Start of funding 01.01.2026

End-to-end Machine Learning Methods for Real-Time Characterization of New Energy Materials

Prof. Dr. Philipp Pelz
Friedrich-Alexander-University of Erlangen-Nuremberg
Materials Science

Prof. Dr. Colin Ophus
Stanford University
Materials Science



The transition to sustainable energy technologies demands a precise understanding of materials at the atomic scale, where structure and chemistry govern performance in batteries, solar cells, and catalysts. Yet modern energy materials are increasingly complex, pushing conventional, manually tuned electron microscopy workflows beyond their limits. This project addresses this challenge by developing autonomous, machine-learning-driven analysis pipelines for four-dimensional scanning transmission electron microscopy (4D-STEM). The central goal is an end-to-end neural framework that directly extracts crystal orientation, crystal structure, and chemical information from raw diffraction data in real time. By replacing hand-engineered algorithms with fully trainable models, the approach enables high-throughput, reproducible, and quantitative characterization. The collaboration unites complementary expertise from FAU Erlangen-Nürnberg and Stanford University. Stanford contributes state-of-the-art microscopes, novel battery materials, and deep experience in ML-based diffraction analysis. FAU provides advanced deep-learning methods, direct electron detection, and a focus on thin-film solar cell materials. Joint research visits will generate benchmark datasets, train prototype models, and prepare high-impact publications. Together, the partners will establish a scalable framework for accelerated discovery of next-generation energy materials.