A methodology for automated quantitative microstructural analysis of transmission electron micrographs

Abstract

It is generally recognized that proper quantification of microstructural behavior is necessary for the optimization of materials properties. In the specific case of polycrystalline thin films, transmission electron microscopy (TEM) is required for microstructural interrogation due to the small (nm–μm) inherent length scales in these systems. While a meaningful study requires large grain populations, typical data sets are relatively small due to the need for human interpretation of the contrast in TEM micrographs. To overcome this limitation, a general methodology has been developed to fully automate the grain boundary identification procedure by using a combination of both conventional and newly developed image processing algorithms to extract and combine information from multiple, optimized TEM micrographs. This technique has been validated by systematically analyzing microstructures of thin Al films, as obtained from TEM micrographs, and comparing these results with those obtained by a conventional, manual approach. Indeed, a statistical analysis shows that the agreement between these two methods is quite good. We further show that based upon a large population (8185 grains) one can estimate the number of grains required to draw meaningful conclusions about this microstructure. While the emphasis of this work is on TEM image processing, the techniques developed here are expected to be sufficiently general and flexible so as to be applicable to other (e.g., focused-ion beam) microscopies.