Amin Behnami
This thesis develops a computational workflow for estimating local composition at GaAs/AlGaAs semiconductor interfaces from HAADF-STEM images. The method combines Anscombe-based denoising, Fourier-domain evaluation, model-based atomic-column fitting, and relative comparison with known well and barrier regions. It also applies a dataset-specific crosstalk correction to refine the estimated aluminum-composition profile while clearly reporting the limitations and uncertainty of the analysis.