Thèse de doctorat en Physique
Sous la direction de Mathis Plapp.
Soutenue en 2007
Phase-field modeling of polycrystalline solidification
The phase-field method has become in recent years the method of choice to model microstructural pattern formation during solidification. For monocrystals, quantitative agreement with experiments and analytical solutions has been obtained. The modeling of polycrystals, which consist of many grains of the same thermodynamic phase, but different orientations of the crystalline lattice, is far less advanced. Two types of models have been proposed: multi-phase-field models use a separate phase field for each grain, and orientation-field models use a small number of fields, but have non-analytical terms in their free energy functional. This work examines various aspects of phase-field modeling of polycrystals and is divided in three parts. In the first, a new possibility of describing the local orientation is explored, using a tensorial order parameter which represents automatically the local symmetry of the system. This approach is tested by developing a phase-field model for the nematic-isotropic phase transition in liquid crystals. The model is applied to simulate the directional ''solidification'' of a liquid crystal. The effect of the coupling between nematic orientation and the interface shape is investigated. The simulation results for the stability of a planar interface agree well with a generalized stability analysis, which takes into account a new anchoring condition at the interface: the nematic orientation at the interface is the result of the interplay between bulk deformation and interface anisotropy. The shape and stability of well-developed cells is also influenced by this effect. Numerically, the use of a tensorial order parameter simplifies the treatment of the symmetries in the system significantly, while the equations of motions become considerably more complicated. In the second part, grain boundaries are investigated on a smaller length scale, using a phase field crystal model, where elastic properties and dislocations appear naturally. With this model, the local order in interfaces is examined and the stability of liquid films between two solid grains is studied below the melting point. This situation can be described by an interaction potential between the two solid-liquid interfaces, which is extracted numerically. The results are compared with a phenomenological model which is found to hold for high-angle grain boundaries, where the dislocations overlap. For low-angle grain boundaries, premelting around dislocation as well as a symmetry breaking (dislocations form pairs) is observed. As a result, the interaction potential becomes nonmonotonous, and consists of a long-range attraction and a short-range repulsion. In the third part, a new phase-field model is developed using an angle variable to describe the crystalline orientation. Contrary to the already existing models, the free energy is constructed without a term proportional to the modulus of the gradient of the orientation field. Instead, the standard squared gradient is used, but it is coupled to the phase field with a singular coupling function. Various benchmark simulations are carried out to test the model. It is found that it presents several artifacts such as spurious grain rotation and interface motion; however, these effects are extremely small, such that the model yields satisfactory results unless the undercooling is very small. Finally, the observed problems are analyzed and ways of obtaining a better description of the dynamics of the angle field are discussed.