Item

Abstract

The economically important dehydrogenation of ethylbenzene (EB) to styrene (ST) is run at 870 K over potassium promoted iron oxide catalysts in the presence of steam. Here we present the continuation of model catalysis studies using thin epitaxial films (10-20 nm) of magnetite Fe3O4, hematite -Fe2O3 and potassium ferrites KFexOy of different composition. They allow the application of surface science methods for pre- and post-reaction surface characterization. The phases are identified from the LEED pattern and Auger (AES) spectrum. Catalytic conversions are measured using an in-situ stagnation point micro-flow reactor. The standard feed consists of EB: H2O=1:10 (molar). The partial pressures of reactive gases (p(EB) and p(H2O) are 3.3 and 33 respectively, the rest working pressure of 1 bar is He. The standard total flow rate is 25 ml/min. In addition, O2 can be introduced or H2O can be switched off. The low film thickness is an advantage; the whole "bulk" material is essentially in thermodynamic equilibrium at reaction conditions. The nature of the films also excludes pore diffusion effects. When starting with Fe2O3 the high initial conversion drops within about 80 min by about an order of magnitude. Simultaneously, the film is reduced to Fe3O4 and covered by coke. Without water in the feed, the deactivation behaviour is similar but substrate reduction proceeds towards metallic iron and coking is heavier. This confirms that water is not involved in the catalytic reaction but it prevents substrate reduction beyond Fe3O4 and help in coke gasification. We conclude further that the final low activity is connected to coke. Addition of O2 prevents both reduction and coking and the high initial conversion is maintained. The necessary amount of O2 corresponds to that necessary to oxidize the produced H2 and to gasify the coking products. Initial deactivation of the promoted catalyst is much slower and occurs mainly by coking which is counteracted by surface K, possibly as K2CO3. K accumulates at the surface but is in equilibrium with the bulk concentration. It is slowly but continuously removed during reaction which accounts for a slow irreversible deactivation. "Steaming" the catalyst by switching off EB but continuing the water admission causes removal of coke but also a fast removal or K. We conclude that in the presence of EB and its reaction- and decomposition-products, surface K is protected against removal, possibly because it forms carbonate. Substrate reduction is also prevented by K. Striking is the similarity of the initial conversion rates for unpromoted and promoted catalysts. Since Fe3O4 also contains Fe3+ but is much less active, we believe that the existence of Fe3+ is necessary but not sufficient. As proposed before, the adsorption strength for EB and St may be essential. Both are bound so tightly on Fe3O4 that they may block the active sites. Investigations over the promoted catalyst with different potassium loadings show that the potassium rich phase (KFeO2) is not the active phase, the ß-ferrite KxFe22O34 is the active phase and the KFeO2 is the potassium storage phase. The activation energy was found to be equal on the clean unpromoted and the promoted catalyst, which support that the active sites on both is the same. Unpromoted real catalysts in the form of pressed powder pellets were used for reactivity studies in the micro-flow reactor under the same conditions as for the model catalyst. The results show a good agreement but the conversion on real powder samples does not scale with the BET-surface, which means that there are diffusion limitations and not all the pores are accessible.