Deep-learning for microscopy image analysis of low-nuclearity catalysts

Low nuclearity catalysts represent a modern class of systems with unique properties, exceptionally positioned in rendering a large variety of conversion processes more sustainable and economically advantageous.  To optimize the design of this catalysts, and further improve their activity, selectivity, and stability, it is key to accurately infer structure-property relationship. In this context, scanning transmission electron microscopy (STEM) is instrumental for a statistically robust characterization of the structural properties of low-nuclearity catalysts. In my research I leverage computer vision algorithms to digitalize the analysis of STEM images, so to speed-up and scale-up their processing, improve process standardization, and access novel structural descriptors. To unlock the potential of the automated characterization of low-nuclearity catalysts my research nurtures a multi-disciplinary exchange with domain-experts in single and few atoms catalysis synthesis, microscopy, electronic structure modelling, and computer vision under the framework of the National Center of Competences in Research in Catalysis (NCCR Catalysis).





ETH Hönggerberg
HCI E129
Vladimir-Prelog-Weg 1
8093 Zurich
 tel: +41 4463 39217