Chemical Engineering Science 62 (2007) 5338 5343 www.elsevier.

com/locate/ces

Development of a complete kinetic model for the FischerTropsch synthesis over Co/Al2O3 catalysts
Carlo Giorgio Visconti a , Enrico Tronconi a , , Luca Lietti a , Roberto Zennaro b , Pio Forzatti a
a Dipartimento di Chimica, Materiali e Ingegneria Chimica Giulio Natta, Politecnico di Milano, Piazza Leonardo da Vinci, 32 20133 Milano, Italy b Eni, Divisione Rening & Marketing, Via Felice Maritano, 26 20097 San Donato Milanese, Italy

Received 16 June 2006; received in revised form 18 December 2006; accepted 26 December 2006 Available online 16 January 2007

Abstract A global kinetic model of the FTS over a Co/Al2 O3 state-of-the-art catalyst is developed in a xed bed micro-reactor under conditions relevant to industrial operation (temperature, 210235 C; pressure, 825 bar; H2 /CO feed molar ratio, 1.82.7; gas hourly space velocity, 20007000 cm3 (STP)/h/gcatalyst ). On the basis of proposed reaction mechanisms, developed according to the carbide theory and the alkyl mechanism, the kinetic expressions for n-parafns and -olens formation are derived. Both the calculated CO conversion and the hydrocarbon distribution (up to n = 49) in FT reaction are satisfactorily predicted. 2007 Elsevier Ltd. All rights reserved.
Keywords: FischerTropsch synthesis; Kinetics; Catalysis; Catalyst selectivity; Cobalt; Mathematical modeling

1. Introduction The FischerTropsch synthesis (FTS) is a catalytic process that converts CO and H2 into a mixture of linear gaseous, liquid and solid hydrocarbons. The main reaction involved in this process can be schematically written as nCO + 2nH2 (CH2 )n +nH2 O HR = 165 kJ/mol. This synthesis has received considerable attention in recent years by both the industrial and academic worlds as a way of exploiting the huge natural gas reserves located in remote areas, leading to high-grade fuels. In fact FischerTropsch derived products are excellent high-performance and clean diesel fuels due to their high cetane number and the absence of sulfur and aromatic compounds. Cobalt-based catalysts have been successfully applied in the industrial processes due to their high FT activity and their
Corresponding author. Tel.: +39 02 23993264; fax: +39 02 70638173.

E-mail address: enrico.tronconi@polimi.it (E. Tronconi). 0009-2509/$ - see front matter 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.ces.2006.12.064

low oxygenates selectivity, which makes this catalysts suitable for the H2 -rich syngas obtained from natural gas. The kinetic description of the FT reaction is a very important task for the industrial practice, being a prerequisite for the industrial FT process design, optimization and simulation. The complexity of the reaction products, however, makes the kinetics of the process quite hard to be accurately described. Two different approaches for the development of FTS kinetic models have been reported in the literature (Pinna et al., 2002). In the rst one a rate law for reactants conversion (often based on empiric laws) and a product distribution model (like AndersonSchulzFlory, ASF, Anderson, 1984) are separately developed. This is theoretically justied only if it can be assumed that the reaction products do not affect or participate in the monomer formation mechanism. In the second more sound approach, all the mechanistic steps in which the CO and H2 are consumed and lead to the nal products are considered jointly. A major shortcoming of the publications implementing this approach so far, however, is the introduction of more or less empirical parameters, such as the chain growth probability ( ), in order to describe the hydrocarbon distribution (Wang et al., 2003; Lox and Froment, 1993). In addition these kinetic treatments often refer to iron-based catalysts and are able to t

