A Screening Workflow for Synthesis and Testing of 10,000 Heterogeneous Catalysts per Day – Lessons Learned Daniel G. Duff, Arne Ohrenberg, Stephan Voelkening,* Matthias Boll Bayer Technology Services GmbH, Process Technology, Building E41, 51368 Leverkusen, Germany Fax: þ49 (0) 214-30 50698; E-mail: stephan.voelkening@bayertechnology.com Received: October 6, 2003; Revised: November 12, 2003; Accepted: November 12, 2003; DOI: 10.1002/marc.200300171 Keywords: catalysis; data mining; design of experiment; high-throughput screening; inorganic materials 1. Introduction Heterogeneous catalysis has been a subject of industrial and academic research for many decades and is a key to developing innovative products and energy-efficient, resource saving processes. New catalysts for new and existing processes are still being sought and discovered. Whereas yield and selectivity are the main character- istics to be optimised, a series of other parameters have also to be considered, such as for example the compo- sition of the by-product mixture, which determines the nature of the work-up, as well as catalyst lifetime, price and patent freedom. The search for a new catalyst represents a multi- parameter optimisation problem in an extremely large and often discrete parameter space. This makes the search for a novel catalyst very difficult. A universal strategy does not exist. Summary: For some years now a trend toward the employ- ment of high-throughput technologies in the development of new catalysts has been apparent in academic and industrial research. One reason is the unmanageably large parameter space present in heterogeneous catalysis, which implies that an increase in throughput in the preparation and testing of candidate substances is necessary, in order to be able to identify a new catalyst in a practicable amount of time. We have tried to locate the current boundaries in achievable throughput. To this end, we developed a workflow in which 10,000 substances per day were synthesised and their activity tested in a heterogeneously catalysed gas phase reaction (alkene epoxidation). The substances were synthesised by dosing precursor solutions onto a single substrate, using ink-jet printer technology and subsequent thermal treatment. For the activity testing, the product stream of each candi- date was conducted through a detection layer, where the target product was converted into a fluorescent substance. By locally resolved fluorescence spectroscopy a rough assess- ment of each candidate was possible. Because a throughput of 10,000 substances per day is still not able to map the whole parameter space in a practicable amount of time, we used a combination of evolutionary optimisation and data mining as our experiment design strategy for the reduction of the experimental parameter space. It became apparent that a throughput of 10,000 substances per day was only possible at the cost of abstraction and simplification with a con- comitant reduction in knowledge gain per individual experi- ment. Because not every reaction is suited to be simplified in such a manner, whether a high-throughput screening approach is effective or not has to be considered on a case by case basis. Methods of synthesis and testing to achieve a throughput of 10,000 catalysts per day. Macromol. Rapid Commun. 2004, 25, 169–177 DOI: 10.1002/marc.200300171 ß 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Feature Article 169