In scanning transmission electron microscopy (STEM), we can collect a full two-dimensional diffraction pattern at every position as we scan the electron beam across a sample. This means that for each point in real space, we also have a detailed map of how electrons are scattered in reciprocal space, resulting in a so-called “4D-STEM” dataset (two dimensions in real space, two in reciprocal space). Each of these diffraction patterns—known as convergent beam electron diffraction (CBED) patterns—contains valuable information about the sample, such as its thickness and orientation, as well as about the properties of the incident electron beam. However, because these patterns are complex and information-rich, it is challenging to extract specific details from them using traditional analysis methods.

In this project, we aim to use convolutional neural networks (CNNs) to automatically extract important information like defocus, sample thickness, and orientation from CBED patterns. To do this, you will first generate simulated CBED patterns for a wide range of sample types, orientations, and defocus values using advanced simulation tools (multislice simulations). Next, you will use these simulated patterns to train a CNN and test how well it can predict the relevant experimental parameters. Finally, you will apply the trained CNN to real experimental data and integrate it into data analysis workflows, helping to improve how we collect and interpret STEM data. All work will be done using open-source Python packages such as abTEM, py4DSTEM and HyperSpy.

Through this project, you will gain hands-on experience with machine learning for image analysis, advanced electron microscopy techniques, electron scattering simulations, and atomic-scale materials characterization. A solid foundation in Python programming and a basic understanding of optics and materials science are required. We will provide guidance and teach you all advanced topics needed for the project. The project is suitable for both Bachelor’s and Master’s students; for Bachelor’s projects, the scope can be adjusted to focus on specific parts of the project.

If you are interested in the project, please contact Dr. Zafran Shah (zafran.shah@rub.de) and Dr. Christoph Flathmann (christoph.flathmann@rub.de).