2025 Theses Doctoral

Thin Film Grain Growth Studies in the Transmission Electron Microscope: Imaging, Segmentation, and Orientation Mapping

Patrick, Matthew Joseph

The microstructure of polycrystalline materials has well documented impacts on their properties, but process development for controlling grain growth remains empirical. In short, predictive models are limited by the multi-dimensional nature of this ensemble problem and the consequent scarcity of time series data. This dissertation presents developments in transmission electron microscopy (TEM)-based approaches for acquiring and analyzing such data, leveraging thin films’ columnar microstructure to study the behavior of microstructures and grain boundaries under geometric constraints and to study the dynamics of grain growth through in situ heating experiments.

While in many instances thin films act as a proxy for bulk materials, geometric constraints are found to introduce unexpected behavior with implications for coarsening. In particular, in two experiments it is shown that the dihedral angles at grain boundary triple junctions do not obey the established Herring equilibrium equations relating triple junction geometry and grain boundary energies, pointing to equilibrium effects related to strain, surface energies, and non-equilibrium effects like triple junction drag, which result in measurable deviations in the morphology of the grain boundary network. Furthermore, the large surface energy contributions that lead to the development of [111]-fiber textures in FCC materials impose geometric restrictions on grain boundary character, leading to the favored growth of high relative energy grain boundaries at the expense of lower energy boundaries, in contrast to bulk materials.

To the end of achieving complete dynamic characterization of grain growth, the longstanding grain/grain boundary identification problem is addressed for brightfield (BF)-TEM images of polycrystalline films with the introduction of two convolutional neural network (CNN)-based segmentation approaches. These models, benchmarked by physical observables, enable the rapid, high-throughput analysis of the thousands of images acquired during an in situ heating experiment, which would not be possible via previous manual methodologies. Demonstrating the use-case, a special-case BF-TEM imaging mode is employed to capture an evolving microstructure during an in situ heating experiment at high time resolution; the images are analyzed automatically to characterize grain size evolution and identify grains and grain boundaries. These microstructural features are spatially correlated to orientation maps acquired before and after heating, demonstrating a framework for tagging dynamically acquired image data with intermittently collected crystallographic data. Given adequate object tracking, this suggests TEM-based thin film grain growth experiments are a viable platform for a complete and high-time resolution characterization of grain growth.

In summary, this dissertation (i) expands our understanding of the effects of thin film geometry on microstructural development, especially with respect to grain boundary character, energy and triple junction behavior, and (ii) develops the software infrastructure and experimental frameworks required for high-throughput TEM-based thin film grain growth studies, establishing an experimental platform for the future development of data-driven models for microstructural evolution.