Sei sulla pagina 1di 20

Reac Kinet Mech Cat (2011) 104:483–502

DOI 10.1007/s11144-011-0369-1

Development of a kinetic model of the Fischer–Tropsch


synthesis reaction with a cobalt-based catalyst

Seung-Ho Kwack • Myung-June Park •


Jong Wook Bae • Kyoung-Su Ha • Ki-Won Jun

Received: 6 April 2011 / Accepted: 23 August 2011 / Published online: 11 September 2011
Ó Akadémiai Kiadó, Budapest, Hungary 2011

Abstract A kinetic model of cobalt-based Fischer–Tropsch synthesis was devel-


oped through the detailed kinetic study of the reaction mechanism. Experimental
evidence and previously reported theoretical analyses were used to suggest the
mechanism and derive reaction rates for the formation of hydrocarbon products by
applying the equilibrium constants of the adsorbents and the quasi steady state
assumption to intermediate species on the surface of the catalyst. The comparison
between experimental data and simulated results with kinetic parameters validated
the effectiveness of the developed model. Further analysis showed that temperature
and H2/CO ratio significantly influenced the entire distribution of hydrocarbon
products. The effects of operating conditions were also predicted in accordance with
previous work, thus demonstrating that the developed model can contribute to a
better understanding of the kinetic mechanism of FT synthesis.

Keywords Cobalt-based catalyst  Fischer–Tropsch synthesis  Kinetics 


Parameter estimation  Mathematical modeling

List of symbols
c Concentration, mol/L
CP Heat capacity, J/g
CWP Weisz-Prater parameter, dimensionless

S.-H. Kwack  M.-J. Park (&)


Department of Chemical Engineering, Ajou University, Suwon 443-749, Korea
e-mail: mjpark@ajou.ac.kr

J. W. Bae
School of Chemical Engineering, Sungkyunkwan University, Suwon 440-746, Korea

K.-S. Ha  K.-W. Jun


Petroleum Displacement Technology Research Center, Korea Research Institute of Chemical
Technology (KRICT), Daejeon 305-600, Korea

123
484 S.-H. Kwack et al.

Dt Tube diameter, cm
E Activation energy, J/mol
Fobj Objective function
DH Heat of formation, J/mol
k Reaction rate constant, mol/kg/s
Kadi Adsorption equilibrium constant of i component, 1/bar
NE Number of experimental conditions
NR Number of reactions
On Olefin (CnH2n)
Pi Partial pressure of i component, bar
Pn Paraffin (CnH2n?2)
rj Rate of reaction j, mol/kg/s
Rn Alkylidene (CnH2n?1)
U Overall heat transfer coefficient, J/cm2/s/K
us Linear velocity, cm/s
v Stoichiometric coefficient
w Weighting factor
z Length of reactor, cm
Greek letters
h Fraction of surface intermediate
qB Bulk density, g/cm3
qg Gas density, g/cm3
Subscripts
i Species
j Reaction
q Objective

Introduction

The Fischer–Tropsch synthesis (FTS) is a promising alternative process for


converting coal, natural gas or biomass into environmentally less damaging fuels
and useful chemicals with low emissions of pollutants [1–4]. Group VIII metals are
commonly employed as Fischer–Tropsch (FT) catalysts. Since Rh, Ir, Pt and Pd
have relatively low activities, most research has focused on Ru, Fe, Co and Ni.
However, Ru is expensive and Ni can transform to nickel carbonyl under high
pressure. Therefore, Fe and Co are considered most adequate for commercial
processes. Cobalt-based catalysts have been successfully used commercially
because they show high activity and selectivity for heavy hydrocarbons. Other
benefits of Co-based catalysts include low oxygenate selectivity when using H2-rich
syngas obtained from natural gas [5], low carbon dioxide emission because of a low
water–gas shift reaction rate [6], and more sensitive product selectivity to H2/CO
ratio, pressure and temperature than iron-based catalysts [7]. Co-based FTS is
further favored since the reaction is little affected by water [8–10].

123
Development of a kinetic model of the Fischer–Tropsch 485

The development of a kinetic model of FTS is important for the simulation,


design, and optimization of commercial FT processes. However, detailed models
that can predict the entire distribution of hydrocarbon products are difficult to create
due to the great number of possible chain lengths. Although many kinetic models
lump products into groups [11–13], they give oversimplified information about
overall conversion and fail to describe the detailed distribution of hydrocarbon
products. Although models such as the Anderson-Schulz-Flory (ASF) model [14]
separately apply a rate law for reactant conversion and product distribution, they are
only applicable if the reaction products can be assumed not to affect or participate in
the monomer formation mechanism. To predict the entire product distribution, van
der Laan and Beenackers [15] developed a kinetic model based on a kinetic
mechanism with olefin re-adsorption reactions included, and the model has been
subsequently expanded [16, 17]. Lox and Froment [18] applied a stricter
methodology to propose a kinetic model that incorporates the fundamentals of
FTS catalysis and kinetics. Recently, kinetic expressions of paraffin and olefin
formation have been developed according to the carbide theory and the alkyl
mechanism by Visconti and coworkers [5].
FTS mechanisms are divided into three categories according to chain growth
pathways; CH2 insertion, CO insertion, and oxygenate (enol) mechanism. The CH2
insertion mechanism [19, 20] originally suggested by Fischer and Tropsch implies
that C ? C coupling is achieved through the polymerization of CH2 intermediates
on the surface (Fig. 1a). This mechanism prevailed until Brady and Pettit [21, 22]
demonstrated the importance of adsorbed CH2 species on the catalyst’s surface.
Brady and Pettit’s work was refined by experimental investigation by Quyoum and
coworkers [23], who showed, by co-feeding 13CH2N2 with syngas, that the 13CH2
intermediates react in the same way as the surface 12CH2 groups formed from 12CO/
H2, leading to their random incorporation into the hydrocarbons. Spectroscopic
studies have also provided direct evidence of the dissociation and the existence of
CH2 on the catalyst’s surface [24]. Cheng et al. [25] found through density
functional theory (DFT) that not only CH2 but also C and CH react with the alkyl
intermediate.
Pichler and Schulz [26] proposed a CO insertion mechanism, which assumes that
CO is directly inserted into the C ? C coupling of adsorbed alkyl intermediates
(Fig. 1b) [27]. This mechanism could explain the formation of oxygenates [20],
something that the CH2 insertion mechanism could not. DFT calculations have
shown that the CH2 species is the C1 species with the lowest stability on metal
surfaces (Co and Ru) [28, 29] and that CH2 ? CH2R (R=H or alkyl) has high
barriers [30], and thus chain growth by the CH2 insertion mechanism may be
difficult, suggesting that the CO insertion mechanism may be the dominant reaction
pathway [27].
In the 1950’s, the oxygenate (enol) mechanism gained widespread acceptance
[31]. It involves adsorbed hydroxyl methylene intermediates dimerizing to form
C ? C coupling (Fig. 1c) [14, 32]. 14C-tracer studies have provided strong support
for this mechanism [33–39].
Recently, Ojeda et al. [40] proposed the H-assisted CO dissociation to describe
the CO activation on cobalt FT catalysts. In their work, they proposed two reaction

123
486 S.-H. Kwack et al.