C.G. Visconti et al. / Chemical Engineering Science 62 (2007) 5338 5343

5339

the hydrocarbons selectivity up to products with carbon atoms number n = 20 (Teng et al., 2006; Yang et al., 2003). In this work a more fundamental approach is reported. Mostly based on chemical enrichment experiments, in which we analyzed the effects of the parafns, olens and alcohols addition to the syngas feed on the product distribution (Fiore et al., 2004), we have at rst dened a detailed FTS mechanism for a cobalt-based catalyst, explaining the synthesis of each product through the evolution of reaction intermediates and adsorbed species. Accordingly, appropriate rate laws have been attributed to each elementary step and the resulting kinetic scheme has been tted to a comprehensive set of FTS runs. The developed model allows to predict simultaneously both the CO conversion and the n-parafns and -olens selectivity up to n = 49, on the basis of the process conditions. The model does not include so far the formation of oxygenates and CO2 , the selectivity of these species being very low with the adopted catalyst. Anyway, a modied kinetic model, considering also oxygenates and CO2 formation is presently under development. 2. Experimental The kinetic measurements were carried out in a fully automated experimental set-up, consisting of a xed bed microreactor loaded with 2 g of a Co/Al2 O3 state-of-art catalyst in powder form (mean average diameter 75 m) to prevent diffusional limitations. The detailed description of the experimental set-up and products analysis can be found elsewhere (Fiore et al., 2004). Experimental conditions were varied in the following ranges: P = 825 bar, T = 210235 C, H2 /CO feed ratio = 1.8 2.7 mol/mol, GHSV = 20007000 cm3 (STP)/h/gcat . In particular, 46 experimental steady state runs were performed following the experimental plan reported in Table 1. Each experimental condition was replicated several times in order to verify the experimental data accuracy and reproducibility. Moreover, the central point (P = 20 bar , T = 230 C, H2 /CO
Table 1 Experimental plan
No. P (bar) 20 8 25 20 20 20 20 20 20 20 20 20 20 20 20 20 H2 /CO ratio (mol/mol) 2.1 2.1 2.1 2.1 1.8 2.3 2.7 2.1 2.1 2.1 2.1 2.1 2.1 2.1 2.1 2.1 T ( C) 230 230 230 230 230 230 230 230 220 235 230 230 230 230 210 220 GHSV (cm3 (STP)/h/gcat ) 5000 5000 5000 5000 5000 5000 5000 5000 5000 5000 5000 4000 7000 5000 2000 2000

feed ratio = 2.1 mol/mol, GHSV = 5000 cm3 (STP)/h/gcat ) was replicated 12 times at constant intervals as an activity check. 3. Results 3.1. Reaction scheme Although great efforts have been devoted to elucidate the FTS reaction mechanism (Yang et al., 2003), there are still many controversies on this point. To develop a comprehensive FTS kinetic model, the following scheme (and the related kinetic rate equations), developed on the basis of a literature analysis and of chemical enrichment experiments (Fiore et al., 2004) and composed by elementary reactions, is adopted in this work: H2 + 2 2H CO+ CO rH2 = kH2 PH2 rM = kM PCO
2

1 H

(1) (2a) (2b) (2c) (2d) (2e) (2f)

CO + C + O C + H CH +
CH + H CH 2+

O + H OH + OH + H H2 O + 2
CH 2 + H CH3 + CH 3 + H CH4 + 2 Rn + CH 2 Rn+1 + Rn + H Pn + 2 Rn On + H O2 + H R2

rI N = k I N

CH 2

(3)
H

rCH4 = kCH4 rG,n = kG

Rn Rn

CH 3

(4)

CH 2 H

n : 1, 49 (5) (6)

rP ,n = kPn

n : 2, 50

rO,n = kOn ,dx rO,2 = kOn ,dx

Rn

kOn ,sx xOn H n : 3, 50 (7) kO2 ,sx xO2

H

R2

(8)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

where Pn and On are the generic linear parafn and -olen is the generic linear growwith n carbon atoms, respectively, Rn ing adsorbed hydrocarbon species, is the fraction of free catalytic sites, i is the fraction of the catalytic sites occupied by species i and xOn is the molar fraction of the -olen n in the liquid phase surrounding the catalyst pellets. In the proposed mechanism, H2 adsorbs irreversibly on two free catalytic sites () in the dissociated state (Eq. (1)), while the CO adsorbs rstly in the molecular state (Eq. (2a)) and then in the dissociated state (Eq. (2b)). Also the formation of the monomeric species CH 2 (in accordance with the carbide theory) was assumed to occur via two steps in series: the reaction between the surface carbon species and the surface hydrogen to form the species CH (Eq. (2c)) and the reaction between this species and the surface hydrogen to form the monomer CH 2 (Eq. (2d)). Concerning the chain growth mechanism we adopted the alkyl mechanism, i.e., we assume that the reaction

