In this thesis, a quantitative analysis framework is constructed to establish the relationship between electron microscopy images to material structure and constitution information. Firstly, the TEM/STEM simulation software based on multislice method with NVIDIA CUDA acceleration is created. The contrast mechanism of HAADF STEM image is fully investigated under the frozen phonon model owing to almost one hundred times acceleration of efficiency brought by graphics processing unit. Then the simulated images based on my software are used to investigate the influence of scanning distortion caused by machine instability and environmental disturbance. A practical scanning distortion correction algorithm accounting for both high and low frequency scanning distortion is proposed. The corrected...[ Read more ]

In this thesis, a quantitative analysis framework is constructed to establish the relationship between electron microscopy images to material structure and constitution information. Firstly, the TEM/STEM simulation software based on multislice method with NVIDIA CUDA acceleration is created. The contrast mechanism of HAADF STEM image is fully investigated under the frozen phonon model owing to almost one hundred times acceleration of efficiency brought by graphics processing unit. Then the simulated images based on my software are used to investigate the influence of scanning distortion caused by machine instability and environmental disturbance. A practical scanning distortion correction algorithm accounting for both high and low frequency scanning distortion is proposed. The corrected images based on simulated and experimental HAADF STEM image fully illustrate the reliability of my scanning distortion correction algorithm. At the same time, the scale space theory is introduced to extract features in HAADF STEM image, and validity of this method is also proved via direct simulation and experimental observation. In addition, the exit wave reconstruction methods are fully reviewed, and a new method called coherent space theory based on calculated exit electron wave is proposed to make three dimensional reconstruction of single layered material. Then, combining the feature extraction methods like scale space theory with exit wave reconstruction methods, systematic theory on precise and accurate deformation analysis for atom resolved images is proposed. These quantitative analysis techniques with GPU acceleration are finally applied to process in-situ videos and tomography data, which possess much larger data amount compared to traditional single image analysis.