(a)
H H M
H branched CH2
H H M
M +CH2 M C R +CH2 M C R 1-alkenes C
H
(R=H) R CH3
M M M CH2

H H H M M CH2 M CH2
H H H +CH2
linear M C R M C R M CH2 M C R M C R M C R
1-alkenes M CH2 M CH2 M C R CH3 H CH3 H CH3
H

MH2C R +CH2
linear MH2C R +H
alkanes M CH2 M CH2
H

(b) M O
H COads O Hads 2Hads Hads COads O
M C M C H M CH2 M CH3 M C
M H -H2O CH3
H
2Hads

O OH
4Hads 2Hads
M CH2 M C M CH2 M C H
-H2O -H2O
(CH2)xCH3 (CH2)xCH3 CH3 CH3

H2
CH3(CH2)xCHO CH3(CH2)xCH2OH

H2
CH3(CH2)x-1CH=CH2 CH3(CH2)xCH3

(c) O O H OH H OH H OH H3C OH
+2H2 -H2O +H2
2CO C C C C C C C
M M M M M M M M

+CO
H3C
O H3C OH
H2C OH +H2 HO CH3 -H2O HO H H3C OH +H2
C C C C C C C
M M M M M M M M

Fig. 1 The mechanisms of the FT synthesis reaction. a CH2 insertion [19, 20], b CO insertion proposed
by Pichler and Schulz [26, 27], and c oxygenate (enol) [14, 19, 32]. All diagrams have been reproduced
from the corresponding references. CH2 and R=H added in the first step of a represent CH2 intermediates
in catalytic surface produced via the combination of carbon and hydrogen molecules in the surface from
the dissociative adsorption of syngas, and growing chains in the surface, respectively

pathways; one is H-assisted CO dissociation resulting in the production of H2O,


while the other (un-assisted CO dissociation) leads to the production of CO2.
In our previous work [41], we have used the published reaction rates [5] to find
the optimal catalyst composition for Co-based FT catalysts and then to quantita-
tively relate the effect of operating conditions on the properties of hydrocarbon
products. However, the mechanism in the literature [5] assumed the irreversible
adsorption of carbon monoxide and hydrogen molecules while there are many
published findings that the adsorption steps are reversible. In addition, the reaction
rates were not completely mechanistic since they included the empirical

123
Development of a kinetic model of the Fischer–Tropsch 487

expressions. Therefore, this work reports the development of a first-principles model


of Co-based FT synthesis based on a detailed kinetic mechanism in which CH2
insertion is dominant. Because of the very low production of oxygenate products in
the experimental tests (less than 3.4%), the CO insertion mechanism and the
oxygenate (enol) mechanism were assumed to be negligible. While most recent
kinetic models assume reversible reactions in the elementary steps and apply rapid
equilibria, the kinetic mechanism considered herein determines the reversibility of
each step through extensive analysis of previously reported experimental and
theoretical backgrounds.

Experimental

Catalyst preparation

In previous work, cobalt-based FT catalysts with various compositions have been


prepared and analyzed to optimize the catalyst with respect to activity [42]. The
catalyst was prepared on SiO2 supports (Davisil; pore size, 15.0 nm; surface area,
300 m2/g) that were modified with zirconium–phosphorous components through
stepwise impregnation with metal nitrate precursors, such as zirconium oxynitrate
hydrate (ZrO(NO3)2xH2O), cobalt(II) nitrate (Co(NO3)26H2O), ruthenium(III)
nitrosyl nitrate (Ru(NO)(NO3)3) and phosphoric acid (H3PO4). Typically, metals at
10 wt% on the SiO2 were co-impregnated with the metal nitrate precursors of
zirconium oxynitrate and phosphoric acid in deionized water. Samples were dried at
110 °C for at least 12 h with 0.2 wt% P in the zirconium–phosphorous-modified
SiO2. The cobalt precursor was successively impregnated in the slurry of the
zirconium–phosphorous-modified SiO2 support with 20 wt% cobalt metal. The Ru
precursor was then impregnated with 0.5 wt% metal based on the previously
prepared catalyst. The prepared catalysts were dried and calcined at 400 °C for 12 h
in air at 10 °C/min in the final step. The composition of the final catalyst was
0.5 wt% Ru/ 16.6 wt% Co/ 7.5 wt% ZrP/ 75.4 wt% SiO2, with a weight ratio of
P/(P ? Zr) of ca. 0.2.

Activity tests

Prior to activity testing, the catalysts were activated at 400 °C in a fixed-bed reactor
(I.D. = 12.7 mm) for 12 h with a 5% H2/He gas mixture. The activity tests were
carried out during the FTS reaction with hot-spot formation minimized by loading
0.3 g catalyst with average particle size of 80–120 lm. Activity testing was
conducted for ca. 70 h. Mathematical modeling was carried out using values of
steady-state activity under the following reaction conditions: T = 210–240 °C,
P = 1.0–2.5 MPa, SV = 3000–7000 L/kgcat/hr and molar ratio of H2/CO =
1.0–2.0. Effluent gas from the reactor was analyzed using an on-line gas
chromatograph (YoungLin Acme 6000 GC) that employed a GS-GASPRO capillary
column. The column was connected to a flame ionization detector (FID) for the
analysis of hydrocarbons. The chromatograph was also connected to a Porapak

123
488 S.-H. Kwack et al.

Q/molecular sieve (5A) packed column that was connected to a thermal conductive
detector (TCD) for the analysis of carbon oxides and hydrogen with Ar as internal
standard. CO conversion and hydrocarbon product distribution in the range of
C1–C7 were obtained by analyzing gas-phase products using on-line GC. In
addition, the selectivity of hydrocarbons above C8? was calculated by sampling the
wax component after reaction and analyzing it using off-line GC in which C-balance
was taken into account by combining the results of TCD and FID analyses. The CO
conversion and product distribution were averaged at 10 h on stream from the
beginning of the reaction as well as at steady-state after 60 h on stream. Further
details of product analysis are reported in a previous work [42].

