You are on page 1of 12

Experimental and modeling studies on the low-temperature water-gas shift

reaction in a dense PdAg packed-bed membrane reactor


Diogo Mendes
a
, Sandra Sa
a
, Silvano Tosti
b
, Jose M. Sousa
a,c,n
, Luis M. Madeira
a,nn
, Ade lio Mendes
a
a
LEPAE, Departamento de Engenharia Qu mica, Faculdade de Engenharia, Universidade do Porto, Rua Dr. Roberto Frias, 4200-465 Porto, Portugal
b
ENEA, Unit a Tecnica Fusione, C.R. ENEA Frascati, Via E. Fermi 45, Frascati (RM) I-00044, Italy
c
Departamento de Qu mica, Escola de Ciencias da Vida e do Ambiente, Universidade de Tras-os-Montes e Alto Douro, Apartado 1013, 5001-801 Vila-Real Codex, Portugal
a r t i c l e i n f o
Article history:
Received 30 July 2010
Received in revised form
24 December 2010
Accepted 17 February 2011
Available online 24 February 2011
Keywords:
Low-temperature water-gas shift reaction
PdAg membrane reactor
Modeling
Simulation
Pure hydrogen
CO conversion
a b s t r a c t
In this work, an experimental and modeling study is described, focusing on the performance of a PdAg
membrane reactor recently proposed and suitable for the production of ultra-pure hydrogen. A packed-
bed membrane reactor (MR) with a nger-like membrane conguration has been used for carrying
out the water-gas shift reaction (WGS) in the region of low temperature operation using a simulated
reformate feed.
The experiments were performed under a broad range of operating conditions of temperature
(200300 1C) and space velocity (120010,800 L
N
kg
cat
1
h
1
); the effect of feed pressure (12 bar)
was also analyzed, as well as the operating mode at the permeate side: vacuum (30 mbar) or sweep gas
(1.0 bar; nitrogen at 1 L
N
min
1
). A one-dimensional, isothermal and steady-state model is proposed,
which assumes axially dispersed plug ow pattern and pressure drop in the retentate side and plug
ow with constant pressure in the permeate side. An innovative composed kinetic model was also used
to describe the catalytic activity of the catalyst for the WGS reaction. In general, the simulation results
showed a good agreement to the experimental data, in terms of carbon monoxide conversion and
hydrogen recovery (and also outlet retentate composition) using only two tting parameters related to
the decline of H
2
permeability due to the presence of CO. Both simulation and experimental runs
showed that the MR achieves high performances, for some operating conditions clearly above the
maximum limit for conventional packed bed reactors. The performance reached is particularly relevant
when hydrogen is recovered via sweep gas mode (a high sweep ow rate was employed), because a
lower partial pressure could be reached than using vacuum pumping. In the rst case, almost complete
CO conversion and H
2
recovery could be reached.
& 2011 Elsevier Ltd. All rights reserved.
1. Introduction
Humanity is facing one of its biggest challenges: the effect of
greenhouse-gas emissions on climate change. One of the main
contributors for the increased global warming is the carbon dioxide.
The increase of CO
2
concentration in the atmosphere is mainly
caused by human activities, as a consequence of the large use of
fossil fuels. In this sense, serious attention is being given to CO
2
abatement (Aresta and Dibenedetto, 2007; Perinline et al., 2008).
Polymer electrolyte membrane fuel cells (PEMFCs) represent
one of the most promising technologies to effectively reduce CO
2
emissions. These devices are electrochemical membrane reactors
that combine hydrogen (the fuel) and oxygen (from air) to
produce electrical power in an efcient way, exhausting just
water vapour (Barbir, 2005). However, strict hydrogen fuel
specications are required for low temperature PEMFC engine
vehicles and therefore ultra-pure hydrogen must be delivered
(limit of CO content depends on the temperature and devices
used, typically in the range from r0.2 ppm (ISO, 2008) till
10 ppm (Zhang, 2008; Srinivasan, 2006)).
From the point of view of the hydrogen fuel processor, such
task may be achieved integrating a hydrogen permselective
membrane, like a PdAg one, into a water-gas shift (WGS) reactor.
The goal of this process is to reduce the CO content in the
hydrogen streams according to the following chemical reaction:
COH
2
O 2CO
2
H
2
. In order to produce high-purity hydrogen
at the highest possible CO conversion and reaction rate, two-stage
adiabatic WGS converters are typically used in industrial practice.
Since each reactor operates in different temperatures range,
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/ces
Chemical Engineering Science
0009-2509/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2011.02.035
n
Corresponding author at: LEPAE, Departamento de Engenharia Qumica,
Faculdade de Engenharia, Universidade do Porto, Rua Dr. Roberto Frias, 4200-
465 Porto, Portugal.
nn
Corresponding author. Tel.: +351 225081519.
E-mail addresses: jmsousa@fe.up.pt (J.M. Sousa),
mmadeira@fe.up.pt (L.M. Madeira).
Chemical Engineering Science 66 (2011) 23562367
distinctive commercial catalysts are used: a Fe-based catalyst for
the high temperature stage (320360 1C) and Cu-based materials
for the low temperature operating stage (190250 1C) (Ratnasamy
and Wagner, 2009). The membrane can then be placed on the
outer surface of the catalytic bed and withdraw hydrogen con-
tinuously from the reaction zone (retentate side). As hydrogen
permeates through the membrane, the chemical equilibrium of
the WGS reaction is shifted towards the reaction products side,
according to Le Chateliers principle, thereby increasing the
carbon monoxide conversion. Increasing the driving force for
hydrogen permeation (difference between the hydrogen partial
pressure values in the retentate and permeate sides), the conver-
sion enhancement should also, in principle, increase. If the
