Applied Catalysis A: General 413414 (2012) 366374

Contents lists available at SciVerse ScienceDirect

Applied Catalysis A: General
j our nal homepage: www. el sevi er . com/ l ocat e/ apcat a
Kinetic analysis of the Ru/SiO
2
-catalyzed low temperature methane steam
reforming
M.A. Soria
a
, C. Mateos-Pedrero
a
, P. Marn
d
, S. Ord nez
d,
, A. Guerrero-Ruiz
b,c
, I. Rodrguez-Ramos
a,b
a
Instituto de Catlisis y Petroleoqumica, CSIC, Marie Curie no. 2, Cantoblanco, 28049 Madrid, Spain
b
Grupo de Diseno y Aplicacin de Catalizadores Heterogneos, UA CSIC-UNED, Madrid, Spain
c
Dpto. Qumica Inorgnica y Tcnica, Facultad de Ciencias UNED, Senda del Rey 9, 28040 Madrid, Spain
d
Department of Chemical Engineering and Environmental Technology, University of Oviedo, Facultad de Qumica, Julin Clavera 8, 33006 Oviedo, Spain
a r t i c l e i n f o
Article history:
Received 10 June 2011
Received in revised form
25 November 2011
Accepted 26 November 2011
Available online 6 December 2011
Keywords:
Hydrogen production
Steamreforming
Kinetic mechanism
Model discrimination
a b s t r a c t
The performance of a Ru/SiO
2
catalyst for methane steam reforming at 450550

C is studied in the
present. These conditions are suitable for coupling the xed-bed reactor with a hydrogen-selective mem-
brane for hydrogen recovery, with the subsequent equilibrium shift. A reaction mechanism based on the
dissociative adsorption of steam and methane has been proposed (from a total of six possible mecha-
nisms compared), in terms of the statistical analysis of reaction data obtained at different temperatures
and contact times in an integral, lab-scale reactor.
The proposed model shows that hydrogen inhibition plays an important role in the reaction. Finally,
the Ru/SiO
2
catalyst prepared in this work is found to be one of the most active catalysts, among other
Ru-based catalysts reported in the literature.
2011 Elsevier B.V. All rights reserved.
1. Introduction
The manufacture of alternative fuels, such as hydrogen or
synthetic gasoline, is currently a mayor challenge in chemical tech-
nology. Methane steam reforming reaction plays a key role in
the production of these fuels from either natural gas or biogas
[1,2]. Nickel catalysts are the most used at industrial scale, which
operates at high temperature (600900

C), pressure and steamto

methane ratio, to minimize coke formation. However, these condi-
tions are not suitable for using novel technologies to separate the
products, H
2
and CO
2
, such as in situ CO
2
capture [3] or membrane
reactors [4,5]. Membrane reactors consist of a multi-tubular xed-
bed reactor equipped with a palladium-based membrane that is
selective to hydrogen. Thus, this gas is separated with high purity
from the reaction products, and at the same time the shift of the
reforming reactor is favored. This reduces the size of the reactor
and the required amount of catalyst.
In these situations, an active catalyst providing high reaction
rates at low temperatures (<550

C) is required. In addition, low

steam to methane ratios are desirable in order to reduce reac-
tor volume, requiring catalyst less prone to bear carbonaceous
deposits. Noble metals (Ru, Rh, Pd, Ir and Pt) are very active for

Corresponding author. Tel.: +34 985103437; fax: +34 985103434.

E-mail address: sordonez@uniovi.es (S. Ord nez).
steamreforming. Among them, Ru and Rh have been shown to be
the most active and stable catalyst, but Ru is signicantly cheaper
[1,68].
Due to the economic importance of commercial steamreform-
ing process, the kinetic of methane steamreforming over Ni-based
catalysts has been extensively studied. Nevertheless, there is not
an agreement in the reaction mechanism and the corresponding
kinetic model for methanesteamreforming. This is explainedbythe
different nature of the catalyst andthe support, the catalyst particle
size and metal loading, the catalyst physical structure, the prepa-
ration method, or gas temperature, pressure and concentration
ranges. Generally, the dissociative methane adsorption reaction is
thought to be the rate determining step at most conditions, but
at low temperature CO formation reaction may became dominant
[3,911].
The kinetic models proposed to describe the kinetic behav-
ior consist of LangmuirHinshelwood, power-laws equations, and
expressions based on microkinetic analysis [12]. Temkin [13] stud-
ied the reforming kinetics on a nickel foil: at high temperature
(900

C) the reaction follows a rst-order kinetic equation with

respect to methane concentration, whereas at low temperature
(470530

C) hydrogen affects the reaction with negative reaction

order. Xu and Froment [14] developed one of the most popu-
lar methane steam reforming kinetic models for Ni/MgOAl
2
O
3
at 500670

C. The model is a LangmuirHinshelwood model

accounting for the adsorption of CO, CO
2
and H
2
on the same
0926-860X/$ see front matter 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.apcata.2011.11.030
M.A. Soria et al. / Applied Catalysis A: General 413414 (2012) 366374 367
Table 1
Summary of methane reforming kinetic models for Ru-based catalysts fromthe literature.
Reference Catalyst Loading Dispersion Kinetic model Conditions
Rostrup-Nielsen and Hansen [16] Ru/MgO 550650

C
1atm
Wei and Iglesia [7] Ru/Al
2
O
3
3.2% 44.2%
r
j
= kp
CH
4
k = 4.7 10
4
exp(10945]1) s
1
kPa
1
550700

C
15bar
Berman et al. [17] Ru/(-Al
2
O
3
+4.8% MnOx) 2%
r
j
=
kp
CH
4
b
CH
4
p
CH
4
+b
H
2
O
p
0.5
H
2
O
b
CH
4
= 4.42 10
6
exp(5694.2]1) atm
1
b
H
2
O
= 8.366 10
6
exp(4531.7]1) atm
0.5
k = 2.68mol]h kg
cat
atm
500900