Reaction mechanism and kinetic model

The kinetic mechanism considered in this work is outlined in Table 1, where R, Pn


and On represent alkylidene (CnH2n?1), paraffin (CnH2n?2) and olefin (CnH2n),
respectively. The chemisorption of hydrogen (step 1 in Table 1) is generally
assumed to be dissociative reversible adsorption [43–45]. Dissociative adsorption of
CO has been demonstrated by X-ray photoelectron spectroscopy (XPS) and pulse
techniques on Ni, Co, Ru, and Fe at elevated temperatures (T [ 350 K) [43].
Further analysis by Khodakov et al. [4] showed that it is almost irreversible under
the FT reaction conditions because of the irreversibility of carbon monoxide
dissociation. Therefore, the adsorption of CO (step 2) was assumed to be reversible
while the dissociation of adsorbed CO (step 5) was considered irreversible.
Kummer et al. [46] found that direct hydrogenation of metal carbide (steps 6 and
7) yields CH- and CH2-surface species, and the CH2 species, in turn, are used as
monomers for FT polymerization. When the monomer (CH2 species) is hydroge-
nated further, CH3 species—precursors for chain growth—are produced (step 10).
Van Barneveld and Ponec [47] reported that the formation of CH3 species is
essentially irreversible. The irreversibility for the hydrogenation of CH2-surface
species also suggests that the hydrogenation of surface carbon and CH-surface
species (steps 6 and 7) should be irreversible [48], although there are also published
findings that some of the hydrogenation steps are reversible [28]. The reaction of
dissociated oxygen atoms with hydrogen atoms (steps 8 and 9), which produces
water at the surface, has been suggested to be irreversible by many researchers
[49–51].
Chain growth begins from the combination reaction of CH3 and CH2 species to
produce alkylidene (R), which grows consecutively through further combination
with CH2 species [52]. Although both the chain growth of the paraffin (step 11) and
termination reactions (steps 12–14) are assumed to be irreversible [49, 50], Iglesia
reported—by analyzing the effects of residence time and olefin addition to H2/CO
feeds—that olefin re-adsorption (steps 15 and 16) plays a significant role in the
chain termination pathways [1]. Schulz et al. [53, 54] reported a possible reaction
mechanism for the re-adsorption of olefins followed by hydrogenation to paraffins
or isomerization to internal olefins via double-bond shift reactions.

123
Development of a kinetic model of the Fischer–Tropsch 489

Table 1 Reaction mechanism for Co-based FT synthesis reaction


No. Mechanism Remarks

1 H2 þ 2  2H h2
KHad2 ¼ P Hh2 Adsorption
H2 

2 CO þ   CO ad
KCO ¼ PhCO
CO
h

3 H2 O þ   H2 O h
KHad2 O ¼ PHH2OOh


4a On þ  ! On KOadn ¼ PhOOnh n2


n

5 kCO
CO þ !C þ O rCO ¼ kCO hCO h

6 kC
C þ H !CH þ  rC ¼ kC hC hH

7 kCH
CH þ H !CH2 þ  rCH ¼ kCH hCH hH

8 kOH
O þ H !OH þ  rOH ¼ kOH hO hH

9 kW
OH þ H !H2 O þ  rW ¼ kW hOH hH

10 kIN
CH2  þ H !CH3  þ  rIN ¼ kIN hCH2 hH Initiation

11 kG
Rn  þ CH2  !Rnþ1  þ  rG;n ¼ kG hRn hCH2 n1 Propagation

12 kCH4 rCH4 ¼ kCH4 hCH3 hH Termination


CH3  þ H !CH4 þ 2
13 kP2 rP;2 ¼ kP2 hR2 hH
R2  þ H !P2 þ 2
14 kPn rP;n ¼ kPn hRn hH n3
Rn  þ H !Pn þ 2
15 kO2 rO;2 ¼ kO2 hR2 h  kOrev2 hO2 hH
R2  þ 
rev
O2  þ H
kO
2

16 kOn rO;n ¼ kOn hRn h  kOrevn hOn hH n3


Rn  þ 
rev
On  þ H
kOn

a
Since the re-adsorption of olefins depends on their solubility
. in the liquid phase filling catalyst pores,
the adsorption equilibrium constant should be KOn ¼ hOn  alOn h where alOn represents the activity of
ad

olefins in the liquid phase. In the study, the vapor–liquid equilibrium with the modified Raoult law is
applied, and thus, the activity is replaced with the partial pressure of the corresponding species

Reaction rates were developed from the elementary steps listed in Table 1.
Details of the development of the kinetic model are provided in Appendix:
rCH4 ¼ kCH4 c1 cH =DEN ð1Þ
rP;2 ¼ kP2 ðA2 c1 þ B2 ÞcH =DEN ð2Þ
( !) ,
X n
n2 ni
rP;n ¼ kPn A A 2 c1 þ A Bi cH DEN n3 ð3Þ
i¼2
n o.
rO;2 ¼ kO2 ðA2 c1 þ B2 Þ  kOrev2 KOadn cH PO2 DEN ð4Þ

123
490 S.-H. Kwack et al.

" ( !) #,
X
NP
rO;n ¼ kOn An2 A2 c1 þ Ani Bi  kOrevn KOadn cH POn DEN n  3 ð5Þ
i¼2

where
" ( ! !)#2
X
NP NP X
X k
k2 ki
DEN ¼ 1 þ cH þ cCH2 þ 1þ A A2 c 1 þ A Bi
k¼2 k¼2 i¼2

kG cCH2 kOrev2 KOadn cH PO2


A2 ¼ ; B2 ¼
kG cCH2 þ kP2 cH þ kO2 kG cCH2 þ kP2 cH þ kO2
kG cCH2 kOrevn KOadn cH POn
An ¼ ; Bn ¼
kG cCH2 þ kPn cH þ kOn kG cCH2 þ kPn cH þ kOn
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
kIN cCH2 cH   1=2 b 2 þ b22 þ 4b1 b3
c1 ¼ ; cH ¼ KHad2 PH2 ; cCH2 ¼
kG cCH2 þ kCH4 cH 2b1
1 
P
kG kOrevn KOadn POn
kG kIN n¼2 ad
b1 ¼ ; b2 ¼ kIN cH þ ; b3 ¼ kCO KCO PCO
kPn þ kOn c1
H kPn þ kOn c1 H

where r, kj, and Kadi denote reaction rates in mol/kg/s, reaction rate constants for the
elementary reaction, j, and adsorption equilibrium constants of species i, respec-
tively. Pi represents the partial pressure of species i. There are many reports
showing values of the C1–C3 range deviating from ASF plots [55]. The amount of
methane produced is usually higher than predicted, particularly over Co and Ru
catalysts. C2 and C3 are produced less than predicted by ASF. This is supported by
ethene and propene being able to polymerize quickly to heavy FT products over a
Co catalyst in the presence of H2, even in the absence of CO [56, 57]. Based on the
above discussions, the termination reaction constants of methane (kCH4 ) and C2 (kP2 ,
kO2 and kOrev2 ) were defined separately.