permeate side is under vacuum, nearly pure hydrogen can be
collected from this chamber (as long as the membrane used is
only permeable towards H
2
). Another alternative which in prin-
ciple also allows increasing the driving force is using steam or an
inert as sweep gas, thereby increasing the hydrogen ux, but a
further separation unit is necessary to produce pure hydrogen.
The gas stream leaving the reactor from the retentate side is
composed predominantly by carbon dioxide, with small amounts
of hydrogen and water vapour. Further condensation of the steam
leaves a stream more concentrated in CO
2
, which may be recycled
into useful products (Olah et al., 2006) or may be compressed and
stored. In this case, the operating costs may be reduced.
Several authors have developed mathematical models with
various level of complexity reporting the advantages of using Pd-
alloy membrane reactors (MRs) for the WGS reaction. Most of
them are steady-state and one-dimensional (Adrover et al., 2009;
Basile et al., 2001; Brunetti et al., 2007; Criscuoli et al., 2000;
Gosiewski et al., 2010), but two-dimensional including axial and
radial gradients are also reported (Markatos et al., 2005).
Isothermal MR operation is generally assumed, but a few other
works include also energy balances, allowing to simulate non-
isothermal operation, adiabatic or non-adiabatic (Adrover et al.,
2009; Brunetti et al., 2007). However, very few studies critically
compare the proposed models with experimental results. More-
over, such comparisons are mostly based only in terms of
conversion (or conversion enhancement over the thermodynamic
value) and take into account data reported in the open literature
with respect to membrane properties and kinetic reaction rates,
most of the times obtained for very specic conditions, some-
times different from those employed in the simulations.
One of the rst works that combined experimental results and
mathematical modeling was performed by Uemiya et al. (1991).
These authors conducted the WGS reaction at 1 bar and 400 1C using
a commercial ironchromium oxide catalyst placed inside a 20 mm-
thick Pd membrane supported on porous glass tube. A simple
stream of CO and H
2
O was fed into the MR and argon supplied to
the permeation side, counter-currently, to sweep the permeated
hydrogen. To validate the experimental results, the authors devel-
oped a model based on simple assumptions, namely isobaric and
isothermal operation conditions and plug-ow pattern for both
retentate and permeate sides, but whose results agreed well with
the experimental data. A similar mathematical model was also
developed by Criscuoli et al. (2000) for a Pd-supported membrane
reactor using a typical reformate gas mixture for the WGS reaction
and kinetic data taken from literature. Also here the model tted
well the experimental results.
The main objective of the present study was to develop a
simplied model and critically compare it with the experimental
results obtained in a self-supported membrane, conceived especially
for the production of ultra-pure hydrogen. Experiments in a broad
range of operation conditions were carried out in a compact PdAg
membrane reactor, which arrangement was recently proposed
by Tosti et al. (2006) and Mendes et al. (2010b), packed with a
low-temperature WGS CuO/ZnO/Al
2
O
3
catalyst. This PdAg alloy
membrane shows both high permeability and selectivity towards
hydrogen (Tosti et al., 2006), being suitable to carry out the WGS
reaction. The innovative kinetic equation used in the model was
dependent on the temperature operation range, according to pre-
vious experiments obtained by the authors (see Section 2.2.2). The
model was validated in terms of CO conversion, retentate composi-
tion and hydrogen recovery for the temperature, feed space velocity
and feed pressure ranges considered. In addition, the model was
used to simulate the performance of the PdAg WGS MR in terms of
CO conversion and H
2
recovery in a wide range of the parametric
space described by Damk ohlers number (Da, ratio between the
reaction rate at the reference temperature and the feed ow rate
(Froment and Bischoff, 1990)) and a parametric contact time
(G, ratio between the characteristic feed ow time and the char-
acteristic permeation time for the reference component (Sousa et al.,
2001)) (cf. Section 3.2). These two parameters describe generally the
possible operation conditions to be used in the reactor, allowing
thus to dene the optimal operating regions in terms of CO
conversion and H
2
recovery.
2. Experimental
2.1. Palladiumsilver membrane tube
The PdAg permeator tube has been produced by cold-rolling
and diffusion welding of an annealed commercial metal foil
(75 wt% of Pd and 25 wt% of Ag, from Johnson Matthey) according
to a previously described technique (Tosti and Bettinali, 2004;
Tosti et al., 2001). This provided a self-supported 50 mm thick
Pd-based membrane tube with a diameter large enough (10 mm)
to allow hosting the catalyst inside it (in opposition to others
commercially available membranes with very low diameter
capillary tubes). In order to give the required mechanical stiffness
and to guarantee the tightness with the membrane module, a
stainless steel VCR
s
connection and a steel plug were brazed at
the ends of the membrane tube, as shown in Fig. 1.
The permeator tube was assembled inside the membrane
module in a nger-like conguration, which allows the elonga-
tion/contraction of the membrane following thermal and hydro-
gen permeation cycles. In this way, mechanical stress is avoided
and a long lifetime for the membrane is expected (Tosti et al.,
2006). The WGS catalyst was packed in the bore side and the
permeated hydrogen was collected in the shell side. The cong-
