Detailed kinetics of the Fischer-Tropsch synthesis over Co-based catalysts containing sulphur

Article abstract of DOI:10.1016/j.cattod.2010.04.007

The effect of sulphur poisoning on the activity and selectivity in the Fischer-Tropsch synthesis of a Co/γ-Al₂O₃ catalyst, investigated in a previous work under conditions relevant to industrial operation, has been described in this work by means of a complete kinetic model, able to predict at the same time the CO conversion and the product distribution (paraffins and olefins from C₁ to C₅₀) on the basis of the process conditions and the amount of sulphur on the catalyst. To reach this goal, a mechanistic kinetic model, developed in a previous work for an unpoisoned catalyst, has been modified introducing the dependence on sulphur of the total number of catalytic active sites (to describe the progressive decrease of the catalyst activity with increasing sulphur loading on the catalyst), of the rate of the chain growth step (to describe the decrease of the selectivity to heavy products observed with sulphur-poisoned catalysts), and of the rates of paraffins and olefins formation from the growing adsorbed intermediate (to describe the variation of the olefin/paraffin ratio in the reaction products collected with the poisoned catalysts).

Full text of DOI:10.1016/j.cattod.2010.04.007

Catalysis Today 154 (2010) 202–209
Contents lists available at ScienceDirect
Catalysis Today
journal homepage: www.elsevier.com/locate/cattod
Detailed kinetics of the Fischer–Tropsch synthesis over Co-based
catalysts containing sulphur
Carlo Giorgio Viscontia,1, Luca Liettia,∗, Enrico Tronconia, Pio Forzattia,
Roberto Zennarob, Stefano Rossinic
a Politecnico di Milano, Dipartimento di Energia, Piazza Leonardo da Vinci, 32- 20133 Milano, Italy
b Eni, Divisione Exploration & Production, Via Emilia, 1 - 20097 San Donato Milanese, Italy
c Eni, Divisione Refining & Marketing, Via Felice Maritano, 26 - 20097 San Donato Milanese, Italy
a r t i c l e i n f o
a b s t r a c t
Article history:
Available online 14 May 2010
The effect of sulphur poisoning on the activity and selectivity in the Fischer–Tropsch synthesis of a Co/␥-
Al2O3 catalyst, investigated in a previous work under conditions relevant to industrial operation, has been
described in this work by means of a complete kinetic model, able to predict at the same time the CO
conversion and the product distribution (paraffins and olefins from C1 to C50) on the basis of the process
conditions and the amount of sulphur on the catalyst. To reach this goal, a mechanistic kinetic model,
developed in a previous work for an unpoisoned catalyst, has been modified introducing the dependence
on sulphur of the total number of catalytic active sites (to describe the progressive decrease of the catalyst
activity with increasing sulphur loading on the catalyst), of the rate of the chain growth step (to describe
the decrease of the selectivity to heavy products observed with sulphur-poisoned catalysts), and of the
rates of paraffins and olefins formation from the growing adsorbed intermediate (to describe the variation
of the olefin/paraffin ratio in the reaction products collected with the poisoned catalysts).
© 2010 Elsevier B.V. All rights reserved.
Keywords:
Fischer–Tropsch synthesis
Catalyst deactivation
Sulphur poisoning
Kinetics
Catalyst selectivity
Cobalt
Mathematical modeling
CO conversion
Product distribution
1. Introduction
heavy paraffins (waxes) are the preferred products, since they can
be easily hydrocracked to the so-called middle distillates, which
The Fischer–Tropsch synthesis (FTS) is a surface catalyzed pro-
cess discovered in 1923, in which the synthesis gas, the mixture
of carbon monoxide and hydrogen, is converted into water, other
oxygenates and hydrocarbons with a broad range of chain length
and functionality [1]. In last decade, this process has received con-
siderable attention by both the industrial and the academic worlds
as a way of exploiting coal, biomasses or the huge and cheap nat-
ural gas reserves located in remote areas, leading to high-grade
synthetic lubricating oils and high-added value fuels (diesel with
an high cetane number, sulphur- and aromatic-free) [2,3].
The FTS can be schematically described as polymerizations pro-
cess, in which CO and H2 react to form a monomer, which then
grows to form the reaction products according to the following
lumped stoichiometry:
represent the FT products with the highest added value.
Due to their cost, activity and selectivity, cobalt and iron are the
most widely studied Fischer–Tropsch catalysts, and the only ones
to be used at the industrial scale. The other metals active in the
synthesis, namely nickel and ruthenium, have been used until now
only as catalyst for research purposes or as promoters of cobalt and
iron-based catalysts [4].
Like for many other catalysts, it is known that S-compounds
contained in coal, biomasses or natural gas are effective poisons
for Fischer–Tropsch catalysts [4–10]. It has been suggested that in
order to minimize the deactivation of both iron and cobalt-based
industrial Fischer–Tropsch catalysts, the sulphur content of syngas
should be kept lower than 0.02 mg/m3 [4]. Besides, alkali promoters
are often added in order to limit the catalyst deactivation by sulphur
[11].