Results and discussion

Parameter estimation

Small catalyst particle sizes allowed no mass transfer resistance to be assumed. The
dimensionless Mears parameter in this study was below 0.15, indicative of
negligible external mass diffusion [58]. The occurrence of any internal pore
diffusion limitation was determined by the Weisz-Prater criterion, where the
dimensionless Weisz-Prater parameter was calculated assuming that the catalyst
particles were filled with liquid waxes. Because CWP \ 0.01 under all the
experimental conditions [58, 59], no internal diffusion limitation existed. The
dispersion number of less than 0.01, defined as the ratio of the transport rate by
dispersion to that by convection (the reciprocal of the dimensionless Peclet

123
Development of a kinetic model of the Fischer–Tropsch 491

number), showed that dispersion contributed little compared to transport, and thus, a
pseudo-homogeneous plug flow model without dispersion was used to simulate the
reactor as follows:
dðus ci Þ X
NR
Mass balance :  qB vi;j rj ¼ 0 ð6Þ
dz j¼1
 
d us CP qg T XNR
4U
Energy balance : ¼ qB ðDH Þj rj þ ðTW  T Þ ð7Þ
dz j¼1
Dt

Boundary conditions : ci ¼ ci;0 ; T ¼ Tin at z ¼ 0 ð8Þ


where us, qB, qg, rj, and vi,j represent linear velocity, bulk and gas densities, reaction
rate and stoichiometric coefficient, respectively. Since the catalyst bed was assumed
to be not isothermal, energy balance was considered, and the overall heat transfer
coefficient was specified as 0.08 J/cm2/s/K, which had been determined in a
previous work [60], under the assumption of better heat transfer than the
commercial scale reactors with heat transfer coefficient around 0.01–0.02 J/cm2/s/
K. Other symbols are defined in the Nomenclature.
To estimate the kinetic parameters, the following objective function was
minimized using lsqcurvefit in MATLAB (The MathWorks, Inc.), where the
Levenberg–Marquardt method was applied:
XNE X  
Xq;calc  Xq;exp 2
Fobj ¼ wq ð9Þ
p q
Xq;exp

where NE and wq are the number of experimental conditions and the weighting
factor, respectively. The subscript q denotes the element of the objective function
(Table 2). Among the experimental conditions in Table 3, the temperature was
varied under four conditions (cases 1, 5–7 in Table 3). Therefore, parameter
estimation was conducted in two steps: First, all the kinetic parameters (total of 11)
were estimated under the conditions of constant temperature (cases 1–4 and 8–11 in
Table 3), where a total of 80 data points (10 elements in Table 2 multiplied by 8
experimental conditions) were used. The estimated results correspond to the kinetic
parameters at the reference temperature (cf. ki,ref in the caption of Table 4). Second,
the estimation was conducted using the data from cases 1 and 5–7 for varying
temperatures and only kCOKad CO, kG, kP2 , kOn and kO2 were assumed to be temperature
dependent, that is, they were considered in the form of ki = ki,ref exp[-Ei(1/T–1/
Tref)/R]. Meanwhile, the other parameters were assumed to be temperature-
independent, i.e. they were fixed at the values of ki,ref during estimation. Since kCH4 ,
which is directly related to methane selectivity, was not included in the second step,
SCH4 in Table 2 was excluded, and thus, the number of data used was 36 (9 elements
multiplied by 4 conditions). In order to select appropriate kinetic parameters for the
estimation of activation energies, some of kinetic parameters provided in Table 1
were selected in a combinatorial manner and considered temperature dependent. In
such a way, the selection of kCOKad
CO, kG, kP2 , kOn and kO2 resulted in the best fitness.

123
492 S.-H. Kwack et al.

The estimated parameters are listed in Table 4. Lower and upper bounds of 95%
confidence interval for estimated parameters are also provided in Table 4. Reported
values of activation energies for olefin formation are between 100 and 110 kJ/mol
[11], and FTS reaction has been reported to have activation energies between 63 and
132 kJ/mol in general [13, 15, 18]. The values of estimated parameters in the
present study coincide with these ranges. P
The average errors, defined as T.avg.error [%] = 100 9 i|(yi,exp-yi,calc)/yi,exp|,
for each element of the objective function are presented in Table 2, along with the
corresponding standard deviations of the errors. The average errors for all the
objective elements were high (ca. 10%), possibly accountable to measurement
errors, which were manifested in Fig. 2 as some experimental data deviate from the
trend with respect to operating conditions. It is also notable that the methane
selectivity (SCH4 ) has relatively high error even in the presence of a specific kinetic
parameter (kCH4 ). This is because the entire distribution is highly correlated and
thus, if we fit the methane fraction with higher resolution, the fitting results for other
objective elements such as averaged chain lengths and HC2 selectivity are
deteriorated. Since the distributions of middle distillates were assumed to be
relatively more important than the methane selectivity, we have reduced the
weighting factor of SCH4 in the objective for the estimation.
Under some experimental conditions (cases 2, 8, 9 in Table 3), experiments were
conducted in a repeated manner and the standard deviations for CO conversion,
methane selectivity, HC2 selectivity and C8? selectivity were found to be less than
20, 11, 13 and 17%, respectively. Therefore, the errors seem to be reasonable
compared to the values of measurement errors. In addition, as shown in Figs. 2 and
3 for the comparison of CO conversion, methane selectivity, olefin selectivity and
the entire distribution of hydrocarbons between experimental data and calculated
values, the calculated values of the objective elements are in satisfactorily
agreement with experimental observations despite relatively high values of
calculated errors. To evaluate the extent of the kinetic model’s coordination with

Table 2 Average errors and standard deviations of errors for each element of the objective function
q Element Error [%] Std. Dev. [%]

1 CO conversion (XCO) 9.85 6.00


2 Weight-averaged chain length of HC (xw,HC) 9.42 9.49
3 Number-averaged chain length of HC (xn,HC) 14.87 9.18
4 Weight-averaged chain length of paraffin (xw,P) 5.54 4.00
5 Number-averaged chain length of paraffin (xn,P) 6.74 5.21
6 Weight-averaged chain length of olefin (xw,O) 4.89 4.99
7 Number-averaged chain length of olefin (xn,O) 6.27 5.18
8 Olefin selectivity (SOn ) 15.49 13.72
9 Methane selectivity (SCH4 ) 21.98 17.27
10 HC2 (ethane and ethylene) selectivity (SHC2 ) 9.60 9.64
Average values 10.46 8.47

123
Development of a kinetic model of the Fischer–Tropsch 493

Table 3 Experimental conditions


Case Temperature [ °C] Pressure [bar] SV [L/kgcat/hr] H2/CO ratio

1 230 20 3000 2
2 230 10 3000 2
3 230 15 3000 2
4 230 25 3000 2
5 210 20 3000 2
6 220 20 3000 2
7 240 20 3000 2
8 230 20 5000 2
9 230 20 7000 2
10 230 20 3000 1.5
11 230 20 3000 1

Table 4 Estimated kinetic parameters in the present study


Kinetic parameters at the reference temperature of 503.15 K
KHad2 3.91 9 10-5 ± (1.62 9 10-6)a bar-1
kCOKad
CO
b
5.82 9 10-2 ± (5.68 9 10-5)a mol/kg/s/bar
b
kIN 3.73 9 10-1 ± (1.62 9 10-1)a mol/kg/s
kCH4 18.9 ± (5.97)a mol/kg/s
kGb 4.33 9 10-1 ± (7.15 9 10-3)a mol/kg/s
kPb2 1.46 ± (4.42 9 10-1)a mol/kg/s
a
kPn 12.1 ± (3.51) mol/kg/s
kOb 2 1.73 9 10-1 ± (3.06 9 10-3)a mol/kg/s
kOb n 8.61 9 10-2 ± (1.16 9 10-3)a mol/kg/s
a
kOrev2 KOadn 16.6 ± (7.49) mol/kg/s/bar
kOrevn KOadn 1.78 ± (1.00 9 10-1)a mol/kg/s/bar
c
Activation energies
(E ? DH)CO -159.8 ± (4.05)a kJ/mol
EG 99.5 ± (2.63)a kJ/mol
EP2 168.0 ± (7.10)a kJ/mol
EO2 70.5 ± (2.93)a kJ/mol
EOn 96.7 ± (3.49)a kJ/mol
a
95% confidence intervals
b
These parameters were assumed to be temperature-independent
h  i
c
ki ¼ ki;ref exp  ERi T1  T1ref ;
where ki,ref kinetic parameter at the reference temperature (Tref = 503.15 K), R gas constant (8.314 J/
mol/K)

the experimental data, Pearson’s product-moment coefficient was calculated for


each experimental condition: values were close to unity (higher than 0.98 for all
cases), indicating that the model coordinated well with the data [61].

