Comparison of microstructure characterization methods by two-point correlation functions and reconstruction of 3D microstructures using 2D TEM images with high degree of phase clustering

Highlights: • Large phase clusters are characterized by cooccurrence correlation functions (CCFs). • Coarseness parameter effectively confirms the consistency of CCFs between 2D and 3D. • The performance of CCFs are compared with that of two-point correlation function. • 3D microstructures with large phase clusters are constructed from CCFs of 2D images. Microstructural characterization methods and their performance in virtual microstructure reconstruction were investigated using two-phase Pebax/PBE membranes. In this study, the microstructural characteristics of two-phase Pebax/PBE membranes with random phase distribution were obtained from two-dimensional (2D) transmission electron microscopy (TEM) images to reconstruct three-dimensional (3D) virtual microstructures. The microstructural characteristics of the 2D TEM images with different volume fractions were expressed using a two-point correlation function (P₂) and cooccurrence correlation functions (CCFs), and the results were compared. Virtual 3D microstructures were reconstructed through a stochastic optimization process using CCFs obtained from 2D TEM images of the membranes. This study further investigated and evaluated the effects of phase clustering on probabilistic functions, namely, P₂ and CCFs. It is confirmed that CCFs accurately represent the characteristics of microstructures with a high degree of phase clustering and that virtual 3D samples can be effectively reconstructed using CCFs obtained from 2D images.