In a previous work [5], we reported the results of an experi-
mental investigation on the reactivity of a bench-scale Co/Al2O3
un-promoted catalyst, pre-poisoned with different amounts of sul-
phur (in the range 0–2000 ppmw) according to an ex situ procedure
(i.e. by impregnation) to overcome the experimental complexity
associated to the use of S-carrier molecules. In this way, the poi-
son concentration was a priori known and constant with the time
on stream and the S-concentration gradients in the catalyst bed,
nCO + 2nH2 → −(CH2)n + nH2O
(1)
The synthesis of the reaction products is strongly exothermic,
releasing 165 kJ per each mole of converted CO. In general, the
∗
Corresponding author. Tel.: +39 02 2399 3272; fax: +39 02 2399 3318.
E-mail address: luca.lietti@polimi.it (L. Lietti).
http://www.lccp.polimi.it.
1
0920-5861/$ – see front matter © 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.cattod.2010.04.007

C.G. Visconti et al. / Catalysis Today 154 (2010) 202–209
203
Table 1
Experimental CO conversion values for samples loaded with different amounts of
sulphur (experimental conditions as in Fig. 1). Data are taken from Ref. [5].
Catalyst sample
Sulphur loading (ppmw
)
CO conversion (%)
S0
S10
S100
S250
S2000
0
10
100
250
2000
24.0
18.8
15.3
9.4
3.0
typical of fixed-bed reactors fed with a S-containing feed, could be
safely neglected.
The feasibility of studying S-poisoning adopting a catalyst pre-
poisoning procedure (ex situ poisoning) has been discussed in our
previous work [5], detailing the possible differences between the
ex situ and the in situ poisoning by S-compounds during the FTS.
It has been shown that in situ poisoning with S-carrier molecules
suffers of two main drawbacks: first, the low concentration which
is required in the feed stream, associated with the high adsorp-
tion capability of S-carrier molecules (typically H2S) in the rig lines,
makes very difficult to precisely know the amount of sulphur reach-
ing the catalyst samples; second, the high adsorption capability
of H2S on the catalyst may lead to the presence of intraparticle
sulphur gradients in the catalyst pellets and, in the case of fixed-
bed reactors, to the presence of axial S-concentration gradients in
the catalyst bed, thus making very complex the kinetic analysis of
experimental data obtained in the presence of sulphur.
An obvious issue of the pre-poisoning method is that the S-
catalyst interactions may be different with that established under
actual reaction conditions. The strategy followed to limit this draw-
back is discussed in [5]. Particular attention was given to the
poisoning process that has been carried out on a pre-reduced
catalyst to simulate the S–Co interactions under actual synthesis
conditions.
From a morphological point of view, we observed the added
sulphur does not lead to appreciable variations in the catalyst
characteristics; on the contrary, sulphur decreased the catalyst
reducibility by hydrogen, preventing the formation of the active
phase in the FT process, namely the metallic cobalt.
Concerning the FT catalytic performances, the comparison
between the product yields of samples loaded with different
amounts of sulphur pointed out that the presence of this atom on
the catalyst surface remarkably affects both the activity and the
selectivity of the reaction.
Fig. 1. Process selectivity in terms of (a) total hydrocarbons, (b) n-paraffins and (c)
␣-olefins as a function of carbon atom number over the various sulphured sam-
ples (experimental conditions: catalyst load 3 g, diluted 1:2 (v/v) with ␣-Al2O3;
Tcat = 493 K; P = 20 bar; H2/CO inlet molar ratio = 2; GHSV = 5000 cm3(STP)/h/gcat, sul-
phur amount on the catalyst = 0–2000 ppmw).
ical model able to predict the catalyst activity loss as a function of
the sulphur content on the surface. A literature power-law expres-
sion (rCO = kP0.74PC−O0.24) [12] was used to describe the activity of
The CO conversion data obtained running catalysts pre-
loaded with different amounts of sulphur (0, 10, 100, 250
)
at the same process conditions (220 ◦C,
H
2
and 2000 ppmw
the unpoisoned catalyst, while the CO conversion rates in the case
of the sulphur-poisoned samples were described by introducing in
the expression of the kinetic constant k a term depending on the
sulphur content on the catalyst. It was found that the experimen-
tal results are nicely described by adopting a relationship with the
following structure:
20 bar, 2 molH /molCO, 5000 cm3(STP)CO+H2 /h/gcat) are reported
2
in Table 1, while the selectivity to the different reaction products
is reported in Fig. 1 in terms of total hydrocarbons (a), n-paraffins
(b) and ␣-olefins (c).
The addition of sulphur determines a progressive decrease of
the CO conversion, so that the activity of heavier sulphured cat-
alyst (2000 ppm) is almost completely compromised. Moreover,
sulphur significantly changes the product distribution, decreasing
the formation of heavier species (both C5+ and C25+), thus driving
the selectivity of the reaction towards light products, and changing
the olefinicity of the products.
