Improved image reconstruction of electrocatalytic surfaces via data-driven optimization

Research question

This project aims to leverage data-driven methods to visualize electrochemical changes in materials at the nanometer scale, supporting the design of electrocatalysts for sustainable technologies such as water splitting and CO₂ utilization. Improving electrocatalytic activity and selectivity in oxygen reactions (OER/ORR) requires modifying the catalyst surface—specifically reducing the native oxide layer and forming a new, controlled oxide layer.

Although in operando Transmission Electron Microscopy (TEM) can visualize these processes, frame-by-frame analysis is time-consuming, and signal loss in liquid electrolytes limits resolution. To address this, we propose enhancing compressed sensing algorithms using prior imaging data and model knowledge. A deep neural network will be trained to identify characteristic features of electrocatalytic surfaces, guiding the reconstruction of high-quality TEM visualizations in real time.

Sustainability aspects

Technology based on direct conversion between chemical and electrical energy is crucial for a sustainable society and to reduce global warming. Electrochemical water splitting and recombination or CO2 utilization is key for electrolysers and fuel cells. However, current catalysts for oxygen reactions rely on scarce, expensive platinum group metals (e.g., Ir, Ru), limiting scalability.  We have
developed thin film electrocatalysts based on transition metals as a sustainable alternative. Modifying the native oxide layer is essential to expose active sites and enhance performance. Additionally, the process allows for post-use catalyst recovery. This project will develop a reconstruction model for TEM micrographs to visualize the formation and evolution of surface oxides and oxide particles, supporting the development of cost-effective, high-performance catalysts.