uration of the single-tube MR used is shown in Fig. 2.
2.2. Experimental set-up
The experimental apparatus used for carrying out the WGS
reaction, as well as for evaluating the membrane separation
Fig. 1. View of the PdAg membrane tube (geometrical characteristics of the
membrane section: 10 mm I.D., 50 mm thick and 50 mm of length).
D. Mendes et al. / Chemical Engineering Science 66 (2011) 23562367 2357
properties, is divided into three sections, namely, (i) the feed
section, consisting of gas cylinders and pressure regulators, mass
ow controllers (Bronkhorst Hi-Tec, model F201) and a Controller
Evaporator Mixer (CEM, Bronkhorst) unit to provide the desired
water vapour ow rate (streams are heated before the reactor
inlet to avoid condensation of water and to pre-heat the reaction
mixture), (ii) the reactor section, consisting of an oven (Memmert,
type UNE200maximum operating temperature of 300 1C) for
heating the membrane reactor, two pressure gauges (Druck, ref.
4010, 7 and 5 bar, respectively), two back-pressure regulators
(Swagelok) for controlling the pressure, a condenser and a
diaphragm vacuum pump (Thomas Instruments, ref.: 7011-
0069), and (iii) the analysis section, which consists of two ow-
meters (Bronkhorst Hi-Tec, model F201) to separately measure
the retentate and permeate streams and a gas chromatograph
(Dani 1000 GC) to analyze the retentate dry gas composition.
Further details concerning the set-up and analysis method can be
found elsewhere (Mendes et al., 2009).
2.2.1. Hydrogen permeation and WGS reaction tests
Pure hydrogen permeation tests were performed at temperatures
in the range 200300 1C, pressures in the range 1.12.5 bar and ow
rates in the range 50190 mL
N
min
1
. The H
2
feed gas owed along
the inner side of the membrane and the permeating stream ow
rate was measured on the shell side at atmospheric pressure with a
mass ow meter after steady state had been achieved (approxi-
mately 1 h), at the desired temperature. Neither sweep gas nor
vacuum was used. Pressures in the two sides of the membrane tube
were monitored via pressure transducers. Frequently, the selectivity
of the membrane was obtained against nitrogen. When pressurized
nitrogen was introduced in the retentate chamber and the module
closed, no decline in the pressure was noticed after 8 h, conrming
the full hydrogen permeation selectivity of the PdAg membrane.
Besides, no change in the membrane performance was noticed
during the experimental campaign. This is in line with a previous
work by Tosti et al. (2006), who observed the complete hydrogen
selectivity and durability of these permeators in long-term tests
(1 year).
As shown in Fig. 2, the MR considered in this work is a tube-
and-shell conguration system. The WGS reaction was performed
in the temperature range 200300 1C by packing 1.5 g of a
commercial CuO/ZnO/Al
2
O
3
catalyst, supplied by REB Research
and Consulting, in the bore side of the inner tube/membrane of
the nger-like membrane reactorcf. Fig. 2. As shown in Fig. 2,
the catalyst was loaded only in the bed section where there is the
membrane. Prior to the reaction runs, the WGS catalyst was
activated in situ with a mixed gas ow of H
2
/N
2
. Details about the
catalyst pre-treatment (reduction) are described elsewhere
(Mendes et al., 2009).
The reaction tests were carried out using the following feed
gas composition: 4.7% CO, 34.8% H
2
O, 28.7% H
2
, 10.2% CO
2
and
balanced in N
2
, which simulates a reformate feed coming from a
typical autothermal reforming of either ethanol (Salemme et al.,
2010) or liquid hydrocarbons (Pasel et al., 2004). The H
2
O/CO
ratio used was 7.4. This is a high value, leading in practice to a
signicant energy consumption (required for water demineraliza-
tion and vaporization). However, it is in the range employed by
industrial packed-bed reactors, where excess of steam is used to
favor the thermodynamic equilibrium shift. Besides, the feed gas
composition used herein is equal to that employed for determin-
ing the reaction kinetics (Mendes et al., 2010a).
Fig. 2. Scheme of the PdAg MR (sweep gas owing counter-currently).
D. Mendes et al. / Chemical Engineering Science 66 (2011) 23562367 2358
The reaction pressure (that is, the pressure in the lumen side
of the membrane) and the feed ow rate were varied in the range
12 bar and 30270 mL
N
min
1
, respectively. As the driving force
for H
2
permeation through the membrane is the difference
between the partial pressures on both sides of the PdAg
membrane, the permeation ux increases by reducing the hydro-
gen permeate pressure. This can be achieved by decreasing the
total permeate pressure (pure hydrogen) or using sweep gas.
Thus, two operating modes were studied: (i) hydrogen was
recovered in the shell side by vacuum pumping (shell side
pressure$30 mbar), and (ii) an inert gas (N
2
, 1 L
N
min
1
) was
fed into the shell side owing counter-currently (cf. Fig. 2). In this
case, a sufcient length of the sweep gas pipe was inserted inside
the oven in a spiral form to ensure pre-heating until the desired
temperature. Additionally, and to simulate the performance of a
packed bed reactor (PBR from now on) for posterior comparison
with the MR operation, the permeate chamber was closed and the
steady-state data were recorded.
Different parameters affecting the catalytic reaction and the
membrane operation were studied, namely the feed ow rate, the
reaction temperature, the feed pressure and the operation mode
(permeate side under vacuum or with sweep gas owing counter-
currently). Such effects have been evaluated in terms of the
carbon monoxide conversion (X
CO
) and hydrogen recovery
(Re
H
2
) at steady-state conditions. Both quantities were calculated
according to Eqs. (1) and (2), respectively:
X
CO
1
u
R,out
p
R,out
CO
u
R,in
p
R,in
CO
1
Re
H
2