Notably, we found that the effects of sulphur very with the
sulphur loading: indeed for low S amounts (<100 ppm) the CO
conversion is decreased whereas the product distribution is not
significantly affected; at higher S loadings a significant decrease in
the formation of heavier products is observed along with a further
decrease in CO conversion.
k0
k =
(2)
(1 + ˛S)m
where k0 is the kinetic constant of the unpoisoned catalyst sample,
S the sulphur content on the catalyst expressed in mgS/kgcat and ˛,
m two positive adaptive parameters estimated upon regression of
the experimental data.
Such a model is extremely simple and for this reason can be eas-
ily applied to estimate the total productivity of lab- and bigger-scale
reactors working with cobalt-based catalysts, fresh or partially poi-
soned with sulphur. However, it does not provide any information
concerning the products distribution, which is well known to be a
crucial aspect for the industrialization of the Fischer–Tropsch syn-
The experimental data of CO conversion decay with S-loading
were also kinetically described in [5] in order to find a mathemat-

204
C.G. Visconti et al. / Catalysis Today 154 (2010) 202–209
thesis and strongly depends on the amount of sulphur loaded onto
the catalyst [5].
where Pn and On are the generic linear paraffin and ␣-olefin
with n carbon atoms, respectively, and Rn* is the generic grow-
ing adsorbed linear hydrocarbon species (with the generic formula
*CH2(CH2)n − 1H, with n = 1 → NP).
The detailed kinetic description of the FTS in the presence of
sulphided catalyst is an important feature for both the industrial
practice and the comprehension of the sulphur effects on the reac-
tion mechanism. In particular, the development of a kinetic model
able to predict at the same time the reactants conversion and the
products distribution (i.e. a detailed kinetic model) either in the
case of fresh or partially poisoned catalyst is a prerequisite for the
design, optimization, and simulation of the industrial processes fed
with sulphur containing syngas. However, such a model is still lack-
ing in the open literature to our knowledge. In order to fulfil this
demand, in the present work, the mathematical description of the
data reported in [5] has been extended also to products selectivity.
A complete kinetic mechanistic model for the Fischer–Tropsch
synthesis over a Co/Al2O3 unpoisoned catalyst, able to predict
simultaneously both the CO conversion and the n-paraffins and
␣-olefins selectivity from C1 to C50 on the basis of the process con-
ditions, had been reported by some of us in [13–15]. A fundamental
approach was followed to develop this model. First, a detailed FTS
mechanism for a cobalt-based catalyst was defined, explaining the
synthesis of each product through the evolution of reaction inter-
mediates and adsorbed species. Then, appropriate rate laws were
attributed to each elementary step. Finally, the resulting kinetic
scheme was fitted to a comprehensive set of FTS runs.
In the proposed mechanism, gas phase H2 dissociates reversibly
on two different free catalytic sites (*) (step 1), while gas phase CO
is first adsorbed reversibly in the molecular state (step 2), then dis-
sociates (step 3). Also the formation of the monomeric species CH2*
(in accordance with the carbide theory [16,17]) occurs via two steps
in series – the reaction between the carbide and the surface hydro-
gen to form the species CH* (step 4) and the reaction between this
species and the surface hydrogen to form the monomer CH2* (step
5) – and the same happens to the adsorbed oxygen atom formed
from the CO* dissociation (step 3), that is first hydrogenated to OH*
(step 6) and then removed as water (step 7).
In accordance with the alkyl theory [18], the chain growth reac-
tion is initiated by the formation of a methyl species (step 8) and the
propagation takes places by the successive insertion of methylene
into the active site-alkyl bond (step 9). Termination is the result of
two different routes: a dual-site reaction between the intermedi-
ate Rn* and an adsorbed hydrogen atom (steps 10 and 11) giving
n-paraffins or a reversible ␤-hydride elimination reaction (steps 12
and 13) giving ␣-olefins.
2.2. The kinetic model
In the present work we have extended this model to make it
suitable to describe also the performances of partially deactivated
catalysts. The developed model allows to predict simultaneously
both the CO conversion and the n-paraffins and ␣-olefins selectivity
up to n = 50, on the basis of the process conditions and the sulphur
loading on the catalyst surface.
An elementary rate law or an equilibrium constant has been
assigned to each step involved in the detailed reaction mechanism
of FTS. Accordingly, the FT kinetics has been detailed as follows
[13–15]:
Step 1 : Keq = ϑH2
P
−1ϑ−2
(16)
∗
H
2
H
2
2. Kinetic modeling
Step 2 : Keq = ϑCO PC−O1ϑ−1
(17)
(18)
(19)
(20)
(21)
(22)
∗
CO
In the following, the kinetic model developed in [13–15] to
describe the activity of the unpoisoned catalyst, and modified in
this work to introduce the effects of sulphur poisoning on the per-
formances of the adopted catalyst, will be briefly recalled.
