Multiscale modelling improved an epoxidation catalyst
More than 90 per cent of chemicals are produced in catalytic processes where the reaction is only feasible if a suitable catalyst is used. The search for better catalysts with higher selectivity and activity has dominated the chemical industry since the late 19th century, as each improvement brings significant financial, environmental and societal benefits. Using computational methods, we screened the entire periodic table to find a better catalyst for the epoxidation of ethylene. While silver is known to be the best performing catalyst, our model predicted that its alloys with Cu/Pb, Cu/Cd or Cu/Tl should outperform it. They were then synthesised experimentally and their superiority was confirmed.
The epoxidation of ethylene is industrially and commercially one of the most important selective oxidations. Silver catalysts have been state of the art for decades, with their efficiency steadily improving through empirical discoveries of dopants and co-catalysts. However, we have computationally scoured the entire periodic table in the hope of finding an improved catalyst. We combine ab initio calculations, scaling relationships and rigorous microkinetic reactor modelling that goes beyond conventional simplified steady-state or rate-determining modelling on invariant catalyst surfaces. We performed detailed quantum chemical simulations on nine noble metals and then used these data to infer the activity of the entire periodic table based on descriptors. It was predicted that Ag/CuPb, Ag/CuCd and Ag/CuTl would outperform the pure Ag catalysts.
To confirm the theoretical prediction, they were synthesised in-house to accurately compare the effects of each dopant and alloy. We ensured that they had an easily scalable synthesis protocol. Characterisation using advanced methods such as high-resolution imaging, X-ray diffraction, XPS, TPD and TEM confirmed their composition and elucidated their structure. Further catalytic testing showed that the model correctly predicted the performance of all catalysts tested, including superior and less active samples.
We have shown that to fully exploit the potential of computational discovery of catalysts, it is essential to include relevant in situ conditions, e.g. surface oxidation, parasitic side reactions and ethylene epoxide decomposition, as neglecting such effects leads to erroneous predictions. The insights from modelling have enabled us to both synthesise novel catalysts and theoretically understand experimental results, bridging the gap between first-principles simulations and industrial applications. We show that computational catalyst design can be easily extended to account for larger reaction networks and other effects, such as surface oxidation.
Authors: Matej Huš, Miha Grilc, Janvit Teržan and Blaž Likozar from the Department of Catalysis and Chemical Reaction Engineering at the National Institute of Chemistry, Sašo Gyergyek from Jožef Stefan Institute and Anders Hellman from Chalmers University of Technology in Gothenburg, Sweden. It was published in Angewandte Chemie.
Link: https://onlinelibrary.wiley.com/doi/10.1002/anie.202305804
Contact person: matej.hus(at)ki.si