Clean data in oxidation catalysis

Sven Richter, Pierre Kube, Spencer Carey, Peter Kraus

Transparent and open handling of data is becoming a topic of crucial importance across scientific fields. In computational science, examples of open data repositories such as NOMAD and QCArc hive are already available. With the launch of the National Research Data Infrastructure (NFDI) in Germany, the Department of Inorganic Chemistry takes part in the FAIRmat consortium.1 Catalysis in general, an d oxidation catalysis in particular, is a complex and highly dimensional problem. While a large amount of high-quality data has been produced over the last 40 years,2 the resulting datasets can usually not be compared directly, as the applied conditions are inconsistent and frequently not complete. Here we propose a “Clean Data Handbook”, which aims to define a minimum standard for catalytic testing and characterisation of materials in selective oxidation catalysis as a test case, enabling comparison of materials under reference conditions.
An initial set of 10 well-known catalysts that exhibit diverse performance has been selected. The catalysts are synthesized in large batch sizes of 10-20 g each and studied in CO, propane (Giulia Bellini, Rania Hanna, Pierre Kube), ethane and butane oxidation using a prescribed schedule of steps in collaboration with the BasCat Laboratory at the Technical University Berlin (Stephen Lohr, Michael Geske, Raoul Naumann d’Alnoncourt, Frank Rosowski). Fresh and spent catalysts are characterized by standard techniques (chemical analysis (Olaf Timpe), XRD (Frank Girgsdies), N2 adsorption (Maike Hashagen), thermal analysis and temperature-programmed oxidation-reduction (Andrey Tarasov), XPS (Spencer Carey), electron microscopy (Liudmyla Masliuk, Christian Rohner, Maxime Boniface, Thomas Lunkenbein) UV/Vis (Gregory Huff) and infrared spectroscopy (Jutta Kröhnert)). Heats of adsorption of the reactants are determined by oxygen and alkane adsorption calorimetry (Sabine Wrabetz, Andrey Tarasov).
The material science of heterogeneous catalysis is characterized by dynamic interactions between catalyst and reaction medium that are modulated by transport phenomena at various length scales. Consequently, catalysts need to be analysed under operation. In the current iteration, contactless conductivity measurements (Peter Kraus), Raman spectroscopy (Yuanqing Wang, Gregory Huff), and near-ambient pressure (NAP) XPS (Spencer Carey, Jinhu Dong, Toyin Omojola, Yuanqing Wang, Gregory Huff, Michael Hävecker, Detre Teschner, Axel Knop-Gericke) are performed in operando.
The work is performed in collaboration with the Theory Department (Luca Ghiringhelli, Matthias Scheffler) with the aim to use the toolbox of artificial intelligence for establishing a bridge between theory and experiment by searching for hidden relations and causally determined connections in experimental data with the help of machine learning and data science. Theory could then predict the combination of optimal conditions and materials for a given process. This is, however, only possible if the generation of the experimental input adequately supports the requirements of machine learning and re-usability of data.



2U. Zavyalova, M. Holena, R. Schlögl, M. Baerns, ChemCatChem 2011, 3, 1935

Go to Editor View