C
Carrara et al. [18] Ru/La
2
O
3
0.6% 5%
r
j
=
K
1
k
2
K
3
K
4
p
CH
4
p
CO
2
K
1
K
3
k
4
p
CH
4
p
CO
2
+K
1
k
2
p
CH
4
+K
3
k
4
p
CO
2
K
1
= 1.46 10
6
exp(7242]1) kPa
1
k
2
= 2.94 10
3
exp(12949]1) mol]g
cat
s
K
3
= 4.05 10
8
exp(15891]1) mol]g
cat
s kPa
K
4
= 2.04 10
8
exp(26226]1) mol](g
cat
s)
510590

C
Jakobsen et al. [10] Ru/ZrO
2
1% 20%
r
j
=
kp
CH
4
[1 +K
CO
p
CO
+KH
2
(pH
2
)
1]2
]
2
k = 4.39 10
7
exp(12990]1) mol](g
cat
hbar)
K
CO
= 2.19 10
5
exp(10454]1) bar
1
)
KH
2
= 7.31 10
6
exp(8540]1) bar
0.5
425-575

C
1.3bar
adsorption sites occupied by CH
4
and H
2
O; the reactions between
adsorbedspecies are assumedtobe the rate determining steps with
reaction order <1 in CH
4
and <0 in H
2
.
Kinetic studies for Ru-supported catalysts are scarcer [15].
Table 1 summarizes the most important works carried out to elu-
cidate the kinetic model for methane reforming over Ru-based
catalysts. Rostrup-Nielsen and Hansen [16] studied Ni and noble
metal catalysts supported on MgO and found that Ru and Rh were
the most active ones. Reaction order with respect to methane
was found to be about 1 for Ru/MgO in the temperature range
of 550650

C. Wei and Iglesia [7], who studied methane steam

and CO
2
reforming in Ru supported on Al
2
O
3
, ZrO
2
, and NaY zeo-
lite, found that the rate limiting step is the dissociative adsorption
of methane. No dependency on H
2
O or CO
2
concentration was
observed, neither the type of catalyst support has inuence on cat-
alyst dispersion. They proposed a rst-order kinetic equation on
methane concentration in the temperature range of 550700

C
and total pressure range of 15bar. In the same way, Berman et al.
[17] workedwithMnO
x
promotedRu/-Al
2
O
3
catalysts, suggesting
reactionorders withrespect tomethane lower than1at 450500

C
and close to 1 at 700900

C (in good agreement with the ndings

of Wei and Iglesia [7]). Carrara et al. [18] focused their attention
on Ru/La
2
O
3
catalysts. After a detailed stability and characteriza-
tion study, they performed a kinetic study and tted a model with
dependency on CH
4
and CO
2
concentrations.
Jakobsen et al. [10] studied the kinetics of steam and CO
2
methane reforming on Ru/ZrO
2
at 425575

C and 1.3bar. The

results were consistently modeled using a LangmuirHinshelwood
expression, developed by considering that the rate limiting step is
methanedissociativeadsorption. Ru-basedcatalysts havealsobeen
proposed to carry out the reforming of organic feed stokes, such as
ethanol [1921].
As shown in Table 1, the kinetic models proposed for methane
reforming on Ru-based catalysts are quite different to each other,
mainly depending on the operating conditions and the catalyst
preparation method. Moreover, the temperature range of inter-
est for palladium-based membrane reactors (<550

C) has been
scarcely studied. The scope of the present work is to elucidate
the behavior of the methane steam reforming reaction at low
temperature (450550

C) for a Ru/SiO
2
catalyst. For accomplishing
this scope, reaction data have been tted to different mechanistic
models proposed in the literature, discriminating between these
mechanisms using statistical criteria. The presence of mass transfer
limitations has been also considered in this manuscript. Finally, the
performance of the catalyst tested in this work has been compared
to other Ru catalyst reported in the literature.
2. Experimental
2.1. Catalyst preparation and characterization
The supported Ru catalyst was prepared using a commercial
support: SiO
2
(silica gel from Fluka). Prior to metal loading the
support was calcined in air at 500

C for 4h after this treatment

the specic surface area of the support was 400m
2
/g. Then the
support was impregnated with Ru (4wt%) by means of the wet
impregnation method using an aqueous solution of RuCl
3
H
2
O
(SigmaAldrich) precursor. After Ruimpregnation, the catalyst pre-
cursor was dried in air at 110

C overnight [22].
The Ru/SiO
2
catalyst was characterized by BET, TPR-H
2
, XPS and
differential COadsorption heat-owmicrocalorimetry. Characteri-
zationresults for this catalyst have beendetailedina previous work
[6]. Metal dispersion measured by CO chemisorption was 16%. The
mean size of Ru particle derived frommetal dispersion was 8.4nm.
Theshapeof theCOadsorptionmicrocalorimetric proleindicates a
homogeneous distribution of the rutheniumsurface centers for the
chemisorption of CO, that is, they mostly have very similar geom-
etry and energetic interaction with the CO. The XPS Ru/Si atomic
ratio for Ru/SiO
2
sample is nearly the same before and after reac-
tion, suggesting that Ru particles still remain well dispersed after
test.
2.2. Experimental set-up
Steamreforming reaction was carried out at atmospheric pres-
sure in a xed-bed tubular reactor. The reactor with an inner
diameter of 11.8mm was heated in an electric furnace equipped
with a programmable temperature controller. A fresh catalyst
368 M.A. Soria et al. / Applied Catalysis A: General 413414 (2012) 366374
Fig. 1. Schematic diagramof experimental methane steamreforming system.
Table 2
Experimental reaction conditions for the steam reforming reaction over a Ru/SiO
2
catalyst.
Run H
2
O:CH
4
(vol.% ratio) W]F
CH
4
(kg
cat
s mol
CH
4
1
) Temperature (

