Three different approaches to model hydrocarbon formation rates of Fischer-Tropsch synthesis have been compared. The model parameters have been determined from experimental data obtained in a stirred tank slurry reactor (STSR) over a wide range of process conditions (temperature: 220, 240 and 260°C; pressure: 8, 15 and 25 bar; feed composition, H₂/CO ratio: 2/3 or 2/1; gas space velocity: 0.5-23.5 NL/g-Fe/h) on a precipitated iron catalyst (100 Fe/4.3 Cu/4.1 K/25 SiO₂). These data are characterized by non-straight line hydrocarbon products distribution, and decrease of 1-olefin with simultaneous increase of n-paraffin and 2-olefin selectivities with conversion and carbon number. The proposed models provide predictions of concentrations of linear paraffins and linear 1- and 2-olefins..

All considered models assume that the chain growth initiates on an active sites of the catalyst by hydrogenation of adsorbed monomer (CH₂,s) to adsorbed methyl group (CH₃,s). Chain propagation occurs via insertion of adsorbed monomer into adsorbed alkyl species (C_nH_2n+1,s), which can terminate to n-paraffin (C_nH_2n+2) by hydrogenation, and to 1-olefin (1-C_nH_2n) or 2-olefin (2-C_nH_2n) by dehydrogenation. The a- and b-olefins (1- and 2-olefins) are considered separately in these kinetic models. All models assume that the 1-olefin can readsorb and form the adsorbed alkyl species, which can subsequently propagate or terminate.

The first model is based on that proposed by Zimmerman et al. (Zimmerman, 1990; Zimmerman et al., 1992). It assumes two types of active sites (s₁ and s₂) on the catalyst surface. Chain initiation, propagation and termination to products as well as 1-olefin readsorption take place on the first type of active sites on the catalyst surface. The 1-olefin can readsorb also on the second type of active sites, different than those on which initiation and propagation occur, leading to adsorbed alkyl species (C_nH_2n+1,s₂). The latter species can also terminate to n-paraffin by hydrogenation and to 1-olefin or 2-olefin by dehydrogenation. There is no propagation on the second type of site. This represents an extension of the original model proposed by Zimmerman et al. (1992).

The second approach is based on the selectivity model developed by Van der Laan and Beenackers (1998). They simplified the Zimmerman's model by considering only one type of sites. This model has been called “olefin readsorption product distribution model” (ORPDM). The original ORPDM considers total olefin formation, i.e. it does not treat formation of 1- and 2-olefins separately. In this work the ORPDM model has been modified by considering formation of both 1- and 2-olefins separately.

The third selectivity model tested in this work was based on the model developed by Nowicki et al. (Nowicki, 2000; Nowicki et al., 2001; Nowicki and Bukur, 2001) and will be referred here as Nowicki model. Nowicki's model is an extension of the Zimmerman's model which includes propagation on the second type of active sites and considers formation of 2-olefins as well. Chain growth is initiated only on the first type of active sites by hydrogenation of adsorbed monomer (CH₂,s₁) to adsorbed methyl group (CH₃,s₁). Termination to 2-olefins occurs only on the second type of active sites.

In all models, the formation of C₁ and C₂ products is considered separately therefore their rate constants are different than others. Also, an assumption has been made that the reaction rate of olefin formation is proportional to its partial pressure in the gas phase and that the reactor behaves as a perfectly mixed flow reactor.

A non-negative values for all model parameters have been obtained and the best fit of experimental data was obtained for the Nowicki's model, and the worst for the ORPDM. However, the differences in predictions between these models are less then 6%. Therefore the ORPDM may be considered as the most preferred model due to its simplicity (i.e. lower number of kinetic parameters).

The Levenberg-Marquardt and the trust-region reflective Newton large-scale (LS) method were successfully employed for minimization of the objective function and kinetic parameter estimation. The 95% confidence intervals of the parameters have been obtained.

References

Nowicki, L., Modelowanie kinetyki reakcji uwodornienia tlenków węgla na wybranych typach katalizatorów w uk³adzie gaz-ciecz-cia³o sta³e, £ódź: Politechnika £ódzka (2000)

Nowicki, L., Ledakowicz, S., Bukur, D.B., Chem. Eng. Sci., 56, 1175-1180 (2001)

Nowicki, L. and Bukur, D.B., Studies in Surface Science and Catalysis, 136, J. J. Spivey, E. Iglesia and T. H. Fleisch (Editors), 2001 Elsevier Science B.V., pp. 123-128 (2001)

Van der Laan, G.P. and Beenackers, A.A.C.M., Studies in Surface Science and Catalysis, 119, 179 (1998)

Zimmerman, W. H., “Kinetic Modeling of the Fischer-Tropsch Synthesis”, Ph.D. dissertation, Texas A&M University, College Station, TX, 1990

Zimmerman, W., D.B. Bukur and S. Ledakowicz, Chem. Eng. Sci., 47, 2707 (1992)

Key words

Fischer-Tropsch, kinetic modeling