123
494 S.-H. Kwack et al.

(a) 1.0
CO conversion

0.8
0.6
0.4
Experiment
0.2
Simulation
0.0
5 10 15 20 25 30 200 210 220 230 240 250 2000 4000 6000 8000 0.5 1.0 1.5 2.0 2.5

(b) 0.4
CH4 selectivity

0.3

0.2

0.1

0.0
5 10 15 20 25 30 200 210 220 230 240 250 2000 4000 6000 8000 0.5 1.0 1.5 2.0 2.5
(c) 0.6
Olefin selectivity

0.4

0.2

0.0
5 10 15 20 25 30 200 210 220 230 240 250 2000 4000 6000 8000 0.5 1.0 1.5 2.0 2.5
Pressure [bar] Temperature [oC] Space velocity [L/kgcat/hr] H2/CO ratio

Fig. 2 Comparisons of a CO conversion and b methane selectivity and c olefin selectivity between
experimental and simulated data. Diagrams have been aligned to show the effects of operating conditions.
Solid circles represent experimental data and empty triangles denote calculated values

Effects of operating conditions

The effects of pressure, temperature, space velocity and H2/CO ratio on CO


conversion, olefin selectivity and the entire distribution of hydrocarbons are shown in
Figs. 2 and 4. Increased pressure resulted in high CO conversion (first diagram,
Fig. 2a) and this feature is in accordance with previous reports [10, 62, 63]. Although
some of experimental data show abnormal trends (e.g. CO conversion with varied
pressure, methane selectivities for SV of 3000 and H2/CO ratio of 1) exceeding the
range of measurement errors, the simulated values show reasonable trends.
CO conversion increased from 35 to 90% as temperature increased from 210 to
240 °C (second diagram, Fig. 2a). Olefin selectivity increased with increasing
temperature (second diagram, Fig. 2c), although the effect of temperature on olefin
selectivity depended on the catalyst [11, 64, 65]. A preference to produce short
rather than long hydrocarbon chains was observed across the entire distribution as
temperature rose from 210 to 240 °C, as the average chain length decreased from 19
to 12 in Fig. 4a, due to the decreased fraction of CH2 intermediates (hCH2 ) in
catalytic surface (cf. diagram (a–1), Fig. 5). Products with decreased carbon
numbers have been reported at increased temperature for cobalt-based catalysts
[62], iron-based catalysts [11, 65, 66], and ruthenium-based catalysts [62].
The simulated temperature gradient in the reactor was lower than 5 °C
(indicating no hot-spots) for all the reaction conditions. This was corroborated by
the methane selectivity in the experimental data not being significantly raised on the
catalysts herein: cobalt-based Fischer–Tropsch reactions under conditions allowing
hot-spot formation generally result in high methane selectivity. Therefore, the

123
Development of a kinetic model of the Fischer–Tropsch 495

-2
Simulation
-4 Experiment
log(Wn/n)

-6
-8
-10 (b) (c)
(a)
-12
-2
-4
log(Wn/n)

-6
-8
-10 (d) (e) (f)
-12
-2
-4
log(Wn/n)

-6
-8
-10 (g) (h) (i)
-12
0 10 20 30 40
-2
Carbon number
-4
log(Wn/n)

-6
-8
-10 (j) (k)
-12
0 10 20 30 40 0 10 20 30 40
Carbon number Carbon number

Fig. 3 Comparisons of entire distribution between experimental (dots) and simulated data (solid lines).
Diagrams a–k correspond to cases 1–11 in Table 3, where case 1 is the reference condition

discussion of the effects of temperature herein is limited to an examination of wall


temperature. However, in the case of large-scale reactor, temperature rise may take
place due to the increased amount of catalysts and wide reactor diameter, and thus,
additional study should be taken into account to observe the temperature profile
inside the reactor.
Changes in space velocity strongly influenced CO conversion as space velocity
corresponds to the inverse of residence time, with CO conversion decreasing with
increasing space velocity (decreasing residence time). However, the change of space
velocity led to little variation of olefin selectivity (third diagram, Fig. 2c) or the
entire distribution (Fig. 4b). The insignificant effect of space velocity on chain
length has been observed experimentally for both Co- and Fe-based catalysts: Darby
and Kemball [67] reported that catalyst bed length had no effect on the chain length
in Co-based FT synthesis and the molecular weights of hydrocarbon products using
iron-based catalysts have been reported not to be affected by changes of space
velocity [68, 69]. Iglesia et al. [70] observed an increase in the average molecular
weight of the products with decreasing space velocity for Ru catalysts. Everson and
Mulder [71] also reported that increased catalyst bed length resulted in increased
chain lengths for Ru-based FT reaction. This feature may infer that significant olefin
reinsertion did not occur over cobalt catalyst in the present study, while olefin
reinsertion seems to be much more facile over Ru than Co based FT catalysts.

123
496 S.-H. Kwack et al.

-2 210 [oC] 3000 [L/kg/hr]


o
-4 220 [ C] 5000 [L/kg/hr]
log(Wn/n)

230 [oC] 7000 [L/kg/hr]


o
-6 240 [ C]

-8
Increasing

-10 (a) (b)


-12
-2 1 10 [bar]
-4 1.5
log(Wn/n)

15 [bar]
2 20 [bar]
-6
25 [bar]
-8
Increasing

-10 (c) (d)


-12
0 10 20 30 40 0 10 20 30 40
Carbon number Carbon number

Fig. 4 Effects of a temperature, b space velocity c H2/CO ratio and d pressure on the entire distribution
of hydrocarbon products (simulated data). Conditions for each diagram are a T = varied (refer to
legends), SV = 3000 L/kgcat/hr, P = 20 bar and H2/CO = 2; b T = 230 °C, SV = varied (refer
to legends), P = 20 bar and H2/CO = 2; c T = 230 °C, SV = 3000 L/kgcat/hr, H2/CO = varied (refer
to legends) and P = 20 bar; d T = 230 °C, SV = 3000 L/kgcat/hr, H2/CO = 2 and P = varied (refer to
legends)

1.0

0.8
2
θCH

H2/CO ratio Pressure


0.6
Temperature increasing increasing
Simulation
(a-1) increaseing (b-1) (c-1) (d-1)
Experiment
0.4
10.0
(a-2) Temperature
7.5 H2/CO ratio
increaseing
3
θ Hx10