C)
1 8:8 10.1 450, 500, 550
2 8:8 5.0 450, 500, 550
3 8:8 3.4 450, 500, 550
4 8:4 20.4 450, 500, 550
6 8:4 10.5 450, 500, 550
7 8:4 8.2 450, 500, 550
8 8:4 6.8 450, 500, 550
ground to 150250m, was diluted with SiC to obtain 50mmbed
height and packed inthe middle of the reactor (plug owis ensured
working with L/d
p
>50). The temperature was monitored by a K-
type thermocouple placed in the centre of the catalyst bed.
The catalyst was reduced in the reactor for 2h at 550

C (max-
imum operation temperature), using a ow rate of 100mL/min of
a mixture of 25vol.% H
2
in He. After reduction, He was used dur-
ing 30min to sweep the H
2
fromthe reactor. The feed consisted of
different mixtures of CH
4
, H
2
O and He. The desired reactant con-
centration was measured-controlled by mass owmeters (Brooks).
A schematic diagramof this installation is presented in Fig. 1.
The steamwas fed by saturating a He owwith distillate water
maintained in a thermostatic bath. The partial pressure of water
was adjusted by controlling the temperature of the bath. After
introducing water, the lines were heated (110

C) to avoid water
condensation. Table 2 summarizes the experimental conditions
used in methane steamreforming over Ru/SiO
2
.
Gas analyses of both reactants and products were carried out
by on line GC (Varian 3400) equipped with a TCD detector. Prior to
Table 4
Different alternative kinetic mechanism associated with methane and steam
adsorption.
Steamadsorption Methane adsorption
a
1
H
2
O+s H
2
O(s) b
1
CH
4
+s CH
4
(s)
a
2
H
2
O+s H
2
+O(s) b
2
CH
4
3s CH
2
(s) +2H(s)
b
3
CH
4
5s C(s) +4H(s)
s denotes the active site on the catalyst.
chromatographic analysis of the outlet reaction gases, water was
condensed employing a cool trap. Porapaq Q and Chromosorb 102
columns were used to separate the sample gas (CH
4
, H
2
, CO and
CO
2
). The carbon balance was close to 100% in all the cases.
3. Modeling
3.1. Kinetic model
Reaction between methane and steam can form many differ-
ent products, e.g. carbon, carbon monoxide, carbon dioxide, and
hydrogen. Using a thermodynamic analysis of the system, Hou
and Hughes [23] demonstrated that there are only three relevant
reactions at the experimental conditions of the present work, as
indicated in Table 3.
The kinetic models used in this work have been developed
by considering different reaction mechanisms. For all the cases,
methane and steam are considered to react as adsorbed species,
instead of species at gas phase [23]. The following possibilities are
considered for the adsorption mechanismof the reactants, as illus-
trated in Table 4 [23]: steam can be adsorbed with or without
Table 3
Thermodynamically relevant reactions, indicating equilibriumconstant (Keq), and approach to equilibriumparameter ().
i Reaction Equilibriumconstant, K
eq,i
Approach to equilibrium
I CH
4
+H
2
O CO+3H
2
exp(53.14 26, 830]1) [Pa
2
] I =
p
CO
p
3
H
2
p
CH
4
p
H
2
O
K
eq,l
II CO = H
2
O CO
2
+H
2
exp(4.04 +4400]1) [Pa
0
] II =
p
CO
2
p
H
2
p
CO
p
H
2
O
K
eq,ll
III CH
4
+2H
2
O CO
2
+4H
2
exp(49.10 22, 430]1) [Pa
2
] III =
p
CO
2
p
4
H
2
p
CH
4
p
2
H
2
O
K
eq,III
M.A. Soria et al. / Applied Catalysis A: General 413414 (2012) 366374 369
Table 5
List of kinetic expression developed fromthe reaction mechanisms considered.
M1 a
1
b
1
rI = kI 0
7
p
CH
4
p
H
2
O
(1 I )
rII = kII 0
3
p
CO
p
H
2
O
(1 II )
0 = (1 +K
H
2
O
p
H
2
O
+K
CH
4
p
CH
4
+KHp
1]2
H
2
+K
CO
p
CO
+K
CO
2
p
CO
2
)
1
M2 a
1
b
2
rI =
k
1
p
H
2
0
5
p
CH
4
p
H
2
O
(1 I )
rII = kII 0
3
p
CO
p
H
2
O
(1 II )
0 =
_
1 +K
H
2
O
p
H
2
O
+K
CH
4
p
CH
4
p
H
2
+KHp
1]2
H
2
+K
CO
p
CO
+K
CO
2
p
CO
2
_
1
M3 a
1
b
3
rI =
k
I
p
2
H
2
0
3
p
CH
4
p
H
2
O
(1 I )
rII = kII 0
3
p
CO
p
H
2
O
(1 II )
0 =
_
1 +K
H
2
O
p
H
2
O
+K
CH
4
p
CH
4
p
2
H
2
+KHp
1]2
H
2
+K
CO
p
CO
+K
CO
2
p
CO
2
_
1
M4 a
2
b
1
rI =
k
1
p
H
2
0
5
p
CH
4
p
H
2
O
(1 I )
rII =
k
II
p
H
2
0
2
p
CO
p
H
2
O
(1 II )
0 =
_
1 +K
H
2
O
p
H
2
O
p
H
2
+K
CH
4
p
CH
4
+KHp
1]2
H
2
+K
CO
p
CO
+K
CO
2
p
CO
2
_
1
M5 a
2
b
2
rI =
k
I
p
2.5
H
2
0
2
p
CH
4
p
H
2
O
(1 I )
rII =
K
II
p
H
2
0
2
p
CO
p
H
2
O
(1 II )
rIII =
k
III
p
3.5
H
2
0
2
p
CH
4
p
2
H
2
O
(1 III )
0 =
_
1 +K
H
2
O
p
H
2
O
p
H
2
+K
CH
4
p
CH
4
p
H
2
+KHp
1]2
H
2
+K
CHO
p
CH
4
p
H
2
O
p
2.5
H
2
+K
CO
p
CO
+K
CO
2
p
CO
2
_
1
M6 a
2
b
3
rI =
k
I
p
3
H
2
0
2
p
CH
4
p
H
2
O
(1 I )
rII =
k
II
p
H
2
0
2
p
CO
p
H
2
O
(1 II )
0 =
_
1 +K
H
2
O
p
H
2
O
p
H
2
+K
CH
4
p
CH
4
p
2
H
4
+KHp
1]2
H
2
+K
CO
p
CO
+K
CO
2
p
CO
2
_
1
dissociation, whereas methane can present more possibilities for
dissociation forming C or the intermediate specie CH
2
. By combin-
ing one mechanism of steam adsorption with one of methane, a
total of six possible kinetic mechanisms are obtained. As an exam-
ple, the set equations that form the so-called mechanism M5 are
indicated below. This kinetic mechanism was found to be the one
providing the best t to the experimental data.
H
2
O+s H
2
+O(s) k
1
p
H
2
0
O
p
H
2
O
0
(1)
CH
4
+3s CH
2
(s) +2H(s) K
2
0
CH
2
0
2
H
p
CH
4
0
3
(2)
CH
2
(s) +O(s) CHO(s) +H(s) K
3
0
CHO
0
H
0
CH
2
0
O
(3)
CHO(s) +s CO(s) +H(s) r
4
= k
4