∗
Steps 3 − 7 : rM = kMϑCO
ϑ
∗
∗
Step 8 : rIN = kINϑCH ϑH
2
∗
∗
Step 9 : rG,n = kGϑR ϑCH
n
2
2.1. The reaction scheme
∗
∗
Step 10 : rCH = kCH ϑCH ϑH
4
4
3
∗
∗
The unassisted CO dissociation version of the carbide theory
[16,17] has been adopted to describe CO conversion, while the alkyl
mechanism [18] has been adopted to describe the chain growth
process. Accordingly, the FT reaction pattern has been detailed as
follows [13–15]:
Step 11 : rP,n = kPϑR ϑH
n
៭
ꢀ
∗
5
∗
Step 12 : rC
= rO,2 − rC
= kO,dxϑC
− kC 4 xC 4 ϑH (23)
H
H
H
H
2
H
2
4
2
4
2
2
៭
ꢀ
∗
∗
Step 13 : rO,n = rO,n − rO,n = kO,dxϑR − kO,sxxO ϑH
(24)
n
n
where ϑ is the fraction of free catalytic sites, ϑi is the fraction of the
Step1 : H2 + 2∗ ꢀ 2H∗
Step2 : CO + ∗ ꢀ CO∗
Step3 : CO ∗ + ∗ → C ∗ + O∗
Step4 : C ∗ + H∗ → CH ∗ + ∗
Step5 : CH ∗ + H∗ → CH2∗ + ∗
Step6 : O ∗ + H∗ → OH ∗ + ∗
Step7 : OH ∗ + H∗ → H2O + 2∗
Step8 : CH2∗ + H∗ → CH3∗ + ∗
Step9 : Rn∗ + CH2∗ → Rn + 1∗ + ∗, n: 1 → NP-1
Step10 : CH3∗ + H∗ → CH4 + 2∗
Step11 : Rn∗ + H∗ → Pn + 2∗, n: 2 → NP
Step12 : C2H5∗ ꢀ C2H4 + H∗
(3)
(4)
catalytic sites occupied by species i and xO is the molar fraction of
n
the ␣-olefin with n carbon atoms in the liquid phase surrounding
the catalyst pellets.
(5)
Both the H2 dissociative adsorption (step 1) and the CO molecu-
lar adsorption (step 2) have been assumed to approach equilibrium
at the actual process conditions [14,15]. For this reason two equi-
librium constants (Eqs. (16) and (17)) have been assigned to these
steps. In contrast, for steps 3–7 we have assumed that CO* dissoci-
ation (step 3) is rate determining in these sequence of consecutive,
non-reversible and kinetically controlled steps that lead to the for-
mation of methylene species and water [14,15]. Accordingly, the
overall rate of all such steps has been described by the rate expres-
sion for CO dissociation (Eq. (18)).
It has been also assumed that the rate constants describing the
elementary steps for the growth of the adsorbed species Rn* (step
9), the formation of the paraffins C2+ (step 11) and the formation
of the olefins C3+ (step 13) are independent of the carbon atom
number of the intermediates involved in the elementary reactions
[13–15].
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
(14)
(15)
Step13 : Rn∗ ꢀ On + H∗, n: 3 → NP

C.G. Visconti et al. / Catalysis Today 154 (2010) 202–209
205
In order to describe the experimental deviation of methane and
ethylene from Anderson–Schulz–Flory (ASF) product distribution,
specificity has been assumed in the kinetic constants involved in
the kinetic expressions for the formation of these species (steps 10
and 12) [13–15].
On the basis of the experimental observation that the reaction
temperature strongly affects CO conversion and the hydrogenation
reactions (and hence the olefin/paraffin ratio) while the effect on
the chain growth probability is weak, only rate expressions (18),
(21) and (22) have been considered as activated [14,15]. In this way,
CO and H2 heats of adsorption are lumped in the activation energy
calculated for steps 3–7, 10 and 11. According to the Arrhenius law,
the kinetic constants involved in the temperature-activated steps
have been written as:
ꢀ
ꢁ
EM
RT
kM = kM0 exp
−
(25)
(26)
(27)
ꢂ
ꢃ
ECH
RT
4
kCH = kC0H exp
−
4
4
ꢀ
ꢁ
EP
kP = kP0 exp
−
RT
Fig. 2. Experimental olefin to paraffin ratio (O/P) as a function of the number of car-
bon atoms in the products (n) for samples loaded with different amounts of sulphur
(experimental conditions as in Fig. 1).
where ki0 is the pre-exponential factor and Ei the activation energy
referred to the kinetic constant ki.
