Catalysis in membrane reformers: A high-performance catalytic system for hydrogen production from methane

Article abstract of DOI:10.1016/j.jcat.2004.12.025

A palladium membrane reactor has been designed and sized to be applied to the dry reforming of methane for pure hydrogen production at a small scale. Three different parameters affecting the reactor operation have been adjusted to tune the reformer and optimise its performance: the extraction conditions, the CO₂/CH₄ ratio in the reactant mixture composition, and the reactants feed flow rate. By forcing H₂ extraction and adjusting the rates of hydrogen production in the reactor and hydrogen permeation through the Pd membrane, it has been possible to obtain very high CH₄ conversion for mixtures with CO₂/CH₄ ratios above unity. Reactant mixtures with CO₂/CH₄ ratios close to 2 offered the best results: high H₂ recovery yields (above 95%) and lower carbon deposition in the catalysts under the severe conditions imposed by the membrane reactor operation. Hydrogen extraction from the reaction side has been shown to enhance the carbon deposition rate on nickel-based reforming catalysts, in which the formation of low-reactivity carbon in the form of fibres has been observed. The dispersion of nickel on high oxygen mobility supports, such as Ce-Zr mixed oxides (Ce_0.5Zr_0.5O₂), which are chemically stable under reaction conditions, results in highly efficient catalysts capable of keeping their surface free of inactive carbon deposits. This kind of oxide provides an extra source of oxygen that equilibrates the net rates of carbon deposition and removal on the nickel surface, thus avoiding the accumulation of carbon in the catalyst.

Full text of DOI:10.1016/j.jcat.2004.12.025

Journal of Catalysis 231 (2005) 331–343
www.elsevier.com/locate/jcat
Catalysis in membrane reformers: a high-performance catalytic system
for hydrogen production from methane
a,∗
a
b
b
P. Ferreira-Aparicio , M. Benito , K. Kouachi , S. Menad
a
Instituto de Catálisis y Petroleoquímica, C/Marie Curie, s/n 28049 Madrid, Spain
b
Département de Chimie, Faculté des Sciences, Université Mouloud Mammeri, Tizi-Ouzou, Algeria
Received 11 November 2003; revised 8 August 2004; accepted 2 December 2004
Available online 17 March 2005
Abstract
A palladium membrane reactor has been designed and sized to be applied to the dry reforming of methane for pure hydrogen production at
a small scale. Three different parameters affecting the reactor operation have been adjusted to tune the reformer and optimise its performance:
the extraction conditions, the CO /CH ratio in the reactant mixture composition, and the reactants feed flow rate. By forcing H extraction
2
4
2
and adjusting the rates of hydrogen production in the reactor and hydrogen permeation through the Pd membrane, it has been possible to
obtain very high CH conversion for mixtures with CO /CH ratios above unity. Reactant mixtures with CO /CH ratios close to 2 offered
4
2
4
2
4
the best results: high H recovery yields (above 95%) and lower carbon deposition in the catalysts under the severe conditions imposed by
2
the membrane reactor operation. Hydrogen extraction from the reaction side has been shown to enhance the carbon deposition rate on nickel-
based reforming catalysts, in which the formation of low-reactivity carbon in the form of fibres has been observed. The dispersion of nickel on
high oxygen mobility supports, such as Ce–Zr mixed oxides (Ce Zr O ), which are chemically stable under reaction conditions, results
0.5 0.5 2
in highly efficient catalysts capable of keeping their surface free of inactive carbon deposits. This kind of oxide provides an extra source of
oxygen that equilibrates the net rates of carbon deposition and removal on the nickel surface, thus avoiding the accumulation of carbon in the
catalyst.

2005 Elsevier Inc. All rights reserved.
Keywords: Palladium membrane reformer; Hydrogen production; Dry methane reforming; Carbon nanofibres; Ni/Al O ; Ni/Ce Zr O ; Ni/ZrO ;
2
3
0.5 0.5
2
2
Ni/ZrO –CeO
2
2
1
. Introduction
and with reduced purification costs would make possible the
commercialisation of fuel cell-powered systems.
Large-scale hydrogen production from natural gas by
steam reforming or autothermal reforming is a well-known
process that uses purification methods such as pressure
swing adsorption or cryogenics. However, H2 production
at a smaller scale from fossil fuels requires further investi-
gation to meet the requirements of purity, economics, and
versatility for its use in fuel processor systems to be coupled
to fuel cells [1]. The utilisation of membrane reactors for this
purpose is one of the most interesting options to consider,
given its advantages related to heat and energy management,
and the integration of two different processes (reaction and
separation) in a single step.
Although hydrogen production has been a matter of great
importance in recent decades, a renewed interest in its pro-
duction processes has emerged recently, driven by the spec-
tacular advances in fuel cell technology. Up to now only a
small fraction of the hydrogen produced is currently used for
energy purposes, but the implementation of fuel cell systems
and the growing demand for zero-emission fuels in the near
future will increase the hydrogen share of the energy market.
Its production from abundant sources in an economical way
*
Corresponding author. Fax: +34 91 585 47 60.
E-mail address: pferreira@icp.csic.es (P. Ferreira-Aparicio).
0021-9517/$ – see front matter  2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2004.12.025