u
P,out
p
P,out
H
2
u
R,out
p
R,out
H
2
u
P,out
p
P,out
H
2
2
where u is the interstitial velocity and p is the partial pressure.
The superscripts R and P stand for retentate and permeate
chambers, respectively, and in and out means inlet and outlet of
the reactor, respectively.
2.2.2. WGS reaction kinetics
The WGS kinetics and mechanisms over the above-mentioned
catalysts (Fe-based for the higher temperatures and Cu-based for
the lower temperatures) have been studied in the past by many
authors (Ayastuy et al., 2005; Koryabkina et al., 2003). However, it
is very controversial if the predominant reaction mechanism for
Cu-based materials is the redox or the associative (Langmuir
Hinshelwood type) one. On the other hand, the predominant
reaction mechanism for Fe-based systems is much less contro-
versial, being the redox mechanism the most accepted. In this
work, it was used a commercial CuO/ZnO/Al
2
O
3
catalyst (supplied
by REB Research and Consulting). Previous experiments with the
same catalyst showed a good relation between activity and
stability for the WGS reaction in the entire range of temperatures
tested (150300 1C) (Mendes et al., 2009). Experimental runs to
collect intrinsic kinetic data with this catalyst were then carried
out in a packed bed reactor, in the temperature range 180300 1C.
According to preliminary studies (Mendes et al., 2010a), two
different kinetic models were proposed. For temperatures
between 180 and 200 1C, the associative mechanism showed the
best tting, while the redox pathway showed the best agreement
in the range 215300 1C, according to Eqs. (3) and (4) below.
For the lower temperatures rangeLT (180200 1C):
R
LT

