Evolving 3D microstructures using a Genetic Algorithm

At the last international meeting on recrystallisation and grain growth there was general consensus that there was a need to incorporate experimental microstructures as starting configurations for computer models of various kinds. In two dimensions (2D) this is relatively straight forward since microstructures can be imaged in 2D at various length scales by a number of well know techniques such as scanning electron microscopy (SEM), transmission electron microscopy (TEM) and atomic force microscopy (AFM). But in three dimensions (3D) it is very problematic because getting real 3D information is experimentally very challenging. In this paper we will describe an general approach to solving this problem of simulation start structures. We will describe a program called MicroConstructor, that will take any two dimensional (2D) scanning electron microscope image (SEM) micrograph and generate a three dimensional (3D) discrete computer microstructure which is statistically equivalent in terms of the microstructral variables of interest. The basis of the code is a genetic algorithm that evolves the 3D microstructure so that its cross-sections match the 2D data. Since this approach is not limited by scale in can be used to generate 3D initial multiscale microstructures. This algorithm will enable multiscale microstructural modellers to use as their starting point, real experimental microstructures without having to acquire 3D information experimentally, a very time consuming and expensive process. We will also comment on the application of this type of approach to design nanomaterials.