Coke on Supported Palladium and Platinum Catalysts

Doctoral thesis, 1997

Catalyst deactivation by coke formation has been investigated in three heterogeneous catalytic processes, propane dehydrogenation, acetylene hydrogenation and hydrogenation of vegetable oil. Palladium and/or platinum supported on alumina were used as catalysts in all reactions. Coke formed during propane dehydrogenation has been studied by weighing in a microbalance reactor and by temperature-programmed oxidation (TPO). Furthermore, the coverage of the metal surface has been investigated by reacting hydrogen and deuterium on coked catalysts. It was found that coke covered the metal surface immediately after the catalyst was exposed to the reaction mixture. It was also found that the presence of hydrogen reduced the deactivation and the coke formation rate. It was even possible to activate the catalyst by changing the reaction conditions to a more hydrogen-rich mixture. This "reversible" deactivation behavior was investigated further in transient experiments, and a model was proposed which included reversible and irreversible coke. The decrease in selectivity to ethene formation with time on stream, during acetylene hydrogenation, was investigated on palladium catalysts. It was found that the change in selectivity was not an effect of coke concentration. Instead, a low surface coverage of hydrogen resulted in less overall coke formation, but the coke was more harmful and affected the selectivity for ethene formation negatively. From experiments, in which deuterium was added to the reaction mixture and the composition of the formed coke was analyzed by TPO, it was proposed that the coke was formed via half-hydrogenated surface species (C2H3). Hydrogenation of vegetable oil in liquid phase was carried out using palladium and platinum catalysts supported on .alfa.- and .gamma.-alumina. It was found that only palladium on .gamma.-alumina deactivated during the hydrogenation process. A mechanism was suggested where conjugated dienes, which are formed to a larger extent on palladium than on platinum, were responsible for the formation of the coke that deactivated the catalyst. A discussion of general aspects in modeling deactivating processes is also included. It was found that when determining steady-state kinetics of a deactivating catalyst, errors are induced if extrapolation of the reaction rate to time zero is used. Instead, a novel transient method was proposed.