Beyond the Eyepiece: 'Learning' Physics from Microscopy Data

Recent advances in microscopy techniques have revolutionized our ability to observe materials at an atomic level, revealing intricate details down to point defects. However, this progress comes with the challenge of managing and interpreting the enormous volumes of data generated. Traditional methods of data acquisition and analysis are no longer sufficient for the fast-paced discovery of advanced materials. This dissertation tackles this issue by harnessing advanced machine learning to efficiently process and analyze microscopic data, offering new insights into material properties and behavior. The research objectives include developing methods to decode complex material processes, linking experimental data to predictive models, and creating an innovative approach to optimize material properties while reducing the need for extensive experimentation. This research combines diverse microscopy datasets across various material systems, illustrating the synergy between data science, cutting-edge microscopy, and innovative modeling. Our findings offer a comprehensive view of material behaviors, highlighting the potential of integrating data science in material research, with broader implications for materials science and engineering.