5.0 increasing

2.5
(b-2) (c-2) (d-2)
0.0
4
H2/CO ratio
3 increasing
2
Σθ Rkx10

1 Temperature Pressure
(a-3) (b-3) (c-3) increasing
increaseing (d-3)
0
0.0 0.4 0.8 1.2 0.0 0.4 0.8 1.2 0.0 0.4 0.8 1.2 0.0 0.4 0.8 1.2
Packing depth [cm] Packing depth [cm] Packing depth [cm] Packing depth [cm]

Fig. 5 Spatial evolution of catalytic surface coverage by intermediates. Each column of diagrams
corresponds to the change of temperature (first), SV (second), H2/CO ratio (third) and pressure (fourth),
and detailed conditions are referred to the caption of Fig. 4

123
Development of a kinetic model of the Fischer–Tropsch 497

CO conversion increased when the hydrogen to carbon monoxide ratio increased


(fourth diagram, Fig. 2a), and the fraction of olefins was reduced (fourth diagram,
Fig. 2c) due to the increased amount of hydrogen. Donnelly and Satterfield [65]
observed a decrease of olefin-to-paraffin ratio from 6 to 1 as the H2/CO ratio
increased from 0.3 to 4. This can also be explained using the termination reaction
rates of paraffin and olefin products (Table 1), where the increased fraction of
surface hydrogen atoms (hH*) with increasing H2/CO ratio increases the rates of
paraffin formation and olefin re-adsorption. This feature is clearly explained in the
coverage of hydrogen atoms provided in the diagram (c-2) of Fig. 5.
Increasing the H2/CO ratio has been reported to result in lighter hydrocarbon
products for all types of FT catalyst [11, 15, 65], and the simulated results agree with
the reported discussions (Fig. 3c). Liu et al. [66] observed a decrease in product chain
length as H2/CO ratio increased when using an iron catalyst. The reduced average
chain lengths can be explained by the values of cCH2 , which describe the coverage of
monomer units (CH2-surface species) on the active sites of the catalyst.
It is worth noting that the averaged olefin selectivities over the entire chain
lengths are considered in Fig. 2c. The olefin selectivity for each chain length was
observed to decrease with respect to carbon number in our experimental data,
indicating that paraffins are more produced than olefins for higher chain length. This
feature is also well reported in the literature [5, 72].
The reaction rates herein exclude the partial pressure of gaseous water, indicating
that the presence of water molecules has no effect on the Fischer–Tropsch synthesis.
With cobalt-based catalysts, the influence of water on the reaction rate has been
reported to be negligible [8–10], resulting in higher productivity at high synthesis gas
conversion. FT synthesis with iron-based catalysts has been reported to be inhibited
by water [13, 64, 73], but lately there have been findings in the literature that water
has no significant influence on the reaction kinetics over iron-FT catalysts [74, 75].
However, recent publication reports positive kinetic effect of water, at least up to
moderate amounts, on the FT rate while negative effect on the space–time yield at a
direct exposure of Co supported on narrow-pore c-Al2O3 to high partial pressures of
water due to a rapid and extensive deactivation of the small Co particles [76].

Conclusions

A detailed kinetic mechanism, which assumed the CH2 insertion mechanism, was
suggested for cobalt-based Fischer–Tropsch synthesis based on experimental
evidence and theoretical analyses of the literature. Other than the adsorption
reactions, the elementary steps were considered irreversible reactions, and the
reaction rates were developed using the equilibrium constants of the adsorbents and
the quasi steady state assumption for intermediate species on the catalyst’s surface.
Kinetic parameters were estimated by fitting experimental data under a variety of
conditions. The model’s prediction of the entire distribution of hydrocarbon
products coincided with experimentally measured data, satisfactorily supporting its
validity. The model suggested herein was able to evaluate the effects of operating
conditions on the product distribution based on detailed governing reactions.

123
498 S.-H. Kwack et al.

Acknowledgments This work was supported by Korea Institute of Energy Technology Evaluation and
Planning (KETEP) and GTL Technology Development Consortium under ‘‘Energy Efficiency and
Resources Programs’’ (Project No. 2006CCC11P011B) of the Ministry of Knowledge Economy,
Republic of Korea. M.-J. Park would also like to acknowledge that this work was supported by Basic
Science Research Program through the National Research Foundation of Korea (NRF) funded by the
Ministry of Education, Science and Technology (No. 2009-0072198).

Appendix

Details about the development of the kinetics model

Rapid equilibrium was assumed in the adsorption steps, while each surface reaction
was assumed to be an elementary reaction. The fractions of surface hydrogen atom
(hH*), surface carbon monoxide (hCO*) and surface olefins (hOn ) were expressed as
functions of vacant site fraction using the adsorption equilibrium constants for
hydrogen molecules, CO molecules and olefins in the gas phase, respectively:
 1=2  1=2
hH ¼ KHad2 PH2 h ¼ cH h where cH ¼ KHad2 PH2 ð10Þ
ad
hCO ¼ KCO PCO h ð11Þ
hOn ¼ KOadn POn h ð12Þ
To develop expressions for the fractions of the other intermediate species on the
catalyst surface, including hC*, hCH* and hCH2 , the quasi steady state assumption
(QSSA) was applied to the corresponding balances (dhC*/dt, dhCH*/dt and dhCH2 /dt,
respectively) as follows:
ad
kCO hCO h kCO KCO
h C ¼ ¼ PCO h ð13Þ
kC hH k C cH
ad
kC kCO KCO
hCH ¼ h C ¼ PCO h ð14Þ
kCH kCH cH
ad
kCO KCO PCO h2
hCH2 ¼ P1  ð15Þ
kIN cH h þ kG n¼1 hRn


Since Eq. 15 includes the summation of hRn*, dhRn*/dt is summed up over all the
values of n, as follows:
 P1 
d n¼1 hRn

¼kIN hCH2 hH  kG hR1 hCH2  kCH4 hR1 hH
dt
þ kG hR1 hCH2  kG hR2 hCH2  kP2 hR2 hH  kO2 hR2 h þ kOrev2 hO2 hH
þ kG hR2 hCH2  kG hR3 hCH2  kPn hR3 hH  kOn hR3 h þ kOrevn hO3 hH
..
. ð16Þ
Here, all the terms related to the growth reaction are canceled. It is assumed that
the specificity for the formation of methane is not extremely significant, and thus,

123
Development of a kinetic model of the Fischer–Tropsch 499

the termination rate for the formation of methane is replaced with the termination
rate of radical R1 with kinetic parameters for paraffins and olefins termination rate
constants, i.e. kCH4 hR1 hH ffi ðkPn hH þ kOn h ÞhR1 . This assumption is validated by


the calculated values of ðkPn hH þ kOn h ÞhR1  kCH4 hR1 hH kCH4 hR1 hH being less
than 0.05 under most of experimental conditions and the contribution of
kCH4 hR1 hH to the sum of termination rates, defined as kCH4 hR1 hH
 