The molar fraction of the ␣-olefin with n carbon atoms in the liq-
uid phase surrounding the catalyst pellets (xOn ) has been evaluated
by means of an explicit correlation, expressing dependence of this
parameter on temperature, number of carbon atoms and partial
pressure of the corresponding olefin in the gas phase, obtained on
the basis of a priori vapor–liquid equilibrium calculation using the
Soave–Redlich–Kwong cubic equation of state. This introduces an
implicit dependence of the chain growth probability on the chain
length, resulting from higher solubilities of olefins with higher car-
bon atom numbers.
carbon atoms (carbon productivity) and the total carbon productiv-
ity of the products, Fig. 1 shows that the presence of sulphur on the
catalyst surface induces a strong decrease of the selectivity to the
high molecular weight products both in the case of paraffins and
in the case of olefins. In our previous work [5], this effect was ten-
tatively ascribed to the ability of sulphur to poison the (ensemble)
sites required to catalyse the chain growth process (step 9). Due to
the polymerization nature of the Fischer–Tropsch synthesis, in fact,
the rate of the chain growth process (propagation step) is directly
responsible for the average molecular weight of the reaction prod-
ucts: upon increasing the chain growth process the selectivity to
heavy products is increased and vice versa.
2.3. The effect of sulphur on the FT kinetics
Starting from the experimental observation that sulphur load-
ings lower than 100 ppmw do not significantly affect the selectivity
of the FTS, while the selectivity of the process is rapidly shifted
to low molecular weight products at higher sulphur contents [5],
for the catalyst samples loaded with sulphur amounts higher than
100 ppmw, we have introduced the dependence on sulphur content
in the kinetic expression associated with the chain growth reactions
(Eq. (20)). In particular, the rate rG was expressed as:
From the data reported in Table 1 in terms of CO conversion,
early discussed in Section 1, it is evident that the activity of the cat-
alyst is greatly decreased by sulphur and that this effect is strong
already at very low poison loadings. As a matter of facts, while the
residual activity of the sample loaded with 10 ppmw of sulphur is
less than 80% of the activity of the unpoisoned catalysts, the activity
of the most sulphided catalyst we tested, loaded with 2000 ppmw
,
is almost nil. This effect was explained in [5] considering the ability
of sulphur both to prevent the complete reduction of cobalt oxides
during the catalyst activation treatment, thus inhibiting the forma-
tion of the active sites, and to poison selectively part of the active
sites available on the catalyst surface.
In order to fit the CO conversion decrease upon increasing
the sulphur content on the catalyst, the number of available cat-
alytic sites (nacs) on the surface of the poisoned catalysts has been
described in the kinetic model according to the following equation:
rG,n = r0
if S ≤ 100 ppmw
(29)
G,n
rG0,n
rG,n
=
if S > 100 ppmw
(30)
(1 + ˛G(S − 100))m
G
where rG0,n is the chain growth rate of the unpoisoned catalyst and
˛G, mG two positive adaptive parameters.
In Fig. 2, the experimental values of the olefin to paraffin ratio in
the products are reported both in the case of unpoisoned and in the
case of poisoned catalyst samples. It is evident that the presence of
sulphur decreases this ratio, in particular for the most sulphided
catalyst we tested. In order to describe this effect, the rates of
both paraffins and olefins formation (Eqs. (11) and (13), respec-
tively) have been expressed as a function of sulphur amount on the
catalysts, according to the following expressions:
nacs0
nacs =
(28)
(1 + ˛sitesS)m
sites
where nacs0 is the number of available catalytic sites on the unpoi-
soned catalyst (named S0 in [5]), S the sulphur content on the
catalyst expressed in mgS/kgcat and ␣sites, msites two positive adap-
tive parameters. The structure of Eq. (28) has been chosen in order
to reflect that of Eq. (2), that is the best deactivation model found in
[5] to describe the CO conversion trend upon changing the sulphur
content on the catalyst.
rP,n = rP0,n(1 + ˛P,nS)
(31)
(32)
0
0
៭
ꢀ
rO,n = rO,n − rO,n(1 + ˛O,nS)
0
0
0
៭
ꢀ
In terms of process selectivity, calculated as the ratio between
the molar productivity of each species multiplied by its number of
where rP,n, rO,n and rO,n are the rates of paraffins formation,
olefins desorption and olefins readsorption for the unpoisoned cat-

206
C.G. Visconti et al. / Catalysis Today 154 (2010) 202–209
Table 2
alyst, respectively, and ˛P,n, ˛O,n two positive adaptive parameters.
The structure of Eq. (31) has been chosen to express the
enhanced hydrogenating activity of the poisoned catalysts, which
in our opinion is one of the reasons of the decreased olefin/paraffin
ratio in the products collected during runs over the poisoned cata-
lysts.
Estimates of the 8 adaptive parameters involved in the deactivation expressions
used for the description of the reactivity of the poisoned catalyst samples.
Parameter
Estimate
Unit
˛
sites
9.0 × 10−2
2.5 × 10−1
6.0 × 10−1
3.0 × 10−1
6.0 × 10−4
6.0 × 10−4
1.5 × 10−1
2.5 × 10−1
ppm−1
–
msites
˛
ppm−1
–
G
The introduction of an effect of sulphur in the olefin readsorp-
mG
៭
tion rate (rO,n, Eq. (32)), on the contrary, has to be considered an
˛
ppm−1
ppm−1
ppm−1
–
P,n
indirect way to correct the explicit correlation we used to compute
the molar fraction of each olefin in the liquid phase surrounding
the catalyst pellets at the reaction conditions (see Eq. (24)). This
correlation, in fact, was developed considering the typical com-
position of the liquid phase trickling in the FT fixed-bed reactors,
working with unpoisoned catalysts; however, in the presence of
sulphur-poisoned catalysts, due to the significant decrease of the
chain growth probability, we expect to have a liquid phase lighter
than usual, in which the solubility of the (light) olefins, and thus
their readsorption rate, is higher.