5340

C.G. Visconti et al. / Chemical Engineering Science 62 (2007) 5338 5343

Table 2 Estimates of the kinetic parameters for the proposed mechanism

Parameter k H2 E H2 kM EM kIN kCH4 Value 3.77 10 267.30 1.36 103 177.44 3.34 101 1.61
5

Unit mmol s1 g1 bar 1 kJ mol mmol s1 g1 bar 1 kJ mol1 mmol s1 g1 mmol s1 g1

1

Parameter kG k Pn kOn ,dx kOn ,sx kO2 ,sx

Value 2.26 10 1.32 101 5.54 103 3.63 1.64 10+2

1

Unit mmol s1 g1 mmol s1 g1 mmol s1 g1 mmol s1 g1 bar 1 mmol s1 g1 bar 1

Fig. 1. Parity plot for calculated and experimental CO conversion.

is initiated by the formation of a methyl species (Eq. (3)) and that the chain growth takes place by the successive insertion of methylene into the active site-alkyl bond (Eq. (5)). In addition it was assumed that methane synthesis occurs via the reaction between a methylene species and a surface hydrogen atom (Eq. (4)) and that water is formed via two consecutive steps involving the reaction between the adsorbed oxygen atom and a surface hydrogen atom (Eq. (2e)), and the reaction between the OH species and another surface hydrogen atom (Eq. (2f)). Termination of the chain growth occurs via two routes. In particular we assumed a dual-site reaction between and an adsorbed hydrogen atom (Eq. (6)) the intermediate Rn for the desorption of a parafn while, for the formation of an olen, we assumed a reversible -hydride elimination reaction (Eq. (7)), with a specic kinetic in the case of ethylene (Eq. (8)) in order to account for the anomalous reactivity of this species. 3.2. Kinetic model An elementary rate law was assigned to each step involved in the detailed reaction mechanism of FTS. Only for the H2 adsorption we adopted a power-law expression.

Furthermore, we assumed that step (2a) is rate determining in the sequence of consecutive non-reversible steps (2a), (2b), (2c), (2d), (2e) and (2f). Accordingly, the overall rate of all such steps is herein described by the rate expression for nondissociative CO adsorption. It was also assumed that the rate constants describing the elementary steps for the growth of the adsorbed species Rn (Eq. (5)), the formation of the parafns C2+ (Eq. (6)) and the formation of the olens C3+ (Eq. (7)) are independent of the carbon atoms number of the intermediates involved in the elementary reactions. On the contrary, in order to describe the experimental deviation of methane and ethylene from the ASF product distribution, a specicity was assumed in the kinetic constants involved in the kinetic expressions for the formation of these species (Eqs. (4) and (8)), extending the approach used by Wang et al. (2003) for methane. On the basis of the experimental observation that the reaction temperature, in the investigated range (210235 C), affects the CO conversion but not the products selectivity (as also reported e.g. by Yates and Sattereld, 1992), we introduced in the model the activation energies only for steps (1), (2) and (3) i.e., the steps involved in the monomer formation, responsible for CO conversion, but not for the product distribution. However, since preliminary regression of experimental data indicated negligible activation energy for step (3), in the following analysis only steps (1) and (2) will be regarded as activated and T-dependent. 3.3. Reactor model The reactor model adopted for describing the lab-scale experimental set-up is an isothermal homogeneous plug-ow model. It is composed of 2NP+2 ordinary differential equations of the type (9) with the initial conditions (10), NP + 2 algebraic equations of the type (11) and the catalytic site balance (12): dFi = dWcat
NR i,k k =1

rk

(9) (10) (11)