332
P. Ferreira-Aparicio et al. / Journal of Catalysis 231 (2005) 331–343
Several types of membranes have been tested in methane-
and resistance to carbon deposition of the different samples,
revealing that the catalytic behaviour of the nickel-based sys-
tems strongly depends on the support characteristics. The
design of high oxygen mobility supports that are chemically
stable under the severe reaction conditions imposed by the
reactor is the key for obtaining an active and stable catalytic
system capable of long-term operation in a membrane re-
former.
reforming reactions for hydrogen separation: ceramic mem-
branes [2,3], porous Vycor glass membranes [4–6], and
palladium-based films [7–10]. However, the application of
these modules to methane-reforming processes poses addi-
tional difficulties compared with other dehydrogenation re-
actions.
The small differences in the kinetic diameters of reactants
and products usually reduce the permselectivity in porous
membranes to the values corresponding to a Knudsen-type
diffusion mechanism of the different species [3]. Because
of this, dense membranes, whose operation is based on a
H2 solution-diffusion mechanism on a permselective film,
are usually the preferred choice. Although metals such as
Nb, Ta, V, and Zr and their alloys, particularly V–Ni, have
greater hydrogen permeability than Pd, they have greater
surface resistance to hydrogen transport due to their easier
oxidation. On the other hand, the high operation temperature
required for these reforming reactions limits the structural
stability of H2 permselective films and the utilisation of ma-
terials for membrane synthesis, fitting, sealing, enamelling,
etc. As a result, the reactor performance becomes limited
and controlled by numerous parameters, such as the thermal
stability of its components, their mechanical resistance, their
geometry, the membrane permeability and selectivity, not to
mention the catalyst stability under the more severe reac-
tion conditions imposed by the removal of hydrogen from
the reaction medium, which finally remains one of the most
important issues in the design of an efficient fuel processor.
Up to now, most of the efforts in this field have been focused
on the membrane itself and/or the reactor characteristics, but
little attention has been devoted to the catalyst behaviour
under the membrane reactor operation conditions. This lat-
ter question remains one of the most important challenges
to be broached for these systems. Usually, noble metals
have been used in membrane reactors to overcome problems
related to carbon deposition [11,12]. However, their high
price drives the development of alternative catalysts for this
process. Some simulation studies concerning the catalytic
performance of membrane reactors in the steam reforming
of methane have been published in recent years [13–15]. In
particular, Hou et al. noted in their study the increased ten-
dency towards catalyst poisoning by carbon deposition and
the stronger effect of contaminants such as H2S when hy-
drogen is removed from the reaction zone [15], especially at
temperatures below 873 K and pressures below 6 bar. In the
case of the dry reforming of methane, the tendency towards
carbon formation on the catalyst is even higher.
2
. Experimental
2
.1. Pd membrane preparation and characterisation
Palladium membranes were prepared by electroless plat-
ing deposition on the external surface of a porous 316L
stainless-steel (PSS) support from Mott Corp. According to
the manufacturer’s specifications, the grade of the PSS was
0
.5 µm, with an average pore diameter of 3 µm and a poros-
ity of 17%. The permeable zone of the membrane module
was designed to have a volume of approximately 2 cm (o.d.
9
part was welded at both extremes to two 15-cm-long pieces
of nonporous 316L stainless steel.
The membrane synthesis involved cleaning of the sup-
port, followed by successive sensitising, activation, and
plating cycles according to the procedure described by
Mardilovich et al. [16]. The activation procedure consisted
of successive immersions in an acidic SnCl2 bath (sensitis-
ing) and in deionised water, followed by an acidic PdCl2
bath at room temperature. Then the membranes were rinsed
with 0.01 M HCl and with deionised water to prevent hy-
3
.5 mm, wall thickness 1.6 mm, length 60 mm). The porous
drolysis of the Pd2 ions. After several sensitizing/activation
cycles, palladium was deposited at 333 K by an autocatalytic
reduction reaction with the use of a plating bath consisting
of a basic solution of Pd(NH3)4Cl2, NaEDTA, and N2H4.
Plating was followed by drying at 393 K. After each plating
cycle, the membranes were tested for He permeation at room
temperature to check for the absence of pores in the palla-
dium film. Membranes that were not dense were submitted
to another activation and plating cycle. Dense membranes
were used for reaction experiments. This procedure made it
possible to synthesise dense palladium membranes with a
minimum thickness and maximise the H2 permeation rate.
The thickness of each Pd layer obtained in this way was es-
timated from the weight increase of the module.
The membrane tube was fixed to a stainless-steel shell
with an inlet and an outlet to feed a sweep gas stream. The
use of independent Swagelok fittings with graphite ferrules
to integrate the membrane in the reactor and to connect it to
the experimental setup provided an easily handled gas-tight
seal, which avoided leakage problems between the retentate
and the permeate sides.
+
This communication reports a detailed study of differ-
ent parameters influencing the operation of a membrane
reformer for methane, with carbon dioxide as oxidant; the
extraction conditions, the feed flow rate, and the reactant
mixture composition have been adjusted to optimise the sys-
tem performance. Several nickel-based catalysts dispersed
on different supports have been prepared and tested in the
membrane reactor. This study analyses the catalytic stability
The permeation properties of the synthesised membranes
were evaluated for He and H2 at several temperatures be-

P. Ferreira-Aparicio et al. / Journal of Catalysis 231 (2005) 331–343
333
tween 623 and 823 K. Selectivity for hydrogen separation
was estimated from single gas permeation experiments.
reaction was analysed: the reactant mixture composition, the
sweep gas flow rate, and the reactant feed rate.
2
.2. Catalyst preparation and characterisation
2.4. Catalyst resistance to carbon deposition
A series of supported nickel catalysts was prepared for
A series of experiments was carried out for the different
catalysts in the membrane reactor under the same conditions
(temperature, extraction flow rate, reactant feed flow rate, re-
actant mixture composition) to determine their resistance to
carbon deposition. After 24 h in operation, spent catalysts
were examined by scanning electron microscopy, and their
carbon deposits were determined by gravimetric analysis af-
ter calcination at 1173 K. The reactivity with hydrogen of
the carbon deposits formed on the several catalysts was also
studied by temperature-programmed hydrogenation. Finally,
the residual coke remaining on the catalysts after hydrogena-
tion at 1073 K was also gravimetrically determined.
the membrane reactor operation. Four different supports
were used: Al2O3 (Puralox Condea), ZrO2 (Mel Chemicals),
ZrO2–CeO2 (Mel Chemicals), and a Ce–Zr mixed oxide
with a Zr:Ce atomic ratio of 1.
The ZrO2 support contained 3.5 wt% SiO2 to stabilise the
zirconia tetragonal phase. It was obtained by calcination of
its hydroxide at 973 K. The ZrO2–CeO2 support, according
to the manufacturer’s specifications, had a CeO2 content of
1
7.5 wt%. Finally, a mixed oxide with a Zr/Ce ratio of 1:1
was prepared by the water-to-oil microemulsion method ac-
cording to the procedure described in [17].
We tested the oxygen exchange ability of the prepared
supports in a volumetric glass system by cycling at 923 K a
known amount of 18O2 through the sample, while analyzing
with a quadrupole mass spectrometer (Balzers QMS 421) the
evolution with time of the gas-phase composition.
Nickel catalysts with 5 wt% metal loading were prepared
on the above-mentioned supports by the incipient wetness
impregnation technique with aqueous solutions of Ni(NO3)2
To clarify the effect of hydrogen removal during reac-
tion, these experiments were also carried out for the several
catalysts in a conventional reactor. They were tested under
comparable conditions by replacing the membrane module
with a nonporous stainless steel tube of identical dimensions
to avoid differences due to the thermal conductivity of the
reactor material and its geometry, the catalyst disposition in
the reactor, and temperature effects.
(
Alfa). After the samples were dried overnight at room tem-
perature, the catalysts were calcined in air at 823 K. Cata-
lysts were pressed and sieved to particle sizes ranging from
3
. Results and discussion
0
.42 to 0.59 mm.
3
.1. Preliminary catalytic study
2
.3. Membrane reactor operation
The Ni/Al2O3 catalyst was initially used for the study of
For operation, the maximum temperature in the external
the different factors affecting the membrane reactor behav-
iour, as it is known to be one of the most efficient catalytic
systems for carrying out hydrocarbon reforming reactions
and, in particular, the dry reforming of methane [18]. A pre-
liminary catalytic study was carried out on this catalyst un-
der differential conditions in a conventional reactor to de-
termine its specific activity and catalytic properties for the
CH4/CO2 reaction. Table 1 summarises the catalytic activity
of this sample for CH4 conversion at different CO2/CH4 ra-
tios in the feed mixture composition. Other parameters, such
as the CO/H2 ratio after reaction and the selectivity for H2,
are also included.
furnace was fixed at 823 K because intermetallic diffusion
between palladium and iron is known to cause the hydrogen
permeation flux in the membrane to decay at higher tempera-
tures [16]. The system was not exposed to H2 at temperatures
below 573 K to avoid hydrogen embrittlement in the palla-
dium layer. Special care was taken during the heating and
cooling processes, in which the temperature was varied at
−
1
low rates (1 K min ) under an inert gas flow.
Some preliminary catalytic tests were carried out in a
conventional reactor to determine the amount of catalyst to
be loaded in the membrane according to its specific activity
for the reaction. For reaction tests, the catalyst was loaded in-
side the membrane tube, between two quartz wool plugs, all
along the permeation zone, and the experimental conditions
were changed to explore the system performance. It should
Table 1
Catalytic properties of the Ni/Al O catalyst for the dry reforming of
2
3
methane
◦
be noted that the reaction temperature was 20–40 C lower
CO /CH
feeding
A
X
(%)
CO/H
2
SH2
(%)
2
4
CH4
CH4
than the external temperature, depending on the operation
conditions. The temperature in the middle of the catalytic
bed was determined with a second thermocouple inserted
along the membrane in the reaction side.
We carried out the methane-reforming reaction by feed-
ing mixtures of CH4 and CO2 in different proportions with-
out any diluents. The influence of three parameters on the
−1 −1
s )
(µmol g
1
1
2.8
3.8
.0
.9
42.6
42.7
33.1
28.3
11.3
11.2
8.7
1.5
2.0
2.7
3.3
80
67
54
46
7.5
Reaction conditions: catalytic bed temperature, 783 K; pressure, 1 atm; CH
4
feed flow rate, 5.95 × 10 6 mol/s.
−