˛
O,n
˛
O,2
mO,2
2.5. Optimization method
In order to obtain the best fit of the 47 experimental data-set
collected over the unpoisoned catalyst (46 experiments described
in Refs. [13–15], 1 experiment described in Ref. [5]) we adopted
Along similar lines, considering the anomalous behaviour of
ethylene, already discussed in Section 2.2, a specific law has been
used to describe the effect of sulphur on the readsorption rate of
this species:
13 kinetic/thermodynamic adaptive parameters (Keq , Keq , kM
,
H
CO
2
kIN, kG, kP, kCH , kO,dx, kO,sx, kC 4 , EM, EP, ECH4 ). Other 8 empiri-
H
cal adaptive pa4rameters (˛sites,2msites, ˛G, mG, ˛P , ˛O,2, mO,2, ˛O,n
)
n
were added to describe the effect of sulphur on the performances
of the adopted catalyst, experimentally investigated at steady state
conditions with catalysts loaded with 5 different sulphur amounts
[5].
The 13 kinetic/thermodynamic parameters referred to the
unpoisoned catalyst were estimated in [13–15] by a non-linear
multi-response regression, performed using the Fortran subroutine
BURENL [19] based on the least-squares method. The algebraic-
differential system constituted by Eqs. (33), (35) and (36) was
integrated numerically with the Fortran subroutine LSODI [20],
that allows to solve stiff problems using Gear’s implicit integration
method with variable step. The regression was performed using
as experimental responses the CO conversion and the CH4, C2H4,
0
0
2
O,2
៭
ꢀ
rC
= rO,2 − rC
(1 + ˛O,2S)m
H
H
2
4
4
0
៭
where rC
is the rate of ethylene readsorption for the unpoi-
H
2
soned catalyst4, and ˛O,2, mO,2 two positive adaptive parameters.
2.4. The reactor model
The reactor model adopted for describing the lab-scale fixed-
bed reactor used to collect the activity data of both the unpoisoned
[13–15] and the poisoned [5] catalysts is an isothermal heteroge-
neous plug-flow model. It is composed by:
C2H6, C3H6, C3H8, C6H12, C6H14, C10H20, C10H22, C12H26, C15H32
,
•
2NP + 2 ordinary differential equations (33), expressing the mate-
rial balances for both the reactants (CO, H2) and the products
C20H42, C25H52, C30H62, C35H72, C45H92, C5+ and olefins selectivity.
The 8 empirical parameters referred to the unpoisoned cata-
lyst were separately estimated in this work, following the same
approach.
(H2O, linear paraffins C1 → CNP, olefins C2 → CNP) of the reaction;
•
2NP + 2 initial condition (34) for the differential equations (33),
assigning the value of flow of each species at the reactor inlet
(Wcat = 0);
•
NP + 3 algebraic equations of the type (35), expressing the mate-
2.6. Model fit
rial balances for the adsorbed species (H*, CO*, CH2*, R∗n n:1→NP);
•
the catalytic sites balance (36).
In our previous works [14,15], the ability of the developed model
to describe the effect of the process conditions on both the CO con-
version and the products distribution was shown in the case of
the unpoisoned catalyst. In this work, those results, which are still
valid for the unpoisoned catalyst, will not be reported anymore;
on the contrary, the ability of the model to describe the sulphur
effects on both the catalyst activity and selectivity will be shown
and discussed.
The estimates of the 8 adaptive parameters used to describe the
effects of sulphur on the reactivity of the adopted catalytic sys-
tem and estimated upon regression of the experimental data-set
described in Ref. [5] are listed in Table 2. The presence of sulphur
decreases both the number of the active sites and their activity in
the chain growth process. In the case of the active sites number, this
effect is very strong already at very low sulphur poisoning, so that
the number of residual active sites is about 85% at 10 ppmw and is
more then halved at 170 ppmw. In the case of the chain growth rate,
the poisoning effect is even stronger than in the case of the loss of
activity, but the deactivation is “delayed” and begins, as discussed
NR
ꢄ
dFi
=
˛i,krk
(33)
(34)
(35)
dWcat
k=1
Wcat = 0 Fi = Fi,0
NR
ꢄ
0 =
˛
j,krk
k=1
NP
ꢄ
∗
∗
∗
∗
ÂR
n
1 = Â + ÂH + ÂCO + ÂCH
+
(36)
2
n=1
where Fi and Fi,0 are the molar flows of the generic species i
(i = CO, H2, H2O, CH4, Pnn:2→NP, C2H4, Onn:3→NP) along the reactor axis
and at the reactor inlet, respectively, Wcat the catalyst mass, ˛i,k and
˛
the stoichiometric coefficients for the ith and for the jth (j = H*,
j,k
CO*, CH2*, R∗nn:1→NP) species, respectively, in the kth reaction, rk
the rate of the kth reaction and NR the number of the elementary
steps (or group of elementary steps if described with the same rate
expression) involved in the process.
in Section 2.3, for sulphur loadings greater than 100 ppmw
.