k
LT
p
CO
p
H
2
O
p
CO
2
p
H
2
=K
e
_ _
1

i
K
a,LT
i
p
i
_ _
2
3
For the higher temperatures rangeHT (215300 1C):
R
HT

k
HT
p
CO
p
H
2
O
p
CO
2
p
H
2
=K
e
_ _
p
CO
1K
a,HT
CO
2
p
CO
2
=p
CO
_ _ 4
where i refers to the i-th component, R stands for the local
reaction rate, k is the forward rate constant, K
e
is the thermo-
dynamic equilibrium constant and K
a
is the equilibrium adsorp-
tion constant of each species. In the membrane reactor, the
pressure of each species is the one at the retentate side. These
two kinetic models were used in the present work to describe the
catalytic activity of the catalyst for the WGS reaction carried out
in the PdAg membrane reactor.
The temperature dependence for the reaction rate and for the
adsorption equilibrium constants is described by the Arrhenius
and vant Hoff laws, Eqs. (5) and (6), respectively:
k k
0
exp
E
k
RT
_ _
5
K
a
i
K
a,0
i
exp
DH
a
i
RT
_ _
6
where k
0
is the pre-exponential factor of the reaction rate
constant, E
k
is the activation energy for the WGS reaction, R is
the gas constant, T is the absolute temperature, K
a,0
refers to the
pre-exponential equilibrium adsorption constant, and H
a
stands
for the enthalpy of adsorption.
Finally, the temperature dependence of the thermodynamic
equilibrium constant is described by Eq. (7) (Moe, 1962):
K
e
exp
4577:8
T
4:33
_ _
7
The parameters k
0
, E
k
, K
a,0
i
, and DH
a
i
were determined for each
temperature range (LT and HT) through the tting of the experi-
mental data and are shown in Table 1. In addition to the mean
estimated values are also given the tting error associated to each
parameter, assuming t-student distribution and for 95% con-
dence level, and computed using the 10 best ttings.
Analyzing the parameters for the LT regime, it can be inferred
that, under the operating condition used in this work, the
adsorption of CO and H
2
O at the catalyst surface is much lower
than the adsorption of CO
2
and H
2
. In this way, the kinetic
equation considered in the simulation calculations performed
along this study was simplied to:
R
LT