334
P. Ferreira-Aparicio et al. / Journal of Catalysis 231 (2005) 331–343
Fig. 1. (a) Theoretical values for equilibrium methane conversion in the dry methane reforming as a function of temperature for different CO /CH ratios in
2
4
the feeding: CO /CH = 1 (–1–), CO /CH = 2 (- -!- -) and CO /CH = 3 (··P··). (b) Methane conversion as a function of the CO /CH ratio at 783 K.
2
4
2
4
2
4
2
4
Circles, experimentally determined data. Dashed line, theoretical values.
It can clearly be seen that as long as the CO2/CH4 ra-
tio in the reactant mixture increases, CH4 conversion and
H2 selectivity decay, and the CO/H2 ratio increases. The
CO2 excess in the reactant mixture with respect to the dry
reforming reaction stoichiometry (CH4 + CO2 ⇐⇒ 2H2 +
middle of the catalytic bed. It must be considered that al-
though the increase in the CO2/CH4 ratio leads to a higher
CH4 conversion under the conditions dictated by thermody-
namics (Fig. 1), with respect to the specific activity of the
catalyst measured under differential conditions (Table 1), the
increase in the CO2/CH4 value causes the specific activity
of the catalyst to decrease, and a higher amount of catalyst
is required to achieve equilibrium.
2
CO) favours the occurrence of the reverse water gas shift
(
RWGS) reaction (CO2 + H2 ⇐⇒ CO + H2O). This sec-
ondary process consuming CO2 and the produced H2 leads
to a large increase in the CO/H2 ratio in the outlet stream
while producing H2O in the reaction medium. The decrease
observed in the specific activity for CH4 decomposition with
an increasing CO2/CH4 ratio in the feed is not due to deacti-
vation of catalytic sites but to the competition of the parallel
RWGS reaction for the nickel adsorption centres. Although
the utilisation of a CO2 excess in the reaction mixture can be
unfavourable from the standpoint of H2 production in terms
of selectivity, thermodynamics suggests the operation of the
dry methane reforming with CO2/CH4 ratios far above unity
to avoid regions with a thermodynamic potential for carbon
formation [19]. This is an important consideration to take
into account when a membrane reactor is used to extract the
hydrogen produced from the reaction medium.
3.2. Membrane reactor tests: adjustment of extraction
conditions, reactant mixture composition, and reactant feed
rate
On the basis of the previous results, several tests were
carried out with the membrane reactor to explore its behav-
iour. We estimated the mass of catalyst to be loaded into the
reactor by considering the specific catalytic activity of the
reforming catalyst (Ni/Al O ) and the permeability range
2
3
of the membranes. The thickness of the prepared palladium
films ranged from 12 to 35 µm, with permeances for H
2
−5
−5
−2 −1
−0.5
between 6.4 × 10 and 28.6 × 10 molm
s
Pa
measured at a temperature of 783 K. Values for hydrogen
permeance were determined from the measurement of the
hydrogen flux across the membranes at 783 K as a function
of the difference in pressure between the two sides of the
membranes. Hydrogen flux was observed to be directly pro-
Under equilibrium conditions, a higher CO2/CH4 ratio
in the feed also causes CH4 conversion to increase, accord-
ing to Le Chatelier’s principle. Fig. 1a shows the estimated
CH4 conversion under equilibrium conditions as a function
of the reaction temperature for different CO2/CH4 ratios in
the feed. These conversion data have been calculated from
the equilibrium constants, which were determined from the
Gibbs free energy values of reactants and products at dif-
ferent temperatures according to the reaction stoichiome-
try [20]. The experimentally determined values in a conven-
tional reactor under equilibrium conditions (Fig. 1b) have
shown a good concordance with those theoretically calcu-
lated at the temperature measured during reaction in the
portional to (P0 –P0 ) values corresponding to the trans-
membrane pressure difference, in agreement with Sievert’s
law for hydrogen solubility in palladium. Depending on the
Pd film thickness of the membrane module used for each
experiment, the maximum rate for hydrogen permeation at
.5
0.5
783 K for a H transmembrane pressure difference of 40 kPa
2
−5
−5
−1
oscillated between 2 × 10 and 9 × 10 mols . For this
permeance, approximately 1 g of sample was loaded along
the porous zone and reduced under pure hydrogen at the re-