On the contrary, sulphur poisoning increases linearly the rate
of paraffins formation, so that in the case of the sample loaded
with 2000 ppmw the rate of formation of these species from the

C.G. Visconti et al. / Catalysis Today 154 (2010) 202–209
207
growing intermediate Rn* is more than doubled with respect to the
unpoisoned sample.
Also, as expected, the presence of sulphur increases the rate
of olefins readsorption, as a consequence of their higher solubil-
ity in the liquid phase surrounding the catalyst pellets packed in
the reactor. This effect is particularly evident in the case of ethy-
lene, possibly as a consequence of the fact that this is the olefin with
the lowest solubility and the higher diffusivity in the waxes.
The ability of the model to describe the CO conversion trend at
fixed process conditions, upon changing the sulphur loading on the
catalyst, is shown in Fig. 3. The model can satisfactorily predict all
the collected experimental data: the average relative error is 16%,
a satisfactory value which is comparable to those obtained in our
previous works with the unpoisoned catalyst [14,15].
In Fig. 4(a)–(d), the experimental and calculated selectivities to
the main reaction products are reported. In the case of methane
(Fig. 4(a)) and ethylene (Fig. 4(b)), the model nicely fits the increase
of the selectivity with increasing the sulphur loading. The same
happens for the decreasing trend shown by the selectivity to C5+
hydrocarbons (Fig. 4(c)). It is worth noticing that the model cor-
rectly describes both the unchanged selectivity to these species
for sulphur loadings up to 100 ppm and the growing/decreasing
trends for higher poison contents. The olefin selectivity is shown in
Fig. 4(d) as a function of the sulphur content on the catalyst. Again,
the model correctly fits the experimental data, accounting for
the high selectivity to unsaturated hydrocarbons experimentally
observed running the catalyst sample loaded with 250 ppmw of sul-
phur. As discussed in [5], this may be associated with the observed
(and modeled) changes in the product distribution upon sulphur
addition, i.e. the decrease of the selectivity to heavy products, which
contain only paraffins, and the increase of the hydrogenating activ-
Fig. 3. Experimental and calculated CO conversion for samples loaded with different
amounts of sulphur (experimental conditions as in Fig. 1). Experimental data are
taken from Ref. [5].
ity of the catalyst. Indeed, while the former effect would increase
the olefinicity of the products, the latter would decrease it, result-
ing in a non-monotonic trend of the olefin selectivity as a function
of the sulphur loading on the catalyst.
In terms of ASF product distribution, in our previous works
[14,15] we showed that the developed model is able to satisfac-
torily describe the product selectivity, accounting for the typical
deviations of the product distribution from the ASF model, namely
Fig. 4. Experimental and calculated selectivities to the main reaction products for samples loaded with different amounts of sulphur: (a) CH4, (b) C2H4, (c) C5+ and (d) olefins
(experimental conditions as in Fig. 1). Experimental data in plots (a), (c) and (d) are taken from Ref. [5].

208
C.G. Visconti et al. / Catalysis Today 154 (2010) 202–209
Fig. 5. Total hydrocarbons (a) experimental and (b) calculated ASF distributions reported in terms of carbon selectivity (Sn) for samples loaded with different amounts of
sulphur (experimental conditions as in Fig. 1). Experimental data are taken from Ref. [5].
Fig. 6. Paraffins (a) experimental and (b) calculated ASF plots reported in terms of carbon selectivity (Sn) for samples loaded with different amounts of sulphur (experimental
conditions as in Fig. 1).
the high methane selectivity, the low selectivity to ethylene and the
change of the ASF slope with growing carbon atoms number. Also,
we showed that the developed model is able to correctly describe
the effects of all the investigated process conditions (temperature,
pressure, H2/CO inlet ratio and gas hourly space velocity) on the
products distributions with a single set of rate parameter estimates.
The model fits of the ASF products distributions in terms of total
hydrocarbons, paraffins and olefins for the samples loaded with
Fig. 7. Olefins (a) experimental and (b) calculated ASF plots reported in terms of carbon selectivity (Sn) for samples loaded with different amounts of sulphur (experimental
conditions as in Fig. 1).