k
LT
p
CO
p
H
2
O
p
CO
2
p
H
2
=K
e
_ _
1K
a,LT
CO
2
p
CO
2
K
a,LT
H
2
p
H
2
_ _
2
3a
3. Membrane reactor model
3.1. Development of the model
The catalytic membrane reactor considered in this study has
the general features described above. The pseudo-homogeneous
1-D model proposed for describing this reactor is based on the
following main assumptions:
Steady-state and isothermal operation.
Axially dispersed plug-ow pattern in the retentate side with
pressure drop described by Ergun equation.
Negligible mass and heat-transfer resistances.
Ideal plug-ow pattern in the permeate side with no
pressure drop.
Ideal gas behavior.
D. Mendes et al. / Chemical Engineering Science 66 (2011) 23562367 2359
The assumption of isothermal operation in a reactor depends
on the extent of the reaction heat compared to the heat loss/gain
through the reactor walls. In low-scale (laboratory) reactors, like
the one considered in the present study, this hypothesis is
acceptable, due to the high ratio between heat transfer area and
the reaction volume or when the consumption or release of heat
is low. However, this condition may not be true when higher
process scales are used (Adrover et al., 2009).
The axial and radial dispersion inside the reactor may have a
large effect on the predicted conversion (Koukou et al., 1996).
However, radial concentration gradients should be negligible
when the permeation ux is comparatively lower than the
convective ux (Tiemersma et al., 2006). Therefore, we considered
in the present model only axial dispersion.
The governing equations for the retentate (reaction) and
permeation sides are as follows:
Retentate sidePartial mass balance:
d
dz
u
R
p
R
i
_ _

d
dz
D
ax
P
R
d
dz
p
R
i
P
R
_ _ _ _

2pr
m
e
b
A
R
RTJ
i

W
cat
e
b
V
R
RTn
i
R 0 8
Total mass balance:
d
dz
u
R
P
R

2pr
m
e
b
A
R
RT

i
J
i

W
cat
e
b
V
R
RT

i
n
i
R 0 9
Pressure drop:
dP
R
dz
150
m
g
1e
b

2
e
b

3
d
p

2
u
R
1:75
r
g
1e
b

d
p
e
b

3
u
R
9u
R
9 0 10
Boundary conditions:
Danckwerts boundary conditions for retentate side (Froment
and Bischoff, 1990):
z 0:
d
dz
p
R
i
P
R
_ _

u
R
e
b
D
ax
p
R,in
i
p
R
i
_ _
P
R
,
u
R
u
R,in

u
F
e
b
and P
R
P
R,in
P
F
11
z :
d
dz
p
R
i
P
R
_ _
0 12
where z is the axial coordinate, is the reactor length, P is the
total pressure, D
ax
is the effective axial dispersion coefcient, r
m
is
the internal radius of the membrane, A
R
is the cross-sectional area
of the retentate chamber, V
R
is the volume of the reaction
chamber, e
b
is the void fraction of the catalyst bed, J is the ux
through the membrane, W
cat
is the mass of catalyst bed, and d
p
is
the catalyst particle diameter. v
i
is the stoichiometric coefcient
of species i, taken negative for the reactants, positive for the
reaction products, and null for components that do not take part
in the reaction.
The viscosity of the gas mixture, m
g
, was obtained by the Wilke
method (Poling et al., 2004). The respective density, r
g
, was
calculated by the virial equation, with the coefcients taken
from Smith et al. (1996). The variation of the effective axial
dispersion coefcient, D
ax
, for the range of conditions tested was
negligible (Perry and Green, 1999); so, the average value of
5.4010
5
m
2
s
1
was used.
Permeate sidePartial mass balance:
d
dz
u
P
p
P
i
_ _
f
2pr
m
A
P
RTJ
i
0 13
Total mass balance:
P
P
du
P
dz
f
2pr
m
A
P
RT