P. Ferreira-Aparicio et al. / Journal of Catalysis 231 (2005) 331–343
335
Fig. 2. Effect of the sweep gas flow rate on the reactor performance for several CO /CH ratios: (a) methane conversion, (b) carbon dioxide conversion, and
2
4
(
c) hydrogen recovery yield. CO /CH = 1 (–1–), CO /CH = 2 (- -!- -), and CO /CH = 3 (··P··).
2
4
2
4
2
4
action temperature. At the previously indicated temperature
and with the mass of catalyst loaded, the system was able
operation improves progressively because of a faster H2 ex-
traction from the reaction medium. However, the reactor per-
formance greatly depends on the reactant mixture.
−5
−1
to convert CH4 at a rate of at least 3 × 10 mols for
CO2/CH4 ratios lower than 3 and produce H2 at a rate of
Fig. 2 compiles the CH4 and CO2 conversion levels ob-
tained in the membrane reactor and the H2 recovery yield
as a function of the sweep gas flow rate for three different
CO2/CH4 ratios ranging from 1 to 3. The recovery yield is
defined as the percentage of hydrogen obtained at the shell
side relative to the hydrogen content in the converted CH4.
As shown in Fig. 2a, a reactant feeding with a CO2/CH4
ratio close to unity causes CH4 conversion to rise when
moderated extraction conditions are applied. For very high
sweeping flow rates, CH4 conversion decays. At the same
time, CO2 conversion continues rising moderately because
of its reaction with carbon species formed from methane and
the remaining H2 (Fig. 2b). These conditions are not suitable
to maintain or increase the CH4 conversion level in the re-
actor, as the CO concentration is too high in the reaction
side. In terms of hydrogen recovery, it can be seen that very
high sweeping gas flow rates cause the H2 yield to decrease
slightly, in concordance with the lower CH4 conversion.
For mixtures with higher CO2/CH4 ratios, the use of high
sweep gas flow rates led to a continuous conversion increase
associated with an enhanced H2 recovery. These results in-
−
5
−1 −1
g . The amount of catalyst loaded was
6
× 10 mols
enough to ensure that the system would not operate lim-
ited by the catalytic activity of the catalyst, as the maxi-
mum conversion rate was similar or even higher than the
permeation rate at a hydrogen transmembrane pressure of
4
0 kPa. Taking into consideration these estimates, the reac-
tant mixture composition, the extraction conditions, and the
reactant feed rate were adjusted to optimise the system per-
formance.
First, two different parameters were changed: the reac-
tant mixture composition and the extraction conditions. As
the driving force for hydrogen permeation through the mem-
brane is proportional to the H2 partial pressure difference
across the Pd film, it is possible to produce a permeance rate
increase by reducing the total pressure at the shell side dur-
ing the reaction or by using a sweep gas flow that causes
the hydrogen partial pressure to come down at the shell side.
In this case a He stream was used as a sweep gas, showing
an important effect on the reactor behaviour. As a general
rule, when the sweep gas flow rate increases, the membrane

336
P. Ferreira-Aparicio et al. / Journal of Catalysis 231 (2005) 331–343
Fig. 3. Composition variation in the reaction side of the membrane reactor along with time: H (–1–), CH (- -!- -), CO (··P··), and CO2 (·-e-·).
2
4
−
5
−1
.
(
a) CO /CH = 1; (b) CO /CH = 2. Sweep gas flow rate, 3.72 × 10 mols
2
4
2
4
dicate that a certain CO2 excess in the feed with respect to
the stoichiometric ratio establishes more favourable condi-
tions for reactor operation. However, the CO2 surplus in the
feeding must be limited to a certain level to reduce as much
as possible H2O formation through the RWGS, given that
the hydrogen recovery yield becomes reduced for too high
CO2/CH4 ratios.
The observed behaviour is the result of a variety of fac-
tors involved in the whole reactor performance: the hydrogen
concentration in the reaction side, which depends on the re-
actant mixture composition, the CH4 conversion level and
the sweeping conditions, the presence of secondary reac-
tions consuming hydrogen from the reaction medium, and
the equilibration of all of these processes, which are inter-
related. The system equilibration is a complex process, as
can be inferred from the time needed to reach a stationary
state in the reaction side composition when low sweep gas
flow rates are used (Fig. 3). For an equimolecular CH4/CO2
mixture (Fig. 3a) or a feed with a CO2/CH4 ratio of 2
reaction between H2 and CO2. With higher sweeping gas
flow rates, the equilibrium situation is reached faster.
It is possible to further reduce the H2 content in the re-
action side by forcing the hydrogen extraction conditions
with the use of reaction mixtures containing excess CO2.
To maximise the H2 recovery yield it is also necessary to
adjust the hydrogen production and permeation rates. There-
fore, the reactant mixture composition and the extraction
conditions are fixed, the reactant feed flow rate would have
to be regulated to tune the reactor and optimise its perfor-
mance. Fig. 4a shows the effect of reducing the CH4 feed
flow rate on conversion. The CO2/CH4 ratio in the reac-
tant mixture was maintained at 1.9, and a sweeping stream
−1
of 0.223 mmol s was used. A large conversion increase
was observed when the CH4 flow rate was reduced below
−
5
−1
mols . It should be noted that for the en-
3.5 × 10
−6
−5
−1
tire range tested (from 6 × 10 to 7.2 × 10 mols of
CH4) the reaction was not limited by the amount of loaded
catalyst. The rate of hydrogen production varied between
−6
−1
(Fig. 3b), H2 concentration inside the membrane remains al-
8.0 × 10 mols (for the lowest feeding flow rate and the
−5
−1
most constant between 16 and 22 vol% when a sweep gas
highest conversion) and 4.3 × 10 mols (for the high-
est feeding flow rate and lowest conversion). By varying the
feeding rate, no significant modification is observed in the
hydrogen permeation rate through the membrane. Assum-
ing that under those conditions the used membrane is able
−1
stream of 0.037 mmol s flows at the shell side. However,
CH4 proportion becomes greatly reduced with time because
of a higher conversion level driven by H2 removal through
the membrane. This CH4 conversion enhancement causes
CO concentration to increase. On the other hand, CO2 con-
version variation is more evident when an excess of CO2
−6
−1
to extract H2 at a rate of 8.4 × 10 mols , the reduc-
tion of the hydrogen formation rate to the same level as the
permeation allows conversion to increase to three times the
equilibrium conversion level (23%) with a hydrogen recov-
ery yield above 95%, as shown in Fig. 4b.
(Fig. 3b) is used, although a similar tendency can be envis-
aged in both cases. Initially the CO2 concentration increases
because of the RWGS reaction avoidance produced by H2
removal, but its proportion in the reaction side is gradually
reduced as result of its consumption associated with CH4
conversion. The larger increment in CH4 conversion as com-
pared with that of CO2 also reveals the inhibition of the
Several studies have analysed parameters that influence
membrane reformer operation. They concur in noting the im-
portance of variables such as the load-to-surface ratio (L/S)
or the weight hourly space velocity (WHSV) [9,21–23]. In