C.G. Visconti et al. / Catalysis Today 154 (2010) 202–209
209
different amounts of sulphur are shown in Figs. 5–7. Thanks to a
single set of the 8 adaptive empirical parameters introduced in the
kinetic scheme in this work, the “modified” model nicely fits also
the variation of the ASF distribution upon increasing the sulphur
loading, accounting both for the increase of the selectivity of the
light species and for the decrease of the selectivity to the heavy
parameters, the model is able to describe the effects of the process
conditions on the products yields, well accounting for the typical
deviations of the products distribution from the ASF model, i.e. the
methane high selectivity, the low selectivity to C2 species and the
change of the slope of the ASF plot with growing carbon number.
Accordingly, the present model can be applied to predict the mod-
ification of the performances of the adopted catalyst due to the
sulphur poisoning both in terms of catalyst activity and selectivity.
products for sulphur loading greater than 100 ppmw
.
3. Conclusions
References
The complete kinetic model of FTS over a Co/Al2O3 un-promoted
catalyst developed in [14,15] for a fresh unpoisoned catalyst has
been extended in this work in order to make it able to describe
also the activity and the selectivity of a sulphur-poisoned catalyst,
[1] E. Iglesia, Appl. Catal. A: Gen. 161 (1997) 59–78.
[2] A. Jess, R. Popp, K. Hedden, Appl. Catal. A 186 (1999) 321–342.
[3] S.T. Sie, M.M.G. Senden, H.M.H. van Wechem, Catal. Today 8 (1991) 371–394.
[4] M.E. Dry, Catal. Today 71 (2002) 227–241.
[5] C.G. Visconti, L. Lietti, P. Forzatti, R. Zennaro, Appl. Catal. A: Gen. 330 (2007)
49–56.
loaded with sulphur loads between 0 and 2000 ppmw
.
This result has been achieved by selectively introducing in the
original model 8 additional adaptive parameters. More specifically,
these parameters have been used to account for the effects of sul-
phur (i) in decreasing the number of the active sites on the catalyst
surface, that in turn induces the decrease of both the reactants con-
version and the products yields, (ii) in decreasing the chain growth
process rate, with the consequential decrease of the chain growth
probability and the shift of the process selectivity towards the light-
est reaction products, (iii) in increasing the hydrogenating ability
of the catalyst, with the consequent decrease of the olefinicity of
the products, and (iv) in increasing the olefins readsorption rate.
For the range of industrially relevant conditions (temperature =
210–235 ◦C, pressure = 8–25 bar, H2/CO inlet molar ratio = 1.7–2.3,
gas hourly space velocity = 4000–7000 cm3(STP)/h/gcat, sulphur
loading on the catalyst = 0–2000 ppmw) the developed model can
accurately predict both the observed CO conversion and the prod-
ucts distribution up to n = 50, in terms of total hydrocarbons,
n-paraffins and ␣-olefins. In particular, using a single set of 13 + 8
[6] H.G. Stenger, C.N. Satterfield, Ind. Eng. Chem. Proc. DD 24 (1985) 415–420.
[7] Z.T. Liu, J.L. Zhou, B.J. Zhang, J. Mol. Catal. 94 (1994) 255–261.
[8] C.H. Bartholomew, R.M. Bowman, Appl. Catal. 15 (1985) 59–67.
[9] T.C. Bromfield, N.J. Coville, Appl. Catal. 186 (1999) 297–307.
[10] Z.T. Liu, Y.W. Li, J.L. Zhou, Z.X. Zhang, B.J. Zhang, Appl. Catal. 161 (1997) 137–151.
[11] J. Li, N.J. Coville, Appl. Catal. 208 (2001) 177–184.
[12] R. Zennaro, M. Tagliabue, C.H. Bartholomew, Catal. Today 58 (2000) 309–319.
[13] C.G. Visconti, E. Tronconi, L. Lietti, R. Zennaro, P. Forzatti, Chem. Eng. Sci. 62
(2007) 5338–5343.
[14] C.G. Visconti, Z. Ballova, E. Tronconi, L. Lietti, R. Zennaro, P. Forzatti, Prepr. Pap.
– Am. Chem. Soc. Div. Pet. Chem. 53 (2) (2008) 91–95.
[15] C.G. Visconti, Z. Ballova, L. Lietti, E. Tronconi, R. Zennaro, P. Forzatti, in: B.H.
Davis, M.L. Occelli (Eds.), Advances in Fischer–Tropsch Synthesis, Catalysts and
Catalysis, CRC Press, Boca Raton, 2009, pp. 293–315.
[16] F. Fischer, H. Tropsch, Berichte der Deutschen Chemischen Gesellschaft 59
(1926) 830–831.
[17] P.M. Maitlis, Pure Appl. Chem. 61 (1989) 1747–1754.
[18] R.C. Brady, R. Pettit, J. Am. Chem. Soc. 102 (1980) 6181–6182.
[19] G. Buzzi-Ferraris, G. Donati, Chem. Eng. Sci. 29 (1974) 1504–1509.
[20] A.C. Hindmarsh, in: R.S. Stepleman, M. Carver, R. Peskin, W.F. Ames, R. Vich-
nevetsky (Eds.), Scientific Computing, North-Holland Publishing Company,
Amsterdam, 1983, pp. 55–64.