2
_
0
CO
0
O

0
CO
2
0
K
eq5
_
(4)
CO(s) +O(s) CO
2
(s) +s r
5
= k
5

2
_
0
CO
0
O

0
CO
2
0
K
eq5
_
(5)
CHO(s) +O(s) CO
2
(s) +H(s) r
6
= k
6

2
_
0
CHO
0
O

0
CO
2
0
H
K
eq6
_
(6)
CO(s) CO+s K
7
=
p
CO
0
0
CO
(7)
CO
2
(s) CO
2
+s K
8
=
p
CO
2
0
0
CO
2
(8)
2H(s) H
2
+2s K
9
=
p
H
2
0
2
0
2
H
(9)
According to this mechanism steam is dissociated when
adsorbed (a
2
) (Eq. (1)). Methane is adsorbed to formthe intermedi-
ate CH
2
(s) (b
2
) (Eq. (2)), which reacts with oxygen in a second step
(Eq. (3)), to formanother intermediate, CHO(s). The CHO(s) specie
canreact inparallel toformCOandCO
2
(Eqs. (4) and(6)). Moreover,
COis alsooxidizedtoCO
2
followingthe water-gas-shit reaction(Eq.
(5)). Based on information well accepted in the literature [23], the
surface reactions, producing COandCO
2
(Eqs. (1)(6)), are assumed
as the rate controlling steps (RCS); the other reactions are con-
sidered to be at equilibrium. Eqs. (1)(9) are developed from the
assumptions previously indicated, and considering each reaction
of the mechanismis an elementary step [3,23].
In order to carry out the tting of this model with experimen-
tal data, the rate expression (Eqs. (1)(6)) must be a function of
the compound partial pressures, as shown in Table 5. This is done
by solving the equilibriumexpressions for the fraction of adsorbed
species (0
j
) and substituting in the rate expressions. The following
apparent kinetic constants are obtained when the rate expressions
are arranged to include the denitions of the equilibriumconstants
(K
eqi
), total concentration of active sites () and approach to equi-
libriumparameters (
i
) of Table 3.
k
I
= k
4

2
K
1
K
2
K
3
K
3]2
9
(10)
k
II
=
k
5
!
2
K
1
K
7
(11)
k
III
= k
6

2
K
2
1
K
2
K
3
K
3]2
9
(12)
The fraction of free active sites (0) is determined from a bal-
ance of active sites (Eq. (13)). The expression of Table 5 is obtained
substituting the expressions for the fraction of adsorbed species,
and dening the corresponding apparent equilibrium adsorption
constants (Eq. (14)).
1 = 0 +0
O
+0
CH
2
+0
H
+0
CHO
+0
CO
+0
CO
2
(13)
370 M.A. Soria et al. / Applied Catalysis A: General 413414 (2012) 366374
1 = 0
_
1 +K
1
p
H
2
O
p
H
2
+K
2
K
9
p
CH
4
p
H
2
+
1
K
1]2
9
p
1]2
H
2
+K
1
K
2
K
3
K
3]2
9
p
CH
4
p
H
2
O
p
2.5
H
2
+
1
K
7
p
CO
+
1
K
8
p
CO
2
_
(14)
This methodologyhas beenfollowedfor therest of kinetic mech-
anisms, resulting in the expressions shown in Table 5. Arrhenius
dependence with temperature of the kinetic constants is assumed.
It should be noted that mechanism M5 is the only one formed by
three rate equations, because considers the formation of interme-
diate specie CHO(s), which can react and form either CO or CO
2
.
Beside this, the kinetic models of Table 5 also differ in the expo-
nents of hydrogen partial pressure and the fraction of free active
sites (0). The rst can be very useful in the determination of the
kinetic mechanismthat best represents the experimental data. For
example, the type of steam adsorption determines the hydrogen
inhibition observed in the water-gas-shift reaction (r
II
), whereas
a similar relationship can be identied for the type of methane
adsorption and the reforming reaction (r
I
).
The kinetic equations canbe simpliedbyassuming that most of
the active sites of the catalyst are free, which means the compound
equilibrium adsorption constants (K
j
) have a low value, and are
neglected. This way, depending on the type of model, the unknown
parameters are reduced to pre-exponential factors and activation
energies for the kinetic constants.
3.2. Reactor model
The laboratory xed-bed reactor is modeled assuming plug ow
andisothermal conditions. The change of the compoundmolar ow
rates (F
j
) with reactor space time (z = W]F
/
0
) is calculated solving
the following set of differential equations:
dF
j
dz
= F
/
0
q
j
r
j
, j = 1. . . C (15)
where j denotes a compound (CH
4
, CO, CO
2
, H
2
O, H
2
and He), F
/
0
is
the inlet ow rate of the limiting reactant (CH
4
), q
j
is the internal
effectiveness factor of compound j, r
j
=