P. Ferreira-Aparicio et al. / Journal of Catalysis 231 (2005) 331–343
337
Fig. 4. Effect of the CH feed flow rate on (a) the methane conversion and (b) the hydrogen recovery yield. Dotted line in (a): equilibrium conversion.
4
particular, Lin et al. have studied the steam methane reform-
ing in a Pd membrane reactor by modelling its behaviour as a
function of these two major working parameters: the WHSV
of CH4 and the L/S in the membrane [10]. In agreement with
our results, they have found a large conversion enhancement
by decreasing the CH4 WHSV and the L/S ratio. However,
their calculation indicates that conversion does not reach the
maximum value (100%), even for a large enough membrane
area (low L/S) or for low WHSV due to H2 removal lim-
itations, which are imposed by the experimental extraction
conditions and the membrane. In the case of the application
of a pressure difference between the reaction and permeation
side, further conversion improvement can be achieved with
a sweep gas stream and vice versa.
Table 2
Weight increase due to carbon deposition for the different nickel catalysts
after 24 h operation in the membrane (MR) and traditional (TR) reactors
Catalyst
Ni
SBET
MR
TR
(wt%)
2
−1
(m g ) ꢀW
C
ꢀW
C
hydrog
hydrog
(mg/gcat) (%)
(mg/gcat) (%)
Ni/Al2O3
Ni/ZrO
Ni/ZrO –CeO 4.8
4.9
4.4
178
110
68
545
464
615
33
29
21
18
81
580
45
100
45
2
2
2
2
2
O
Ni/Ce Zr
4.4
62
100
100
0.5 0.5
2
Ni (%), nickel loading in wt%.
SBET, catalyst specific surface area.
ꢀ
W, gravimetric determination of the deposited carbon in spent catalysts
after its calcination in air at 1173 K.
, percentage of hydrogenatable carbon from the total amount de-
C
hydrog
posited.
3
.3. Nickel-based reforming catalysts: resistance to carbon
side. (In a conventional reactor CO and CO2 concentrations
for identical conditions of temperature and feed are 20 and
44 vol%, respectively.)
deposition
The previous results will now allow us to study one of
the most important components of the reactor: the reforming
catalyst. Four different nickel-based catalysts were tested:
Ni/Al2O3, Ni/ZrO2, Ni/ZrO2–CeO2, and Ni/Ce0.5Zr0.5O2.
Their catalytic behaviour with respect to carbon deposi-
tion has been studied and analysed after operation in the
membrane reactor under the previously established optimum
reaction conditions. For these experiments, the membrane
reactor was fed with a mixture of CO2/CH4 at a ratio of
Table 2 compiles some of the catalyst characteristics and
the carbon deposition on them after 24 h of operation in the
membrane reactor and after 24 h of operation in a traditional
reactor under identical conditions but without hydrogen ex-
traction. It is clear from the results that, in the case of cat-
alysts such as Ni/Al2O3 or Ni/ZrO2, hydrogen extraction
from the reaction medium favours faster carbon deposition
on the catalyst surface. However, it should be noted that no
deactivation was observed in the reactors, although a sig-
nificant pressure drop appeared progressively in the cases
where catalysts with very high coke deposition rates were
used (Ni/ZrO2–CeO2). These results indicate that, despite
the high coking rate on the catalysts, the nickel surface sites
remain free for reactant activation. Moreover, it is important
to consider that catalyst fractions located at different heights
in a 5-cm-long fixed-bed reactor operate under different lo-
cal reaction atmospheres, differences that become even more
−6
−1
1
.9 and at a total flow rate of 5.97 × 10 mols and at-
mospheric pressure. H2 was extracted with a sweep gas flow
−1
rate of 0.223 mmol s . It is worth noting the high CO con-
centration in the effluent coming from the reaction side of
the membrane reactor under these conditions, even when a
mixture with a CO2/CH4 ratio close to 2 is used. During op-
eration, the CO concentration in the outlet gas composition
exceeds 35 vol%, although CO2 is maintained at a high level
(
46 vol%) to reduce CO disproportionation in the reaction

338
P. Ferreira-Aparicio et al. / Journal of Catalysis 231 (2005) 331–343
Fig. 5. Temperature programmed hydrogenation experiments for the supported nickel catalysts: (a) Ni/Al O , (b) Ni/ZrO , (c) Ni/ZrO –CeO , and
2
3
2
2
2
(
d) Ni/Ce Zr O . Left side, after 24 h operation in a traditional reactor. Right side, after 24 h operation in the palladium membrane reactor. Solid line, CH
0.5 0.5 2
4
production. Dashed line, CO production. Dotted lines, C –C hydrocarbons.
2
3
pronounced in the case of a membrane removing H2 from the
reaction medium. Thus, under the conditions used, a higher
carbon deposition at the end of the catalytic bed due to the
occurrence of the CO disproportionation reaction (2CO ⇐⇒
C + CO2) is expected.
bon deposition corresponding to the formation of reactive
species that can be completely removed by hydrogenation.
However, by the forcing of H extraction, high amounts of
2
low-reactivity deposits are obtained. A similar tendency is
observed in the Ni/ZrO sample under membrane reactor
2
A recent study regarding the use of fixed- and fluidised-
bed reactors applied to dry methane reforming with Ni/SiO2
catalysts reveals the superior performance of fluidised bed
reactors over fixed-bed reactors. In these latter reactors it is
confirmed that the used catalysts show a non-uniform coke
distribution deposited mainly at the end of the bed via the
Boudouard reaction [24]. The authors attribute catalyst de-
activation to the formation of less reactive carbon species
caused by a lower CO2/CH4 ratio in that zone. That study
was carried out with a clean model biogas with a CO2/CH4
ratio of 1.0. In our case, the CO2 excess in the reactant acts
as a “moderator” of the Boudouard reaction. Froment has re-
cently remarked the importance of modelling the behaviour
of steam and CO2 reformers to simulate the net rate of car-
bon formation on Ni catalysts [25]. This kind of study pro-
vides better insight into the role of the numerous variables
influencing carbon deposition, which increases with the dis-
tance in the catalytic bed and a higher CH4 partial pressure
in the feed. Wurzel et al. have also stressed in their stud-
ies the performance improvement obtained when fluidised
bed reactors are used for the CO2 reforming of CH4 [26].
They have attributed this better performance to lower deac-
tivation rates in the catalyst caused by the solid mixing into
the CO2-rich distribution zone, where carbon deposits are
removed by the reverse Boudouard reaction.
conditions, although this catalyst shows a higher disposition
for generating this low-reactivity coke in a traditional reactor
than does Ni/Al2O3.
The use of ceria-modified zirconia carriers changes the
catalytic behaviour dramatically, depending on the structure
and properties of the oxide support. Whereas Ni/ZrO2–CeO2
favours the formation of large amounts of coke in both mem-
brane and traditional reactors (about 60 wt%), small amounts
of reactive carbon species (about 4 wt% in the sample),
which can easily be removed by hydrogenation, are formed
on the Ni/Ce0.5Zr0.5O2 catalyst.
The temperature-programmed hydrogenation (TPH) pro-
files obtained from the different samples used in the tra-
ditional and membrane reactors provide more information
about the type of carbon species formed on each cata-
lyst (Fig. 5). In general, the main hydrogenation product
was CH4, although in some cases some light hydrocarbons
(C2–C3) and CO were also detected. As regards their re-
activity with hydrogen, two types of hydrogenatable car-
bonaceous species have been found. The higher reactivity
species, which can be hydrogenated at temperatures below
6
23 K and evolve as CH4 (or light hydrocarbons in the
case of the alumina-supported sample), have been assigned
β
in the literature to the so-called C carbon. These species
As can be inferred from the results in Table 2, apart from
the operation conditions imposed by the reactor, the sup-
port used to disperse the metallic phase is a key element
in preventing carbon deposition. In a conventional reactor,
the Ni/Al2O3 catalyst shows a good behaviour: low car-
are probably formed from an amorphous structure in which
carbon–carbon, carbon–metal, and hydrogen–carbon bonds
are present. This type of carbon may include several sub-
species, which differ in their mobility [27]. At temperatures
γ
above 673 K, low-reactivity graphitic carbon (C ) reacts

