Computational Study of the Evolution of the Grain Boundary Network During Anisotropic Grain Growth

The grain boundary network (GBN) of polycrystalline materials changes during grain growth, affecting the material's properties. This research presents the results of fully anisotropic grain growth simulations. We perform simulations using three GB energy functions: an energy function that considers a GB's 5 degrees of freedom, the Read-Shockley model, and isotropic. We analyze the impact of these energy functions on the morphological evolution and various microstructural statistics, such as crystallographic texture and triple junction distribution. The results demonstrate that while individual grain evolution varies with GB energy function, certain microstructural statistics reach similar steady states across different models. In addition, we perform simulations using a diverse set of initial microstructures sampled from the texture hull to investigate their influence on the evolutionary trajectory during grain growth. We find a universal increase in the sharpness of orientation distribution functions (ODFs) and a positive correlation between texture strength and its sharpening rate. Additionally, the evolution of the GBN exhibits an increase in low-angle grain boundaries. Finally, we develop a method for predicting microstructural evolution as an alternative to expensive traditional grain growth simulations. Using the dataset from the previous simulations, we train a diffusion model to reconstruct the morphology of the evolved microstructure and a GBN spectrum predictive model to reconstruct the texture. The alternative method is almost ten times faster at producing an evolved microstructure than a traditional grain growth level set method. Overall, this work enhances our understanding of grain growth and the impact of the GB energy and initial texture on the evolution of the microstructure.