N
R
i=1

ij
r
i
is the reaction
rateof compoundj,
ij
is thestoichiometric coefcient of compound
j in the reaction i, and r
i
is the reaction rate determined using one
of the kinetic models of Table 5.
When the internal mass transfer resistance is neglected, the
internal effectiveness factor is 1. Otherwise, it is calculated using
the generalized Thiele modulus (
j
) by means of the following
expression corresponding to a rst order kinetic equation and
spherical catalyst particles [24]:
q
j
=
1

j
_
1
tanh(3
j
)

1
3
j
_
(16)

j
=
_
d
P
6
_

r
j

2
_

_
D
jc

C
c
j
_
c
jeq

r
j

dc
j
_

_
1]2
(17)
where d
P
is the mean catalyst particle size (200m),
C
is the cat-
alyst density, c
j
is the molar concentration, and D
jc
is the effective
diffusioncoefcient of compoundj, whichis calculatedfrommolec-
ular and Knudsen diffusion coefcients. In these calculations, the
textural properties of the catalyst measured in the BET test have
been used (mean pore size 6nm, and internal porosity 0.64).
The set of differential equations are solvedusingfunctionode15s
of MATLAB. Methane conversion and carbon monoxide yield are
calculated fromthe outlet molar owrates:
X
CH
4
= 1
(F
CH
4
)
outlet
F
/
0
(18)
Y
CO
=
(F
CO
)
outlet
F
/
0
(19)
3.3. Fitting of model parameters
For given kinetic parameters, the kinetic and reactor models can
be solved together, and the outlet methane conversion and car-
bon monoxide yield determined. When the kinetic parameters are
unknown, the models can be solved iteratively in order to deter-
mine the kinetic parameters that best t the experimental data. The
tting problem is formulated as a weighted least-square problem
with two dependent variables (X
CH
4
and Y
CO
):
min

i
w
i
Nexp

k
(y
model
y
exp
)
2
(20)
The weight function (w
i
), necessary to normalize the objective
function, consists of the square maximum experimental value for
each dependent variable, methane conversion or carbon monoxide
yield.
All the calculations of the tting problemhave been written in
a MATLAB code, where the least-square problem is solved using
lsqnonlin function. This function solves the optimization problem
using the trust-region-reective algorithm. To ensure the conver-
gence of the algorithm, a goodapproximationto the optimummust
be introduce as a starting guess. For this reason the tting is done
in two steps: rst, the kinetic constants are t for each individual
temperature, and then the pre-exponential factors and activation
energies calculated using the Arrhenius plot are supplied as ini-
tial guess in the tting performed with all the experimental data.
This methodology has been successfully applied in previous works
[25,26].
4. Results and discussion
4.1. Preliminary experiments
The aimof the preliminary experiments is to examine the cata-
lyst stability and the inuence of diffusional effects.
The results published in a previous work [6] show the catalyst
is very stable at 550

C and a H
2
O/CH
4
molar ratio of 1:1. Methane
conversion decreases by about 9% after 15h of time on stream. For
this reason, all the kinetic measurements carried out in the present
work have been performed after ageing the catalyst for 15h.
The inuence of diffusional effects, e.g. internal and external
mass and heat transfer, has been examined using the criterions
proposed in the literature [27].
The WeiszPrater criterion of absence of internal mass trans-
fer, (r
vc
)
obs
r
2
p
]D
jc
c
jS
-1, was fullled for most of the experimental
conditions of Table 2. However, it was found that for some cases
the criterion slightly excess 1. Instead of discarding these data, the
reactor model of Section 3.2 has been used, and the internal mass
transfer is calculated for the actual kinetic expression by means of
the effectiveness factor. Once the kinetic model is t, the inuence
of internal mass transfer will be re-examined.
The criterion of absence of external mass transfer limitations,
(r
vc
)
obs
r
p
]k
C
c
jC
-0.15, was fully satised (the maximum value of
the left-handtermwas foundto be 0.002 for methane consumption
and 0.05 for CO formation).
M.A. Soria et al. / Applied Catalysis A: General 413414 (2012) 366374 371
Fig. 2. Inuence of space time in methane conversion and carbon monoxide yield (symbols: experiments; lines: model predictions). Left graphs: CH
4
/H
2
O=4:8. Right graphs:
CH
4
/H
2
O=8:8. Temperature: ( ) 450