i
J
i
0 14
Boundary conditionsVacuum mode:
z 0: u
P
0 and p
P
i
P
P
15
Sweep gas mode (counter-current operation):
z : p
P
i
p
P,in
i
and u
P
u
P,in
16
where A
P
is the cross-sectional area of the permeate chamber and
f is dened as 1 or +1 according to the ow (co-current or
counter-current, respectively).
Membrane permeation equation: permeation of hydrogen
through a dense palladium membrane occurs via a solution-
diffusion mechanism (Dittmeyer et al., 2001). Richardsons
equation (based on Sieverts law) is typically used to describe
the overall permeation rate of hydrogen through the membrane
(Basile, 2008) and is used herein. Due to the high membrane
radius/membrane thickness ratio, the PdAg membrane is
approached to a at shape, despite its cylindrical geometry.
The transport properties of the PdAg membrane used in this
investigation were previously studied based on single permeation
measurements (for details, see (Mendes et al., 2010b)). The PdAg
membrane showed ideal selectivity towards H
2
, therefore J
i
0
except for i H
2
, Eqs. (8) and (9) and (13) and (14).
The H
2
permeating ux through the PdAg membrane is
expressed in terms of Richardsons equation (Eq. (17)) and
corrected for the temperature using an Arrhenius type depen-
dence (Eq. (18)):
J
H
2
z
L
H
2
d

p
R
H
2
z
_

p
P
H
2
z
_ _ _
17
Table 1
Calculated parameters for mechanistic-derived rate equations. Fitting errors of parameters are for 95% condence level (Mendes et al., 2010a).
Parameter T180200 1C (LT) T215300 1C (HT)
k
0
(for temperatures 180-2001Cmol g
1
cat
h
1
Pa
2
;
for temperatures 215-3001Cmol g
1
cat
h
1
Pa
1
)
1.18870.000 1.84110
3
70.21010
3
E
k
(kJ mol
1
) 36.65870.000 6.71070.399
K
a,0
CO
Pa
1

2.28310
24
70.00010
24
K
a,0
H2O
Pa
1

1.95710
28
70.00010
28
K
a,0
CO2
Pa
1

5.41910
4
70.00210
4
6.34310
1
70.727 10
1a
K
a,0
H2
Pa
1

2.34910
4
70.00010
4
DH
a
CO
kJ mol
1

45.99670.158
DH
a
H2O
kJ mol
1

79.96370.172
DH
a
CO2
kJ mol
1

16.47470.009 19.45970.402
DH
a
H2
kJ mol
1

13.27970.192
a
This value is dimensionless.
D. Mendes et al. / Chemical Engineering Science 66 (2011) 23562367 2360
L
H
2
L
0
H
2
exp
E
p
RT
_ _
18
where L
H
2
is the membrane permeability towards hydrogen, d is
the membrane thickness, E
p
is the permeation activation energy
and L
0
H
2
is the pre-exponential factor.
Gaseous species such as H
2
O, CO, CO
2
, and N
2
inevitably co-
exist in the WGS reaction process, affecting the hydrogen per-
meation of the Pd-based membranes, though without affecting
the H
2
selectivity (Barbieri et al., 2008; Unemoto et al., 2007). In
the present MR model, the hindrance effect due to the presence of
CO was taken into account using an extended equation, pre-
viously proposed by Barbieri et al. (2008). By including a correc-
tion factor, the authors arrived to a modied equation, named
SievertsLangmuir formulation:
J
H2
z 1c
K
p
CO
p
CO
1K
p
CO
p
CO
_ _
L
H
2
d
_ _