P. Ferreira-Aparicio et al. / Journal of Catalysis 231 (2005) 331–343
339
with hydrogen to yield CH4, although it can also produce
CO, depending on the support characteristics.
The TPH profiles obtained for the catalysts used under
the traditional reactor conditions reveal important differ-
ences among the studied nickel samples, depending on the
β
γ
oxide carrier properties. In Ni/Al2O3 C and C species
are formed, showing the particularity of producing small
amounts of light hydrocarbons at low temperature. Ni/ZrO2
γ
and Ni/ZrO2–CeO2 form mainly C carbon, which at high
temperature (above 773 K) evolves as CO and continues
its production at temperatures above 1073 K. The presence
β
of CeO2 in the support stabilises some C species in the
Ni/ZrO2–CeO2 catalyst, in contrast with the unpromoted
sample. The production of certain amounts of CO during
hydrogenation at high temperature is caused by the reduc-
tion of the support to a certain degree. The reducibility of
ZrO2, CeO2, and Zr–Ce mixed oxides is related to the abun-
dance of mobile oxygen in their structure and their capability
to release it [28,29]. A completely different behaviour is
obtained for the Ni/Ce0.5Zr0.5O2 catalyst, for which only
β
high-reactivity carbon (C ) is obtained, in agreement with
the gravimetric analysis.
Fig. 6. Oxygen exchange rate of the oxides. Isotopic exchange carried
18
out at 923 K with
O in the gas phase. (2) Ce Zr O , (") CeO ,
2 0.5 0.5 2 2
The distribution of the hydrogenated carbon species is
different for the samples used under the membrane reactor
(Q) ZrO –CeO , (a) ZrO .
2 2 2
β
γ
conditions. As a general trend, a transformation of C in C
catalytic systems for the dry reforming of methane [17,31].
This is the case for mixed oxide Ce/Zr supports. Oxygen
exchange experiments between the oxides and 18O2 in the
gas phase reveal the higher oxygen mobility in the Ce/Zr
1:1 mixed oxide in comparison with the pure oxides, ZrO2
and CeO , and a CeO -impregnated ZrO support, as shown
β
is observed. C carbon only appears for the Ni/Ce0.5Zr0.5O2
catalyst, in which some CO is also observed at high tem-
γ
perature as result of the reaction of C carbon with oxygen
γ
released from the support. For the rest of the catalysts, C
species are observed to increase considerably in some cases,
2
2
2
β
to the detriment of the more reactive C carbon.
in Fig. 6. Studies by computer-simulation techniques us-
ing atomistic models on surfaces of cubic CeO –ZrO solid
The high tendency towards carbon deposition in the
Ni/ZrO2–CeO2 sample under the traditional reactor oper-
ation conditions contrasts with the Ni/ZrO2 behaviour. In
general, CeO2 addition to reforming catalysts as a promoter
causes conversion to increase, although ceria-supported met-
als usually show very low activities for this reaction [29–31].
The utilisation of supports able to provide an extra source
of oxygen (by favouring CO2 activation through carbonate-,
bicarbonate-, and/or formate-type intermediates or storing/
releasing oxygen through a redox process) opens alternative
pathways for the reaction, enabling a bifunctional mecha-
nism [32–36]. However, in the case of CeO2-supported cat-
alysts and metals dispersed on CeO2-impregnated supports,
a typical behaviour has been observed: after an initial pe-
riod of higher activity, a strong decay of the initial activity
in the DMR reaction is usually found [17,30,37]. The redox
properties of ceria and the high mobility of lattice oxygen
are probably the factors that initially promote catalytic ac-
tivity in the DMR reaction, but its easier reducibility under
the reaction conditions probably causes a loss of the oxy-
gen storage capacity of the support to provide active oxygen
species.
2
2
solutions have suggested that the increase in the oxygen
storage capacity experimentally observed in these systems
relative to pure ceria can be explained by the introduc-
tion of zirconia into the ceria lattice, a fact that decreases
the Ce4 /Ce reduction energy on stable surfaces and in-
creases the oxygen vacancies’ tendency to segregate to these
surfaces [38].
+
3+
Therefore, the striking differences observed among the
ZrO and the two Ce-modified ZrO nickel catalysts regard-
2
2
ing to their resistance to carbon deposition can be attributed
to the support characteristics associated with their different
structures, as revealed by the XRD patterns of the samples
(Fig. 7). The diffractogram corresponding to the Ni/ZrO2–
CeO catalyst shows mainly the peaks corresponding to
2
2
2
2
2
cubic ZrO , although signals corresponding to the mono-
clinic ZrO phase can be also appreciated. No diffraction
peaks corresponding to CeO are observed, indicating that
the promoter oxide is well dispersed on the ZrO surface.
The diffraction pattern of Ni/ZrO indicates a different struc-
2
ture corresponding to tetragonal zirconia. Although cubic
and tetragonal structures in ZrO are difficult to differen-
2
The design of supports with high oxygen mobility that
are chemically stable under reaction conditions has been
proposed as the key to obtaining highly efficient and stable
tiate, the broadening of the diffraction peaks at 2θ values
close to 50.5 and 60 , probably due to diffraction line split-
ting, suggests the formation of the tetragonal phase [39].
◦
◦

340
P. Ferreira-Aparicio et al. / Journal of Catalysis 231 (2005) 331–343
Fig. 7. XRD diffraction patterns of the fresh catalysts: (a) Ni/ZrO –CeO ;
2
2
Fig. 8. XRD patterns of Ni/Al O samples after 24 h operation in the re-
2 3
(
b) Ni/ZrO ; (c) Ni/Ce Zr O . Dotted lines, diffraction lines corre-
2 0.5 0.5 2
forming reaction (CO /CH = 1.9): (a) membrane reactor; (b) traditional
2
4
sponding to the cubic ZrO phase. (e) main diffraction peak of the NiO.
2
reactor. Dotted lines, diffraction lines corresponding to carbon fibres.
On the other hand, the mixed-oxide support prepared by the
water-to-oil microemulsion method shows only the signals
corresponding to the tetragonal Ce0.5Zr0.5O2 phase diffrac-
tion [40]. Nagai et al. have recently studied the local struc-
ture of CeO2–ZrO2 mixed oxides with the same composition
ratio. They have observed that the local arrangement in the
oxide at the atomic level is the main factor determining the
oxygen storage capacity (OSC) of this kind of material [41]:
separated ZrO2 and CeO2 phases, or solid solution forms
with Ce- and Zr-rich domains, show a much lower OSC than
solid solutions of the same average composition homoge-
neously arranged.
Used catalysts were also examined by XRD after 24 h of
operation in both membrane and traditional reactors. Figs. 8,
9
, and 10 show the diffraction patterns obtained for the
Ni/Al2O3, Ni/ZrO2, and Ni/ZrO2–CeO2 samples. In agree-
ment with the gravimetric weight variation analysis com-
piled in Table 2, the appearance of diffraction peaks at 2θ
◦
◦
values of 26 and 46 is observed, corresponding to car-
bon nanofibres in samples showing high weight increases
after reaction [42,43]. In the case of the Ni/ZrO2–CeO2 sam-
ple, signals corresponding to these carbon fibres are present
in both samples used in the membrane reactor and in the
conventional one. However, it can be appreciated that the
reaction conditions affect the support structure, given that
under the traditional reactor operation conditions the mono-
clinic ZrO2 phase undergoes transformation into the tetrag-
onal one. The Ni/Al2O3 and Ni/ZrO2 catalysts give place to
fibre formation only in the case of membrane reactor opera-
tion. However, in the latter, the incipient formation of these
Fig. 9. XRD patterns of Ni/ZrO (3.5% SiO ) samples after 24 h opera-
2
2
tion in the reforming reaction (CO /CH = 1.9): (a) membrane reactor;
2
4
(b) traditional reactor. Dotted lines, diffraction lines corresponding to car-
bon fibres.
structures for the traditional reactor can be guessed from the
XRD results, in concordance with the weight increase ob-
served in the sample after reaction. No changes in the XRD

