Mesoscopic Modeling and Experimental Investigations of the Last-Stage Solidification of Globular-Equiaxed Grains in Metallic Alloys

Hot tearing is a major defect of solidification processes and welding. It occurs in metallic alloys while they are in the semi-solid state and is caused by a lack of liquid feeding in the mushy zone to compensate solid network openings due to tensile and shear strains. These strains are localized in the thin liquid films that remain at grain boundaries. Therefore, it is important to know when the solid grains coalesce and percolate to form a fully coherent solid, i.e., when the material can transmit stresses and strains but also withstand thermal contraction without rupture. Coalescence is defined as the disappearance of a liquid film in between two distinct grains, forming a solid grain boundary, while percolation is defined as the gradual transition of isolated grains or clusters surrounded by a continuous liquid film to a continuous solid network across a domain, even if some liquid regions remain. If the simulation of hot tearing has to account for localization of strains and liquid feeding, it is mandatory to use a granular approach. Two previous theses done recently at EPFL have developed such granular models of hot tearing, first in 2D (PhD of S. Vernède) and then in 3D (PhD of M. Sistaninia). However, these models have used a Voronoi tessellation for the description of globular microstructures in which equiaxed grains were approximated by polygons and polyhedra, respectively. From the experimental point of view, an innovative method, based on the Bridgman furnace principle, has been developed. The standard Bridgman furnace has been substantially modified in order to obtain the desired globular-equiaxed grain structure and reduce macrosegregation in the quenched sample. This modified furnace has been characterized then in terms of thermal measurements and observed final grain size. The samples obtained with the modified Bridgman furnace were then observed post-mortem by ex situ X-ray tomography. In addition, X-ray tomography experiments under similar conditions have been performed in situ with the laser-heated furnace installed on the TOMCAT synchrotron beam line at PSI. A postprocessing analysis of the principal curvature distribution (ISD plots) was performed on both ex situ and in situ observations for various compositions of inoculated Al-Cu alloys. The ISD plots of the X-ray tomography observations indicated the formation of liquid cylinders along triple line and, at a more advanced solidification stage, the formation of liquid pocket at quadruple vertices. From the modeling point of view, a new mesoscopic model has been developed both for 2D and 3D simulations. This model was inspired by the granular model developed by Phillon et al. that considers polyhedral grains based on a Voronoi tessellation of space, but allow to obtain smoother grain shapes and more progressive coalescence. After its validation with multiphase-field simulations, the mesoscopic model has been used to predict the various percolation transitions of the solid phase. In addition, by estimating the solid moment at which the liquid starts to form isolated pockets, it was possible to predict the temperature and solid fractions for which the mushy zone is most vulnerable to hot tearing. The simulations have been performed for various conditions and copper compositions, indicating the nominal compositions that are more sensitive to hot tearing. Finally, the 3D mesoscopic model predictions were compared with the X-ray tomography observations.