NP

Wcat = 0,
NR

Fi = Fi,0 rk

0=
k =1

j,k

1= +

CH 2

+
n=1

Rn

(12)

C.G. Visconti et al. / Chemical Engineering Science 62 (2007) 5338 5343

5341

Fig. 2. Typical model ts.

where Fi and Fi,0 are the molar ows of the generic species i (i = CO, H2 , H2 O, CH4 , Pn , C2 H4 , On ) along the reactor axis and at the reactor inlet, respectively, Wcat the catalyst mass, i,k and j,k the stoichiometric coefcients for the i th and for the j th (j = H , CH 2 , Rn n : 1 N P ) components, respectively, in the k th reaction, rk the rate of the k th reaction and NR and NP the number of the reactions and of the products, respectively, involved in the process. The molar fraction of the -olen n in the liquid phase surrounding the catalyst pellet, xOn , is evaluated by means of an explicit correlation, expressing the xOn dependence from temperature, number of carbon atoms and partial pressure of the corresponding olen in the gas phase, obtained on the basis of an external VLE (Vapour Liquid Equilibrium) calculation based on the RKS cubic equation of state. 3.4. Optimization method The non linear regression was performed using the Fortran subroutine BURENL (Donati and Buzzi-Ferraris, 1974) based

on the least-squares method. The algebraic-differential system constituted by Eqs. (9), (11) and (12) was integrated numerically with the Fortran subroutine LSODI (Hindmarsh, 1983), that allows to solve stiff problems using implicit integration methods with variable step. In order to obtain the best t of the 46 experimental data considered, we adopted 11 adaptive parameters. Extradiagonal terms in the correlation matrix were lower than 0.9 in over 90% of the cases. The regression was performed using as experimental responses the CO conversion and the CH4 , C2 H4 , C2 H6 , C3 H6 , C3 H8 , C6 H12 , C6 H14 , C10 H20 , C10 H22 , C12 H26 , C14 H30 , C16 H34 , C18 H38 , C22 H46 , C26 H54 , C30 H62 , C5+ and olens selectivity. 3.5. Results and discussion The kinetic parameters estimated upon regression of the experimental data set are listed in Table 2. As previously reported only steps (1) and (2a) are activated and T-dependent;

5342

C.G. Visconti et al. / Chemical Engineering Science 62 (2007) 5338 5343

Fig. 3. Effect of the H2 /CO feed molar ratio on the product distribution: comparison between (A) calculated and (B) experimental data (T = 230 C, P = 20 bar , GHSV = 5000 cm3 (STP)/h/gcat ).

accordingly activation energy values were obtained for these steps only. The estimated values are high, suggesting that the adsorption of the reactants on the Co based catalyst represents the rate limiting step. Anyway, these aspects need further clarication, and are presently under investigation. Using the parameters reported in Table 2 the model satisfactorily predicts both the CO conversion and the observed product distribution up to n = 49, in terms of total hydrocarbons, parafns and olens for all the experimental conditions considered in this study. The parity plot in Fig. 1 shows the ability of the model to estimate the CO conversion at the conditions previously reported. The average relative error calculated on these data is 14.5%, a satisfactory value which is directly comparable to those reported by Yang et al. (2003) in terms of CO conversion and by Teng et al. (2006) in terms of syngas consumption rate. In terms of product distribution, Fig. 2 shows some typical model ts at different process conditions. The model satisfactorily describes the products selectivity and also accounts for the typical deviations of the product distribution from the ASF model, namely the high methane selectivity, the low selectivity to C2 species and the change of the ASF slope with growing carbon atoms number. It is worth noticing that the model predicts the hydrocarbons selectivity up to n = 49, a range of product distribution which is wide if compared to other literature works. Notably, with a single set of rate parameter estimates, the present model can also correctly describe the small effects of all the investigated process conditions on both CO conversion and product distribution. For example, Fig. 3 compares the experimental and the calculated ASF product distributions as a function of H2 /CO feed ratio. Model predictions indicate a decrease in the heavy product distribution on increasing the H2 /CO ratio (Fig. 3A): the same trend is apparent in the experimental data (Fig. 3B), although the effect is less evident.