P P1 P
1
kCH4 hR1 hH þ kPn 1n¼2 h R h
n H þ k On n¼3 h R  h  k rev
n  On h O h 
n H , is less than
n¼3
0.1, indicating that this assumption may cause an error of 0.005 at most. In a similar
manner, kP2 , kO2 , and kOrev2 are assumed to be equal to kPn , kOn , and kOrevn , respectively,
in Eq. 16, although they are estimated separately in the estimation step.
It is worth noting that ethane is well known to follow the ASF product
distribution, differently from ethylene. However, our experimental data showed
that, if we consider the specificity for ethylene only, the deviated fraction of C2
products were unable to be explained due to low olefin selectivity. Therefore, we
assumed the ethane to be deviated from the ASF plot.
By applying the assumptions discussed above to Eq. 16 and making assumption
of quasi steady state, then, inserting Eqs. 10 and 12, the following expression is
obtained:
 P1  ! !
d n¼1 hRn
 X1 X 1
¼k
~ IN hCH2 hH  kPn hRn hH  kOn hRn h
dt n¼1 n¼1
! ð17Þ
X
1
rev
þ kOn hOn hH ¼ 0
n¼2
P1   P1 
X
1
kIN hCH2 hH þ kOrevn n¼2 hOn hH
  kIN hCH2 þ kOrevn KOadn n¼2 POn h
h ¼
Rn ¼ ð18Þ
n¼1
kPn hH þ kOn h kPn þ kOn c1 H

Eq. 18 can be inserted into Eq. 15 to give the following quadratic equation:
b1 h2CH2 þ b2 h hCH2  b3 h2 ¼ 0
 P1 
kG kIN kG kOrevn KOadn n¼2 POn ad
where b1 ¼ ; b2 ¼ kIN cH þ ; b3 ¼ kCO KCO PCO
kPn þ kOn c1
H kPn þ kOn c1 H
ð19Þ
which can be solved to an expression of hCH2 , in terms of h*, as follows:
0 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1
b2 þ b22 þ 4b1 b3
hCH2 ¼ @ Ah ¼ cCH h
2
2b1
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð20Þ
b2 þ b22 þ 4b1 b3
where, cCH2 ¼
2b1
The QSSA is also applied to hR1 , as follows:

123
500 S.-H. Kwack et al.

 
kIN hCH2 hH kIN cCH2 cH
hR1 ¼ ¼ h ¼ c 1 h
kG hCH2 þ kCH4 hH kG cCH2 þ kCH4 cH
ð21Þ
kIN cCH2 cH
where, c1 ¼
kG cCH2 þ kCH4 cH
And the application of QSSA to hR2 and hRn (n C 3) produces the following
expressions:
kG hR1 hCH2 þ kOrev2 hO2 hH kG cCH2 hR1 þ kOrev2 KOadn cH PO2 h
hR2 ¼ ¼ ¼ A2 hR1 þ B2 h
kG hCH2 þ kP2 hH þ kO2 h kG cCH2 þ kP2 cH þ kO2
kG cCH2 kOrev2 KOadn cH PO2
where A2 ¼ ; B2 ¼
kG cCH2 þ kP2 cH þ kO2 kG cCH2 þ kP2 cH þ kO2
ð22Þ
kG hRn1 hCH2 þ kOrevn hOn hH kG cCH2 hRn1 þ kOrevn KOadn cH POn h
hRn ¼ ¼ ¼ An hRn1 þ Bn h
kG hCH2 þ kPn hH þ kOn h kG cCH2 þ kPn cH þ kOn
kG cCH2 kOrevn KOadn cH POn
where An ¼ ; Bn ¼
kG cCH2 þ kPn cH þ kOn kG cCH2 þ kPn cH þ kOn
ð23Þ
Eqs. 22 and 23 are considered in the form Xn ¼ An Xn1 þ Bn h to solve hRn
algebraically:
X 2 ¼ A 2 X1 þ B 2 h
X3 ¼ A3 A2 X1 þ ðA3 B2 þ B3 Þh
..
.
ð24Þ
Xn ¼ An An1    A2 X1 þ ðAn An1    A3 B2 þ An An1    A4 B3 þ    þ Bn Þh
!
Xn
n2 ni
¼ A A2 X1 þ A Bi h 
i¼2

Here, A3 ¼ A4 ¼    ¼ An ¼ A for n C 3, since kPn is fixed, while the values of


Bn depend on chain length due to the presence of POn . Finally, it is assumed that
vacant sites, surface hydrogen atoms, surface monomer units, and living chains on
the surface play important roles in the site balance, and thus, the developed
expressions for hH , hCH2 , and hRn , i.e. Eqs. 10, 20, and 24 are inserted into the site
P
NP
balance, 1 ¼ h þ hH þ hCH2 þ hRn to develop an expression for the fraction of
n¼1
vacant sites:
" ( ! !)#1
X
NP NP X
X NP
k2 ki
h ¼ 1 þ cH þ cCH2 þ 1þ A A2 c 1 þ A Bi ð25Þ
k¼2 k¼2 i¼2

Eq. 25 can be inserted into Eqs. (10), (12), (21) and (24) to obtain the final
expressions for hH , hOn , hCH3 = hR1 , and hRn , respectively, and the formation rate of

123
Development of a kinetic model of the Fischer–Tropsch 501

hydrocarbon products follows by inserting the expressions for the fraction of the
corresponding intermediate species on the catalyst surface, as follows:
rCH4 ¼ kCH4 c1 cH =DEN ð26Þ
rP;2 ¼ kP2 ðA2 c1 þ B2 ÞcH =DEN ð27Þ
( !) ,
X n
n2 ni
rP;n ¼ kPn A A 2 c1 þ A Bi cH DEN n3 ð28Þ
i¼2
n o.
rO;2 ¼ kO2 ðA2 c1 þ B2 Þ  kOrev2 KOadn cH PO2 DEN ð29Þ
" ( !) #,
XNP
rO;n ¼ kOn An2 A2 c1 þ Ani Bi  kOrevn KOadn cH POn DEN n  3 ð30Þ
i¼2
    2
P
NP
k2
P
NP P
k
ki
where DEN ¼ 1 þ cH þ cCH2 þ 1þ A A2 c 1 þ A Bi .
k¼2 k¼2 i¼2