C, ( ) 500

C, ( ) 550

C.
Regarding internal and external heat transfer resis-
tances, the criteria

^H
r

(r
vc
)
obs
r
2
p
]z
c
1
S
-0.75R
g
1
S
]
u
and

^H
r

(r
vc
)
obs
r
p
]h
C
1
C
-0.15R
g
1
C
]
u
proposed, respectively, by
Anderson and Mears have been used. All the experiments fulll
these criteria, the higher left-hand terms are 4.610
3
and
3.710
3
, respectively.
4.2. Kinetic experiments
The kinetic experiments have been planned to study the inu-
enceof threevariables: methanetosteamratio, spacetime(W]F
/
0
),
and temperature, as indicated in Table 2.
The results from the experiments are shown in Figs. 2 and 3
(symbols), where the experiments corresponding to a different
methane to steamratios are shown in separate plots. The increase
of the reactor space time increases methane conversion and carbon
monoxide yield, which is in agreement with the expected behavior
of a continuous xed bed reactor. At higher temperatures, a higher
conversion is observed, as a consequence of both kinetic and ther-
modynamic effects (reforming equilibriumis shifted on increasing
temperature).
Fig. 3 depicts the experimental data together with the equilib-
rium (dashed lines). As shown, experimental methane conversion
data points are half-way to equilibrium, which means the kinetic
behavior can be elucidated from the experimental data. However,
the data are not completely free fromthe inuence of equilibrium,
which means that the reverse reaction must be taken into account
using the thermodynamic equilibrium constants, as indicated in
Section 3.1. Regarding carbon monoxide yield, the experimental
data obtained at the lower temperature are near to the equilibrium,
so these points are not useful to t the kinetic model.
4.3. Kinetic modeling and model discrimination
In this section, the kinetic models proposed in Section 3.1 (see
Table 5) are used to t the experimental data. The methodology
explained in Section 3.3 has been followed to determine the kinetic
parameters. All the kinetic models have been tted successfully,
Table 6
Statistical information of the model ttings used in the model discrimination.
SSE R
2
adj
M1 0.078 0.83
M2 0.036 0.90
M3 0.032 0.92
M4 0.038 0.88
M5 0.028 0.92
M6 0.033 0.92
In bold the model providing the best tting.
and the one providing the best t of the experimental data has
been identied following a discrimination analysis based on three
criteria.
The rst one uses the physical characterization of the kinetic
parameters [23]. Thus, models with negative kinetic constants or
activation energies out of the range for these reactions are rejected.
None of the models present such tting parameter values.
The second criterion is based on statistical information regard-
ing the quality of the tting. In this work, the sumof square errors
(SSEs) andthe adjustedregressioncoefcient (R
2
adj
) have beenused,
as shown in Table 6. The adjusted regression coefcient takes into
account that there are models with different number of tting
parameters. Basedonthis information, the best model is model M5,
because is the one with the lowest sumof square errors (0.028) and
the highest adjusted regression coefcient (0.92). However, differ-
ences with other models are very small, in particular with model
M3. For this reason, a third discrimination criterion is used.
The third criterion is based on the work of Box and Hill [28], that
presents a methodology to discriminate between different models,
using the probability of each individual experimental data belong-
ing to a particular kinetic model. Assuming normal distribution
of the experimental error, which is also a required condition of
most least-square based ttings, the following probability density
function of the i-th observation under the m-th model is obtained:
p
im
=
1
_
2o
2
m
exp
_

(y
im
y
i
)
2
2o
2
m
_
(21)
372 M.A. Soria et al. / Applied Catalysis A: General 413414 (2012) 366374
Fig. 3. Inuence of temperature in methane conversion and carbon monoxide yield (symbols: experiments; solid lines: model predictions; dashed lines: equilibrium). Left
graphs: CH
4
/H
2
O=4:8, space time ( ) 6.8, ( ) 8.2, ( ) 10.5, ( ) 20.4kg
cat
s mol
CH
4
1
. Right graphs: CH
4
/H
2
O=8:8, space time ( ) 3.4, ( ) 5.0, ( ) 10.1kg
cat
s mol
CH
4
1
.
where o
2
m
, the variance of the m-th model predictions, is approxi-
mated by the normalized mean square error.
The models are compared to each other using the posterior
probability (
im
), which is a measurement of the accumulated
probability of the m-th model fromthe rst to the i-th observation:

im
=

i1,m
p
im

m=1
Nm

i1,m
p
im
(22)
The posterior probability for the rst observation is assumed to
be the same for the six models,
0m
= 1]N
m
= 1]6. This method-
ology has been successfully applied by different authors [28,29].
Fig. 4 shows the trend followed by the posterior probability of
the models, fromthe rst to the i-thobservation. For the rst obser-
vations, most of the models exhibit a similar probability, whereas
fromobservation8 to the end, model M5 is clearly positionedas the
most probable. It should be pointed out that the graphical tness
of the best models (M5 and M3) is very similar; therefore model
discrimination is based on statistical analysis.
0%
10%
20%
30%
40%
50%
60%
70%
0 5 10 15 20 25

i
n
Exp ID
M1
M2
M3
M4
M5
M6
Fig. 4. Evolution of posterior probability on increasing the experimental data for
the different kinetic models considered.
Summarizing, model M5 has been found to be the best using the
statistical information obtained in the tting and calculating the
probabilities associated with each model. Recalling the develop-
ment of model M5, two important mechanistic assumptions were
made: both steam and methane dissociate when adsorbed, and
methane forms the intermediate specie CH
2
(s) that is oxidized in a
second step into CHO(s). Mechanism M5 is the only one that con-
siders the presence of intermediate specie CHO(s), which can react
to formeither COor CO
2
. On the contrary, all the other mechanisms
postulate that CO
2
is only produced fromCOby the water-gas-shift
reaction.
This mechanismcan be postulated as the most probable, among
the different mechanisms considered in this work. This model pre-
dicts a strong inhibition due to hydrogen, which is in agreement
with the ndings of other authors [9,10,14].
Table 7 summarizes the kinetic parameters corresponding to
model M5; activation energies for the reforming reactions (I and
III) are within the usual range reported in the literature for these
reactions and similar catalysts [9,10,14]. The quality of the tting
can be examined using the residual and parity plots of Fig. 5. The
residuals distribute randomly, which validates the adequacy of the
model to t the experimental data.
The predictions of model M5 have been depicted in Figs. 2 and 3
as solidlines. Ingeneral, it canbe concludedthat the model predicts
the experiments with satisfactory results. The mayor discrepancies
are reported for the lower space time, where the model predicts
slightly higher methane conversion and carbon monoxide yield.
Table 7
Values of the kinetic parameters corresponding to mechanism M5 that best t the
experimental data.
Reaction Kinetic constant at 500