Fig. 4. Effect of temperature on CO conversion: comparison between calculated and experimental data (P = 20 bar , GHSV = 5000 cm3 (STP)/h/gcat , H2 /CO feed ratio = 2.1 mol/mol).

Fig. 4, on the other hand, shows the experimental and predicted CO conversions upon changing the reaction temperature: again, model predictions well describe experimental results. Inspection of the calculated surface coverage of the intermediate species nally revealed that the surface concentration of is typically of the same order of magnitude as the species Rn that of CH , 2 i.e., the C1 species associated with CO adsorption/conversion. This implies that the coverage of catalytic sites by the synthesis products has a signicant inuence on the CO conversion rate, which conicts with the traditional approach of developing separate models for CO conversion and for product distribution.

C.G. Visconti et al. / Chemical Engineering Science 62 (2007) 5338 5343

5343

4. Conclusions A complete kinetic model for FTS over a state-of-the-art Co catalyst is derived, on the basis of the carbide theory and of the CH2 insertion alkyl mechanism. For the range of industrially relevant conditions, the developed model could accurately predict both the observed CO conversion and the product distribution up to n = 49, in terms of total hydrocarbons, n-parafns and -olens. In particular the model is able to describe the typical deviations of the product distribution from the ASF model, i.e., the methane high selectivity, the low selectivity to C2 species and the change of the slope with growing carbon atoms number. Finally, the model can also correctly describe the effects of all the investigated process conditions on both CO conversion and product distribution. Accordingly it can be applied to identify optimized process conditions which are suitable to grant the desired conversion with the requested product distribution. References
Anderson, R.B., 1984. The FischerTropsch Synthesis. Academic Press, Orlando, FL. Donati, G., Buzzi-Ferraris, G., 1974. Chemical Engineering Science 29, 1504.

Fiore, F., Lietti, L., Pederzani, G., Tronconi, E., Zennaro, R., Forzatti, P., 2004. Reactivity of parafns, olens and alcohols during FischerTropsch synthesis on a Co/Al2 O3 catalyst. Studies in Surface Science and Catalysis 147, 289294. Hindmarsh, A.C., 1983. ODEPACK, A Systematized Collection of ODE Solvers. In: Stepleman, R.S. et al. (Eds.), Scientic Computing. NorthHolland, Amsterdam, pp. 5564. Lox, E.S., Froment, F., 1993. Kinetics of the FischerTropsch reaction on a precipitated promoted iron catalyst. 2. Kinetic modeling. Industrial and Engineering Chemistry Research 32, 7182. Pinna, D., Tronconi, E., Lietti, L., Zennaro, R., Forzatti, P., 2002. Rassegna di modelli cinetici per la sintesi di FischerTropsch. La rivista dei combustibili 56 (2), 6985. Teng, B.T., Chang, J., Zhang, C.H., Cao, D.B., Yang, J., Liu, Y., Guo, X.H., Xiang, H.W., Li, Y.W., 2006. A comprehensive kinetics model of FischerTropsch synthesis over an industrial FeMn catalyst. Applied Catalysis A: General 301, 3950. Wang, Y.N., Ma, W.P., Lu, Y.J., Yang, J., Xu, Y.Y., Xiang, H.W., Li, Y.W., Zhao, Y.L., Zhang, B.J., 2003. Kinetics modeling of FischerTropsch synthesis over an industrial FeCuK catalyst. Fuel 82, 195213. Yang, J., Liu, Y., Chang, J., Wang, Y.N., Bai, L., Xu, Y.Y., Xiang, H.W., Li, Y.W., Zhong, B., 2003. Detailed kinetics of FischerTropsch synthesis on an industrial FeMn catalyst. Industrial and Engineering Chemistry Research 42, 50665090. Yates, I.C., Sattereld, C.N., 1992. Hydrocarbon selectivity from cobalt FischerTropsch catalyst. Energy & Fuels 6, 308314.