References

1. Iglesia E (1997) Appl Catal A Gen 161:59–78


2. Dry ME (2002) Catal Today 71:227–241
3. Davis BH (2005) Top Catal 32:143–168
4. Khodakov AY, Chu W, Fongarland P (2007) Chem Rev 107:1692–1744
5. Visconti CG, Tronconi E, Lietti L, Zennaro R, Forzatti P (2007) Chem Eng Sci 62:5338–5343
6. Newsome DS (1980) Catal Rev Sci Eng 21:275–318
7. Jager B, Espinoza R (1995) Catal Today 23:17–28
8. Yates IC, Satterfield CN (1991) Energy Fuels 5:168–173
9. Schulz H, Claeys M, Harns S (1997) Stud Surf Sci Catal 107:193–200
10. van Berge PJ, Everson RC (1997) Stud Surf Sci Catal 107:207–212
11. Dictor RA, Bell AT (1986) J Catal 97:121–136
12. Dry ME (1976) Ind Eng Chem Prod Res Dev 15:282–286
13. Huff G Jr, Satterfield CN (1984) Ind Eng Chem Prod Res Dev 23:696–705
14. Anderson RB (1984) The Fischer–Tropsch synthesis. Academic Press, New York
15. van der Laan GP, Beenackers AACM (1999) Catal Rev Sci Eng 41:255–318
16. 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)
Fuel 82:195–213
17. Kim YH, Hwang D-Y, Song SH, Lee SB, Park ED, Park M-J (2009) Korean J Chem Eng
26:1591–1600
18. Lox ES, Froment GF (1993) Ind Eng Chem Res 32:71–82
19. Dry ME (1996) Appl Catal A Gen 138:319–344
20. Gaube J, Klein H-F (2008) J Mol Catal A Chem 283:60–68
21. Brady R, Pettit R (1980) J Am Chem Soc 102:6181–6182
22. Brady R, Pettit R (1981) J Am Chem Soc 103:1287–1289
23. Quyoum R, Berdini V, Turner ML, Long HC, Maitlis PM (1998) J Catal 173:355–365
24. Hindermann JP, Hutchings GJ, Kiennemann A (1993) Catal Rev Sci Eng 35:1–127
25. Cheng J, Gong X-Q, Hu P, Lok CM, Ellis P, French S (2008) J Catal 254:285–295
26. Pichler H, Schulz H (1970) Chem Ing Tech 42:1162–1174
27. Zhuo M, Tan KF, Borgna A, Saeys M (2009) J Phys Chem C 113:8357–8365
28. Gong X-Q, Raval R, Hu P (2005) J Chem Phys 122:024711
29. Ciobı̂ca˛ IM, Frechard F, Jansen APJ, van Santen RA (2001) Stud Surf Sci Catal 133:221–228
30. Liu Z-P, Hu P (2002) J Am Chem Soc 124:11568–11569

123
502 S.-H. Kwack et al.

31. Scorch HH, Goulombic N, Anderson RB (1951) The Fischer–Tropsch and related syntheses. Wiley,
New York
32. Davis BH (2001) Fuel Process Technol 71:157–166
33. Kummer JT, Podgurski HH, Spencer WB, Emmett PH (1951) J Am Chem Soc 73:564–569
34. Kummer JT, Emmett PH (1953) J Am Chem Soc 75:5177–5183
35. Hall WK, Kokes RJ, Emmett PH (1957) J Am Chem Soc 79:2983–2989
36. Hall WK, Kokes RJ, Emmett PH (1960) J Am Chem Soc 82:1027–1037
37. Blyholder G, Emmett PH (1959) J Phys Chem 63:962–965
38. Blyholder G, Emmett PH (1960) J Phys Chem 64:470–472
39. Raje A, Davis BH (1996) Catalysis vol 12. The Royal Society of Chemistry, Cambridge
40. Ojeda M, Nabar R, Nilekar AU, Ishikawa A, Mavrikakis M, Iglesia E (2010) J Catal 272:287–297
41. Kwack S-H, Bae JW, Park M-J, Kim S-M, Ha K-S, Jun K-W (2011) Fuel 90:1383–1394
42. Bae JW, Kim S-M, Park S-J, Lee Y-J, Ha K-S, Jun K-W (2010) Catal Commun 11:834–838
43. Ponec V, van Barneveld WA (1979) Ind Eng Chem Prod Res Dev 18:268–271
44. Rofer-DePoorter CK (1981) Chem Rev 81:447–474
45. Ernst KH, Schwarz E, Christmann K (1994) J Chem Phys 101:5388–5401
46. Kummer JT, DeWitt TW, Emmett PH (1948) J Am Chem Soc 70:3632–3643
47. van Barneveld WAA, Ponec V (1984) J Catal 88:382–387
48. van Steen E, Schulz H (1999) Appl Catal A Gen 186:309–320
49. Bell AT (1980) Catal Rev Sci Eng 23:203–232
50. Kellner CS, Bell AT (1981) J Catal 70:418–432
51. Hovi J-P, Lahtinen J, Liu ZS, Nieminen RM (1995) J Chem Phys 102:7674–7682
52. Joyner RW (1988) Catal Lett 1:307–310
53. Schulz H, Beck K, Erich E (1988) Fuel Proc Technol 18:293–304
54. Schulz H, van Steen E, Claeys M (1993) Selective hydrogenation and dehydrogenation. DGMK,
Kassel
55. Henrici-Olive G, Olive S (1984) The chemistry of the catalyzed hydrogenation of carbon monoxide.
Springer-Verlag, Berlin
56. Eidus YT (1967) Russ Chem Rev 36:338–351
57. Kibby CL, Pannell RB, Kobylinski TP (1984) ACS Division of Petroleum Preprints 29:1113–1119
58. Fogler HS (1999) Elements of chemical reaction engineering. Prentice-Hall, New Jersey
59. Boudart M (1972) AIChE J 18:465–478
60. Lim H-W, Park M-J, Kang S-H, Chae H-J, Bae JW, Jun KW (2009) Ind Eng Chem Res
48:10448–10455
61. Rodgers JL, Nicewander WA (1988) Am Stat 42:59–66
62. Dry ME (1981) Catalysis-science and technology. Springer-Verlag, New York
63. Yan Z, Wang Z, Bukur DB, Goodman DW (2009) J Catal 268:196–200
64. Anderson RB (1956) Catalysts for the Fischer–Tropsch synthesis. Van Nostrand Reinhold, New York
65. Donnelly TJ, Satterfield CN (1989) Appl Catal 52:93–114
66. Liu Y, Teng BT, Guo X-H, Li Y, Chang J, Tian L, Hao X, Wang Y, Xiang H-W, Xu Y-Y, Li Y-W
(2007) J Mol Catal A Chem 272:182–190
67. Darby PW, Kemball C (1959) Trans Faraday Soc 55:833–841
68. Bukur DB, Patel SA, Lang X (1990) Appl Catal 61:329–349
69. Feimer JL, Silveston PL, Hudgins RR (1981) Ind Eng Chem Prod Res Dev 20:609–615
70. Iglesia E, Reyes SC, Madon RJ (1991) J Catal 129:238–256
71. Everson RC, Mulder H (1993) J Catal 143:166–174
72. Yang J, Liu Y, Chang J, Wang Y-N, Bai L, Xu Y-Y, Xiang H-W, Li Y-W, Zhong B (2003) Ind Eng
Chem Res 42:5066–5090
73. Vannice MA (1976) Catal Rev Sci Eng 14:153–191
74. Botes FG (2008) Catal Rev Sci Eng 50:471–491
75. Zhou L-P, Hao X, Gao J-H, Yang Y, Wu B-S, Xu J, Xu Y-Y, Li Y-W (2011) Energy Fuels 25:52–59
76. Lögdberg S, Boutonnet M, Walmsley JC, Järås S, Holmen A, Blekkan EA (2011) Appl Catal A Gen
393:109–121

123

Potrebbero piacerti anche