P. Ferreira-Aparicio et al. / Journal of Catalysis 231 (2005) 331–343
341
Fig. 10. XRD patterns of Ni/ZrO (17.5% CeO ) samples after 24 h oper-
2
2
ation in the reforming reaction (CO /CH = 1.9): (a) membrane reactor;
2
4
(b) traditional reactor. Dotted lines, diffraction lines corresponding to car-
bon fibres.
pattern were obtained for the Ni/Ce0.5Zr0.5O2 catalyst, re-
gardless of the reaction conditions.
Fig. 11. SEM images obtained for the Ni/Al O catalyst after 24 h opera-
2
3
These results regarding the presence of carbon fibres were
confirmed by analysis of the samples by scanning electron
microscopy (SEM). SEM photographs gave direct evidence
of the kind of filamentous carbon structures formed on the
different catalysts. Fig. 11 shows micrographs obtained from
two distinct zones of the Ni/Al2O3 catalytic bed used in the
membrane reactor. Although some catalyst particles remain
free of filamentous carbon deposits, there are others that
appear to be completely covered by very long carbon nanofi-
bres with widths lower than 100 nm. The heterogeneity in
the carbon deposition distribution is due to the presence of
particles coming from different portions of the catalytic bed.
Whereas conditions at the initial part of the fixed bed inside
the membrane resemble those existing in a conventional re-
actor, catalyst particles located at the end of the catalytic bed
operate under more unfavourable conditions due to the lack
of H2 and the high CO concentration.
tion in the membrane reactor. (a) and (b) are photographs of two different
portions of the same catalytic bed.
On the Ni/ZrO2 and Ni/ZrO2–CeO2 samples carbon
deposits were more homogeneously distributed, proba-
bly because in these catalysts carbon deposition is more
favourable, even when no H2 is extracted from the reac-
tion medium. Figs. 12 and 13 show micrographs obtained
from these samples after 24 h of operation in the membrane
reactor, in which carbon filaments are covering the catalyst
surface.
Fig. 12. SEM image obtained for the Ni/ZrO catalyst after 24 h operation
2
in the membrane reactor.
face was found in most of the sample (Fig. 14a). However,
the presence of small amounts of segregated ZrO , detected
2
by energy dispersive spectroscopy, that do not form part of
the mixed oxide, should be noted. These small particles,
over which some carbon fibres have been found, are prob-
ably a very small portion of the whole catalyst, as they have
not been detected in the XRD pattern, and the weight incre-
A completely different aspect was found in the case of
the Ni/Ce0.5Zr0.5O2 catalyst, for which a carbon-free sur-

342
P. Ferreira-Aparicio et al. / Journal of Catalysis 231 (2005) 331–343
supported catalysts [43]. Their presence in distinct amounts
on the several studied catalysts gives an indication of the
relative rates between carbon deposition and removal. Tran-
sient kinetic studies on nickel catalysts have revealed that
CH4 adsorption and decomposition on Ni is a fast and re-
versible reaction step [45]. CO2 dissociation is also fast, but
the recombination of adsorbed carbon and oxygen species
has been proposed as the process limiting the reaction rate.
Despite the CO2 excess used in the feed, the carbon deposi-
tion rate in some of the catalysts under the membrane reactor
conditions is faster that its removal by reaction with CO2,
depending on the nickel support. This situation finally leads
to the accumulation of huge amounts of carbon in the form
of fibres. On the other hand, mixed-oxide-supported nickel
Fig. 13. SEM image obtained for the Ni/ZrO –CeO catalyst after 24 h
operation in the membrane reactor.
(Ni/Ce0.5Zr0.5O ) has exhibited excellent behaviour in reac-
tion, even under the most severe conditions. Its capability to
2
2
2
keep its surface free from carbon fibres seems to be due to
its ability to release continuously extra oxygen from its lat-
tice, to give it up to the nickel surface, thus enhancing the
oxidation rate of the adsorbed carbon species.
One of the key issues for a membrane reformer is obtain-
ing a long-lived, stable, high-performance catalyst that is re-
sistant to carbon deposition. However, even when no carbon
deposition is expected in the catalyst, it is desirable to use
fluidised beds to obtain a homogeneous catalytic bed. Re-
cently, interest in reactors that combine hydrogen-permeable
membranes with fluidised beds for methane reforming has
been growing [46–49]. This kind of system combines several
advantages of fluidised beds as chemical reactors, particu-
larly catalyst bed temperature uniformity and improved heat
transfer, with advantages offered by permselective mem-
brane technology: shifting the conventional thermodynamic
equilibrium and obtaining a reaction product in a very pure
state. In light of the presented results for a fixed-bed mem-
brane reactor, for which carbon deposition in the reforming
catalyst is associated with the different reaction atmosphere
existing along the reactor length, fluidised-bed membrane
reactors appear as a good alternative for solving these ques-
tions. The danger of abrasion, which is in fact a problem
for unsupported palladium membranes in bubbling beds, is
eliminated in the case of supported palladium films if they
are synthesised on the surface opposite the reaction side.
This would be one more step in reactor engineering design
to use with stable catalytic systems to obtain a long-lived,
high-performance reformer.
Fig. 14. SEM images obtained for the Ni/Ce Zr
O catalyst after 24 h
2
0
.5 0.5
operation in the membrane reactor: (a) surface of a typical catalyst particle;
b) detail of a particle covered with a segregated ZrO phase in which some
carbon nanofibres have grown.
(
4. Conclusions
2
A palladium membrane reactor has been designed and
sized for its application to the dry reforming of methane
for pure hydrogen production. Several operation parame-
ment corresponding to these carbon structures is negligible
(Fig. 14b). Surface segregation of ZrO2 in these mixed ox-
ides has been described previously by Hori et al. [44].
Although carbon fibres are low-reactivity species with re-
gard to H2, they can be gasified and removed by reaction
with CO2, as found by Takenaka et al. over different nickel
2 4 4
ters (sweep gas flow rate, CO /CH ratio, and CH feed
flow rate) have been adjusted to optimise the system perfor-
mance with regard to CH conversion, hydrogen recovery
4
yield, and reduced carbon deposition. Four supported nickel