C Activation energy
I 4.6mol kgcat
1
s
1
Pa
0.5
140kJ mol
1
II 1.7 10
4
mol kgcat
1
s
1
Pa
1
4.2kJ mol
1
III 0.41mol kgcat
1
s
1
Pa
0.5
144kJ mol
1
M.A. Soria et al. / Applied Catalysis A: General 413414 (2012) 366374 373
Fig. 5. Residuals (upper graph) and parity (lower graph) plots. ( ) Methane con-
version. ( ) Carbon monoxide yield.
4.4. Assessment of internal mass transfer
As explained in Section 4.1, preliminary estimations of inter-
nal mass transfer indicated that for some experiments the internal
diffusional resistance cannot be neglected. For this reason, the t-
ting of the experimental data was done using the reactor model
that includes the calculation of the internal effectiveness factor. In
this section, the predictions of this model using the kinetic model,
which was found to be the best (M5), are used to assess the inu-
ence of internal mass transfer.
In Fig. 6, the effectiveness factor prole is simulated for one the
experiments (550

C and CH
4
/H
2
O=8:8) for methane, steam, and
carbon monoxide, as being the compounds more representative
(for simplicity the effectiveness factor of the other compounds is
not shown). At the entrance of the reactor, internal mass transfer
Fig. 6. Effectiveness factor reactor proles obtained by simulating the kinetic model
M5. Conditions: CH
4
/H
2
O=8:8, 550

C, 10.1 kg
cat
s mol
CH
4
1
.
becomes important (q<1), because of the higher reaction rate pro-
duced by higher methane and steamconcentrations. Nevertheless,
fromFig. 6, it can be said that at the worst experimental conditions,
more than 82% of the bed has an effectiveness factor higher than
90%. Hence, it can be concluded that internal mass transfer has lit-
tle inuence on the experimental data, though this inuence has
already been taken into account in the modeling.
4.5. Comparison of Ru-based catalysts
The different Ru-based catalysts appearing in Table 1 have been
compared to each other and with the catalyst developed in the
present work. A direct comparison of methane conversion data
is not possible, because the experiments were carried out at dif-
ferent operating conditions, e.g. different catalyst weight, gas ow
rate, methane to steamratio, temperature, etc. For this reason, the
kinetic models proposed by these authors have been used instead
of direct experimental data.
The kinetic models of Table 1 have been used to calculate
the forward reaction rate of methane at the same operating con-
ditions, e.g. CH
4
/H
2
O=4:8, 1bar, and 20% methane conversion.
Temperature has been varied in the range 450550

C. The con-
centration of the different compounds has been calculated using
these operating conditions and assuming the water-gas-shift is in
Fig. 7. Comparison of the kinetic model of this work with other models fromthe literature in terms of the turn-over frequencies (CH
4
molecules/s Ru atoms) of the forward
reforming reaction. WGS reaction at equilibrium. Conditions: 4% CH
4
and 8% H
2
O feed concentrations, 20% methane conversion, 1bar pressure, and 450550

C temperature
range.
374 M.A. Soria et al. / Applied Catalysis A: General 413414 (2012) 366374
equilibrium. Nevertheless, even reaction rate per weight of catalyst
(mol/s kg
cat
) cannot be used to compare the performance of the
catalysts, because they were prepared with different metal load-
ings and dispersions, which would introduce an important bias in
the comparison. The use of the turn-over frequency (TOF), which
accounts for the reacted molecules of CH
4
per unit of time and
exposed atomof Ru, is a better choice to carry out the comparison:
TOF = r
j
_
M
Ru
Loading Dispersion
_
(23)
The TOF for the different catalysts is depicted as a function of
temperature in Fig. 7. When using logarithm coordinates for the
TOF, the catalysts can be classied into two groups, according to
the trend observed when increasing temperature. Thus, slope of
the line can be related with the apparent activation energy. On one
hand, the catalysts of Wei and Iglesia [7], Jakobsen et al. [10] and
this work, whichexhibit a highslope, andonthe other handthe cat-
alysts of Carrara et al. [18] andBermanet al. [17] witha lower slope.
The differences in TOF observed for the catalysts can be related to
the preparation method, which determines that a higher fraction of
exposed Ru atoms are placed in favored positions of the Ru particle.
The activity of the catalyst studied in this work is one of the
highest. Thus, for temperatures higher than 500

C, it is the most
active catalyst, whereas belowthis temperature the catalyst of Car-
rara el al. [18] performs better. This statement must be considered
carefully because the model proposed by Carrara el al. [18] was
developed for the temperature range 510590

C, and the extrap-

olation belowthis temperature may be inaccurate.
5. Conclusions
In this work, the methane steam reforming over a Ru/SiO
2
catalyst has been studied at low temperature (450550

C) by
means of a detailed kinetic study. The experiments have been done
varying the most important reaction parameters: space velocity
(020kg
cat
/s mol), methane to steamratio (4:8 and 8:8), and tem-
perature (450550

