A fuzzy logic approach to image analysis of HREM micrographs of III-V compounds

High resolution electron microscopy can within certain limits provide quantitative information on morphology and composition of crystalline materials. In layered structures of III-V semiconductor compounds changes resulting from interdiffusion phenomena across interfaces may result in image contrast changes because of composition sensitive electron scattering. A new approach for extracting the corresponding information from cross-section images of such interfaces is introduced. Local variations of similarity, a measure of comparability with respect to previously defined templates of known material composition on both sides of the interface, are recorded by image analysis. On the basis of the well-worked out fuzzy set theory variations of the underlying chemical composition are evaluated by applying fuzzy logic criteria of inference to image regions of about 1 nm x 1 nm in size. With comparatively little computational effort this new approach can be used to localize regions of significant composition changes, i.e. edge detection, and to determine the composition across the interface region, i.e. chemical mapping, respectively. The method is introduced and its reliability demonstrated by means of simulated contrast patterns of GaAs/GaAs1-xPx as well as of experimental images of AlAs/AlxGa1-xAs.