Kinetic and selectivity modelling of the iron-based low-temperature Fischer-Tropsch synthesis

The Fischer-Tropsch (FT) synthesis provides a method of converting syngas, obtained from the gasification of crude oil alternative sources such as coal, natural gas or biomass, to liquid fuels suitable for use in standard motor vehicles. The economic viability of a commercial FT plant is highly dependent on the crude oil price, which has fluctuated considerably during the past few decades. Probably due to the risk of an ill-timed slump in crude oil prices, world-wide application of the technology has been somewhat restricted. However, due to the current high oil prices, there is at present a renewed interest in coal and natural gas as alternative energy sources, and consequently the FT process is receiving wide-spread attention. The FT synthesis has been applied commercially in different forms, but the focus of this study has been specifically on the "iron-based Low-Temperature Fischer-Tropsch" (Fe-LTFT) synthesis. This process is characterised by the use of an alkali-promoted iron catalyst, synthesis temperatures around 240°C, and a product slate that extends well into the wax range. An accurate prediction of the process performance is very important for the design of a commercial FT plant. There are three main aspects that should be addressed in terms of the modelling of the Fe-LTFT synthesis, namely the rate of CO conversion to hydrocarbons (FT kinetics), the rate of CO conversion to CO2 (WGS kinetics) and the distribution of hydrocarbon products (product selectivities). FT kinetics Even though water (or carbon dioxide) has traditionally been included in FT rate expressions for the iron-FT synthesis, a critical literature review revealed that there is no conclusive evidence for the premise that either of these two components adversely affects the FT kinetics per se. Instead is was shown that the observed influence of water can also be explained by its effect on the water-gas-shift (WGS) reaction rate, which in turn affects the gas phase partial pressures of the reactants (CO and hydrogen) and therefore indirectly also the hydrocarbon formation rate. The FT rate equation proposed in this study did originally contain a water term, but after testing the expression against a variety of existing data sets, it was concluded that there was no statistical basis for including water in the kinetic model. Following an experimental design procedure, new kinetic data were measured which could conclusively discriminate between the traditional rate equations (accounting for an influence of water) and the following new kinetic expression (which assumes no influence of water on the FT kinetics): ( )2 0.5 1 2 CO CO H CO FT k P P P r A + = WGS kinetics Originally it was proposed in literature that the WGS kinetics can be described by a simple first order expression in CO, but more recently models derived from mechanisms based on the formation of a formate intermediate seem to be favoured. In this study it was shown that a first order rate equation in CO is a reasonable description of the WGS rate, but contains systematic deviations indicative of its empirical nature. It was further found that models based on the formate mechanism described the historic data the best. After evaluating the rival equations with newly measured data, the following expression emerged as the preferred WGS kinetic model: 2 0.5 2 2 2 2 2 2 2 1 1 ¿ ¿ ¿ ¿ ¿ ¿ ¿ ¿ + + - = H P P k P k P P K P P r A H O H O H O OH H CO WGS CO H O WGS The new kinetic models for the Fe-LTFT synthesis imply that the FT and WGS reactions occur on different types of sites. Whereas the FT sites are predominantly covered with adsorbed CO or C1 intermediates, the WGS sites are mostly covered with adsorbed water and OH species. Selectivity modelling Two types of product characterisation models for the FT synthesis are known in literature, namely double-a models and olefin reinsertion models. In this study it was shown that there is a high degree of cross-correlation between the independent parameters of double-a models. Due to the low propensity of iron-FT catalysts for secondary olefin reactions, it was also concluded that olefin reinsertion models are not appropriate for the Fe-LTFT synthesis. Consequently, a new product characterisation model was proposed. According to the chain length dependent desorption model, olefin and paraffin formation is governed by three generic reactions: chain growth by one carbon atom at a time; chain desorption, resulting in the formation of an olefin; chain hydrogenation, resulting in the formation of a paraffin. The cornerstone assumption is that the rates of chain growth and hydrogenation are independent of chain length, but that the rate of desorption is a function of carbon number due to the physisorption of the chain on the catalyst surface. The longer the chain, the greater the physisorption and the slower the rate of desorption relative to growth and hydrogenation; consequently, there is an increase in growth probability and saturation with chain length. The model could accurately describe the olefin and paraffin distributions in the C3 to C10 range. After making some mechanistically-rationalised adjustments to the model equations for the case of the C2 intermediate, the model could be extrapolated to the C1 and C2 products as well. This is a true extrapolation, as the C1 and C2 data were not used for the estimation of the parameter values. This may be the first product characterisation model that can successfully be extrapolated to the C1 and C2 components without introducing additional (unique) parameter values for these products. It was found that the chain length dependent desorption model overestimates the ethylene / ethane ratio and predicts a higher olefin / paraffin ratio for the C2 fraction than for the C3 fraction. Consistent with the results of ethylene co-feeding studies reported in literature, this was ascribed to the secondary hydrogenation of ethylene. Data measured in a laboratory slurry reactor operated under recycle were used as further support for some of the assumptions and implications of the product characterisation model. These results indicated negligible reinsertion rates of both ethylene and propylene, high rates of secondary ethylene hydrogenation and very low (almost negligible) rates of secondary propylene hydrogenation. After estimating rate constants for secondary hydrogenation, it was shown that the predicted primary ethylene / ethane ratio was higher than the predicted primary propylene / propane ratio, consistent with the chain length dependent desorption model.