C).
Different alternative reaction mechanisms between methane
and steam have been proposed, and the corresponding kinetic
expressions developed. The expression that best t the experimen-
tal data has been identied as the one where steam and methane
dissociate when adsorbing on the catalyst. Methane originates
specie CH
2
(s), that is oxidized into CHO(s). This mechanismis sim-
ilar to other mechanisms for this reaction, but with other catalysts
[3,14,23].
Finally, the performance of the catalyst has been compared to
other Ru-based catalysts fromthe literature in terms of TOF. It was
found that the metal dispersion can affect the observed appar-
ent activation energy. The catalyst tested in this work is one of
the most active at the low temperature range. For this reason, the
authors recommend the use of this catalyst to carry out the steam
reforming of methane in membrane reactors, which operate a low
temperature (450550

C), but with hydrogen permeation through

the reactor wall.
Acknowledgment
This work was nanced by the Spanish Ministry for Science
and Innovation (contract CIT-120000-2008-4, Applied Collabora-
tive Research Program2008).
References
[1] J.R. Rostrup-Nielsen, J. Sehested, J.K. Nrskov, Advances in Catalysis, Academic
Press, 2002, pp. 65139.
[2] K. Aasberg-Petersen, J.H. Bak Hansen, T.S. Christensen, I. Dybkjaer, P.S. Chris-
tensen, C. Stub Nielsen, S.E.L. Winter Madsen, J.R. Rostrup-Nielsen, Applied
Catalysis A: General 221 (2001) 379387.
[3] M.H. Halabi, M.H.J.M. de Croon, J. van der Schaaf, P.D. Cobden, J.C. Schouten,
Applied Catalysis A: General 389 (2010) 8091.
[4] R. Sanz, J.A. Calles, D. Alique, L. Furones, S. Ord nez, P. Marn, P. Coren-
gia, E. Fernandez, International Journal of Hydrogen Energy, in press,
doi:10.1016/j.ijhydene.2011.08.102.
[5] S. Tosti, International Journal of Hydrogen Energy 35 (2010) 1265012659.
[6] M.A. Soria, C. Mateos-Pedrero, I. Rodrguez-Ramos, A. Guerrero-Ruiz, Catalysis
Today 171 (2011) 126131.
[7] J. Wei, E. Iglesia, The Journal of Physical Chemistry B 108 (2004) 72537262.
[8] G. Jones, J.G. Jakobsen, S.S. Shim, J. Kleis, M.P. Andersson, J. Rossmeisl, F.
Abild-Pedersen, T. Bligaard, S. Helveg, B. Hinnemann, J.R. Rostrup-Nielsen, I.
Chorkendorff, J. Sehested, J.K. Nrskov, Journal of Catalysis 259(2008) 147160.
[9] J. Wei, E. Iglesia, Journal of Catalysis 224 (2004) 370383.
[10] J.G. Jakobsen, T.L. Jrgensen, I. Chorkendorff, J. Sehested, Applied Catalysis A:
General 377 (2010) 158166.
[11] M.H. Halabi, M.H.J.M. de Croon, J. van der Schaaf, P.D. Cobden, J.C. Schouten,
Applied Catalysis A: General 389 (2010) 6879.
[12] M. Maestri, D.G. Vlachos, A. Beretta, G. Groppi, E. Tronconi, Journal of Catalysis
259 (2008) 211222.
[13] M.I. Temkin, in: H.P.D.D. Eley, B.W. Paul (Eds.), Advances in Catalysis, Academic
Press, 1979, pp. 173291.
[14] J. Xu, G.F. Froment, AIChE Journal 35 (1989) 8896.
[15] M.M. Yung, W.S. Jablonski, K.A. Magrini-Bair, Energy & Fuels 23 (2009)
18741887.
[16] J.R. Rostrup-Nielsen, J.H.B. Hansen, Journal of Catalysis 144 (1993) 3849.
[17] A. Berman, R.K. Karn, M. Epstein, AppliedCatalysis A: General 282(2005) 7383.
[18] C. Carrara, J. Mnera, E.A. Lombardo, L.M. Cornaglia, Topics in Catalysis 51
(2008) 98106.
[19] A. Iulianelli, T. Longo, S. Liguori, P.K. Seelam, R.L. Keiski, A. Basile, International
Journal of Hydrogen Energy 34 (2009) 85588565.
[20] P.D. Vaidya, A.E. Rodrigues, Industrial & Engineering Chemistry Research 45
(2006) 66146618.
[21] S. Tosti, A. Basile, F. Borgognoni, V. Capaldo, S. Cordiner, S. Di Cave, F. Gallucci,
C. Rizzello, A. Santucci, E. Traversa, Journal of Membrane Science 308 (2008)
250257.
[22] P. Ferreira-Aparicio, C. Mrquez-Alvarez, I. RodrIguez-Ramos, Y. Schuurman, A.
Guerrero-Ruiz, C. Mirodatos, Journal of Catalysis 184 (1999) 202212.
[23] K. Hou, R. Hughes, Chemical Engineering Journal 82 (2001) 311328.
[24] G.F. Froment, K.B. Bischoff, Chemical Reactor Analysis and Design, John Wiley
and Sons, 1979.
[25] P. Marn, C. Hamel, S. Ord nez, F.V. Dez, E. Tsotsas, A. Seidel-Morgenstern,
Chemical Engineering Science 65 (2010) 35383548.
[26] P. Marin, S. Ordonez, F.V. Diez, Catalysis Today 147 (2009) S185S190.
[27] M.E. Davis, R.J. Davis, Fundamentals of Chemical Reaction Engineering,
McGraw-Hill, 2003.
[28] G.E.P. Box, W.J. Hill, Technometrics 9 (1967) 5771.
[29] K. Jarosch, T. El Solh, H.I. de Lasa, Chemical Engineering Science 57 (2002)
34393451.