) combined to a line search method to identify a direction of sufficient decrease of the merit function. The implementation of a new algorithm to perform the constrained minimization of the Gibbs energy is justified by the difficulty to identify, in specific cases, the correct multiphase assemblage of a system where the thermodynamic behaviour of the equilibrium phases is described by one of the previously quoted models using the FactSage software (ex.: solid_CSA+liquid_MQMPA; solid1_CSA+solid2_CSA). After a rigorous validation of the constrained Gibbs energy minimization algorithm using several assessed binary and ternary systems found in the literature, the CVM and the CSA models used to describe the energetic behaviour of metallic solid solutions present in systems with key industrial applications such as the Cu-Zr and the Al-Zr systems are parameterized using fully consistent thermodynamic an structural data generated from a Monte Carlo (MC) simulator also implemented in the framework of this project. In this MC simulator, the modified embedded atom model in the second nearest neighbour formalism (MEAM-2NN) is used to describe the cohesive energy of each studied structure. A new Al-Zr MEAM-2NN interatomic potential needed to evaluate the cohesive energy of the condensed phases of this system is presented in this work. The thermodynamic integration (TI) method implemented in the MC simulator allows the evaluation of the absolute Gibbs energy of the considered solid or liquid structures. The original implementation of the TI method allowed us to evaluate theoretically for the first time all the thermodynamic mixing contributions (i.e., mixing enthalpy and mixing entropy contributions) of a metallic liquid (Cu-Zr and Al-Zr) and of a solid solution (face-centered cubic (FCC) Al-Zr solid solution) described by the MEAM-2NN. Thermodynamic and structural data obtained from MC and molecular dynamic simulations are then used to parameterize the CVM for the Al-Zr FCC solid solution and the MQMPA for the Al-Zr and the Cu-Zr liquid phase respectively. The extended thermodynamic study of these systems allow the introduction of a new type of configuration-dependent excess parameters in the definition of the thermodynamic function of solid solutions described by the CVM or the CSA. These parameters greatly improve the precision of these thermodynamic models based on experimental evidences found in the literature. A new parameterization approach of the MQMPA model of metallic liquid solutions is presented throughout this work. In this new approach, calculated pair fractions obtained from MC/MD simulations are taken into account as well as configuration-independent volumetric relaxation effects (regular like excess parameters) in order to parameterize precisely the Gibbs energy function of metallic melts. The generation of a complete set of fully consistent thermodynamic, physical and structural data for solid, liquid, and stoichiometric compounds and the subsequent parameterization of their respective thermodynamic model lead to the first description of the complete Al-Zr phase diagram in the range of composition based on theoretical and fully consistent thermodynamic properties. MC and MD simulations are performed for the Al-Zr system to define for the first time the precise thermodynamic behaviour of the amorphous phase for its entire range of composition. Finally, all the thermodynamic models for the liquid phase, the FCC solid solution and the amorphous phase are used to define conditions based on thermodynamic and volumetric considerations that favor the amorphization of Al-Zr alloys. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://publications.polymtl.ca/871/1/2012_JeanPhilippeHarvey.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