P. Ferreira-Aparicio et al. / Journal of Catalysis 231 (2005) 331–343
343
catalysts have been tested under the established conditions:
Ni/Al2O3, Ni/ZrO2, Ni/ZrO2–CeO2, and Ni/Ce0.5Zr0.5O2.
A detailed analysis of their behaviour under the severe con-
ditions imposed by the reactor indicates that the oxide sup-
port clearly determines the catalyst resistance to carbon de-
position. Among the tested catalysts, Ni/Ce0.5Zr0.5O2 is the
only one able to keep its surface free from carbon fibres, be-
cause of the high oxygen mobility in the mixed oxide and its
chemical stability under reaction conditions. This high oxy-
gen mobility support provides an extra source of oxygen for
the oxidation of adsorbed carbon species on the metal sur-
face, equilibrating in this way the rates of carbon deposition
and removal on nickel and avoiding fibre formation.
[18] D. Sutton, S.M. Parle, J.R.H. Ross, Fuel Process. Technol. 75 (2002)
5.
4
[
[
19] A.M. Gadalla, B. Bower, Chem. Eng. Sci. 43 (1988) 3049.
20] I. Barin, O. Knacke, Thermochemical Properties of Inorganic Sub-
stances, Springer, Berlin/New York, 1973.
[21] Y.-M. Lin, G.-L. Lee, M.-H. Rei, Catal. Today 44 (1998) 343.
[22] Y.-M. Lin, M.-H. Rei, Int. J. Hydrogen Energy 25 (2000) 211.
[23] S. Goto, T. Tagawa, S. Assabumrungrat, P. Praserthdam, Catal. To-
day 28 (2003) 223.
[24] A. Effendi, K. Hellgardt, Z.-G. Zhang, T. Yoshida, Catal. Commun. 4
(2003) 203.
[
[
[
[
25] G.F. Froment, J. Mol. Catal. A: Chem. 163 (2000) 147.
26] T. Wurzel, S. Malcus, L. Mleczko, Chem. Eng. Sci. 55 (2000) 3955.
27] T. Koerts, M.J.A.G. Deelen, R.A. van Santen, J. Catal. 138 (1992) 101.
28] H. Vidal, J. Kaspar, M. Pijolat, G. Colon, S. Bernal, A. Cordón, V.
Perrichon, F. Fally, Appl. Catal. B: Environ. 30 (2001) 75.
29] H.S. Roh, K.-W. Jun, W.-S. Dong, J.-S. Chang, S.-E. Park, Y.-I. Joe,
J. Mol. Catal. A: Chem. 181 (2002) 137.
[
[
[
Acknowledgment
30] J.A. Montoya, E. Romero-Pascual, C. Gimon, P. del Angel, A. Mon-
zón, Catal. Today 62 (2000) 71.
31] F.B. Noronha, E.C. Fendley, R.R. Soares, W.E. Alvarez, D.E. Resasco,
Chem. Eng. J. 82 (2001) 21.
P.F.-A. acknowledges the Spanish Ministry of Science
and Technology for a Ramón y Cajal contract (call 2002).
[
[
32] J.H. Bitter, K. Seshan, J.A. Lercher, J. Catal. 171 (1997) 279.
33] P. Ferreira-Aparicio, I. Rodríguez-Ramos, J.A. Anderson, A. Guer-
rero-Ruiz, Appl. Catal. A: Gen. 202 (2000) 183.
References
[34] Z. Zhang, X.E. Verykios, S.M. MacDonald, S. Affrossman, J. Phys.
Chem. 100 (1996) 744.
[
[
1] D.G. Löffler, K. Taylor, D. Mason, J. Power Sources 117 (2003) 84.
2] L. Paturzo, F. Gallucci, A. Basile, G. Vitulli, P. Pertici, Catal. To-
day 82 (1–4) (2003) 57.
[35] J.H. Bitter, K. Seshan, J.A. Lercher, J. Catal. 176 (1998) 93.
[36] M.C.J. Bradford, M.A. Vannice, Appl. Catal. A: Gen. 151 (1997) 223.
[37] S. Wang, G.Q. Lu, Appl. Catal. B: Environ. 19 (1998) 267.
[38] G. Balducci, J. Kaspar, P. Fornasiero, M. Graziani, M.S. Islam, J. Phys.
Chem. B 102 (3) (1998) 557.
[3] B.S. Liu, L.Z. Gao, C.T. Au, Appl. Catal. A: Gen. 235 (2002) 193.
[4] A.K. Prahbu, S.T. Oyama, J. Membr. Sci. 176 (2000) 233.
[5] S. Uemiya, N. Sato, H. Ando, T. Matsuda, E. Kikuchi, Appl. Ca-
tal. 67 (1) (1990) 223.
[
39] U. Martin, H. Boysen, F. Frey, Acta Crystallogr. Sect. B: Struct. Sci.
[
[
[
6] T. Ioannides, X.E. Verykios, Catal. Lett. 36 (1996) 165.
7] A. Basile, L. Paturzo, Catal. Today 67 (1–3) (2001) 55.
8] J. Múnera, S. Irusta, L. Cornaglia, E. Lombardo, Appl. Catal. A: Gen.
4
9 (1993) 403.
[
[
40] S. Meriani, G. Spinolo, Powder Diffr. 2 (1987) 255.
41] Y. Nagai, T. Yamamoto, T. Tanaka, S. Yoshida, T. Nonaka, T. Okamo-
to, A. Suda, M. Sugiera, Catal. Today 74 (2002) 225.
42] S. Wang, Energy Fuels 12 (1998) 248.
43] S. Takenaka, E. Kato, Y. Tomikubo, K. Otsuka, J. Catal. 219 (2003)
245 (2003) 383.
[9] E. Kikuchi, Catal. Today 56 (2000) 97.
[
[
[
10] Y.-M. Lin, S.-L. Liu, C.-H. Chiang, Y.T. Chu, Catal. Today 82 (1–4)
(2003) 127.
176.
[
[
[
11] E. Kikuchi, Fuel & Energy Abstracts 38 (2) (1997) 106.
12] S. Irusta, L.M. Cornaglia, E.A. Lombardo, J. Catal. 210 (2002) 263.
13] J. Shu, B.P.A. Grandjean, S. Kaliaguine, Appl. Catal. A 119 (2) (1994)
[
[
[
44] C.E. Hori, H. Permana, K.Y. Simon Ng, A. Brenner, K. More, K.M.
Rahmoeller, D. Belton, Appl. Catal. B: Environ. 18 (1998) 105.
45] Y. Schuurman, V.C.H. Kroll, P. Ferreira-Aparicio, C. Mirodatos, Catal.
Today 38 (1997) 129.
46] A.M. Adris, C.J. Lim, J.R. Grace, Chem. Eng. Sci. 52 (1997) 1609.
[47] T. El Solh, K. Jarosch, H.I. de Lasa, Appl. Catal. A: Gen. 210 (2001)
315.
305.
[
[
[
14] J.S. Oklany, K. Hou, R. Hughes, Appl. Catal. A 170 (1998) 13.
15] K. Hou, M. Fowles, R. Hughes, Chem. Eng. Sci. 54 (1999) 3783.
16] P.P. Mardilovich, Y. She, Y.H. Ma, M.-H. Rei, AIChE J. 44 (2) (1998)
310.
[
17] S. Menad, P. Ferreira-Aparicio, A. Guerrero-Ruiz, I. Rodríguez-
[48] K. Jarosch, H.I. De Lasa, Chem. Eng. Sci. 54 (1999) 1455.
[49] Z. Chen, Y. Yan, S.S.E.H. Elnashaie, Chem. Eng. Sci. 58 (2003) 4335.
Ramos, Catal. Lett. 89 (1–2) (2003) 63.