p
R
H
2
z
_

p
P
H
2
z
_ _ _
19
c and K
p
CO
are adjustable parameters and were obtained by tting
the H
2
recovery from the theoretical model to the respective
experimental results, as described below (Section 4.1). In this
equation, the term K
p
CO
p
CO
=1K
p
CO
p
CO
denes the fraction of the
membrane surface covered by adsorbed CO. The proportionality
coefcient, c, accounts for the dimensionless reduction of the
permeable area hindered by CO molecules. This parameter
depends only on the temperature, while K
p
CO
depends on the
temperature and CO pressure.
3.2. Dimensionless equations
The model variables were made dimensionless with respect to
the feed conditions (u
F
), to hydrogen species L
H
2
and to the
reactor length . The reference pressure was considered 100 kPa
and the reference temperature 273 K. Changing for dimensionless
variables and introducing suitable dimensionless parameters,
Eqs. (3a)(16) and (19) become as follows:
R
LT

exp g
k,LT
1
1
T

_ _ _ _
p

CO
p

H2O

CO
2
p

H
2
Ke
_ _
1K
a,0,LT
CO2
exp
DH
a,LT
CO
2
RT
_ _
P
ref
p

CO2
K
a,0,LT
H2
exp
DH
a,LT
H
2
RT
_ _
P
ref
p

H2
_ _
2
20
R
HT

exp g
k,HT
1
1
T

_ _ _ _
p

CO
p

H
2
O

CO
2
p

H
2
Ke
_ _
p

CO
1K
a,0,LT
CO
2
exp
DH
a,LT
CO
2
RT
_ _
p

CO
2
p

CO
_ _ 21
d
dx
u
R
p
R
i
_ _

1
Pe
d
dx
P
R
d
dx
p
R
i
P
R
_ _ _ _
GT

i
DaT

n
i
R

0 22
d
dx
u
R
P
R
GT

i
J

i
DaT

i
n
i
R

0 23
dP
R
dx
am

g
u
R
br

g
u
R
9u
R
9 0 24
x 0:
1
Pe
d
dx
p
R
i
P
R
_ _
u
R,in
p
R
i
p
R,in
i
_ _
P
R
,
u
R
u
R,in
and P
R
P
R,in
25
x 1:
d
dx
p
R
i
P
R
_ _
0 26
d
dx
u
P
p
P
i
_ _
f GsT

i
0 27
d
dx
u
P
P
P
_ _
f GsT

i
J

i
0 28
Vacuum mode:
x 0: p
P
i
P
P
and u
P
0 29
Counter-current mode:
x 1: p
P
i
p
P,in
i
and u
P
u
P,in
30
J

H
2
x 1c
K
p
CO
P
ref
p
R
CO
1K
p
CO
P
ref
p
R
CO
_ _
exp g
p
1
1
T

_ _ _ _
_ _

p
R
H
2
x
_

p
P
H
2
x
_ _ _
31
where g
k,LT
E
k,LT
=RT
ref
, g
k,HT
E
k,HT
=RT
ref
, g
p
E
p
=RT
ref
, p

i
p
i
=
P
ref
, P

P=P
ref
, u

u=u
ref
, L

H
2
1, x z=, a 150 1e
b

2
m
ref
u
ref
=e
b

3
d
p

2
P
ref
, b 1:751e
b
M
ref
u
ref

2
=d
p
e
b

3
RT
ref
, s
e
b
A
R
=A
P
, Pe u
ref
=D
ax
, GA
m
RT
ref

P
ref
_
L
ref
T
ref
= e
b
u
ref
A
R
. The
superscript
*
stands for dimensionless variables, and the subscript
ref stands for reference component or conditions. Pe is the Pe clet
number for mass transfer, G is a dimensionless contact time (ratio
between the characteristic feed ow time and the characteristic
permeance time of the reference component). The Damk ohler
number (ratio between the reaction rate at the reference tem-
perature and the feed ow rate to the reactor) was dened
independently for each of the kinetic models, depending on the
temperature range. For the lower temperature range, Da W
cat
R
T
ref
k
LT
T
ref
P
ref

2
=e
b
u
ref
A
R
, while Da W
cat
RT
ref
k
HT
T
ref
P
ref
=e
b
u
ref
A
R
for the higher temperature range. The remaining symbols are
reported in the nomenclature.
3.3. Numerical solution strategy
To simulate the WGS membrane reactor, it is necessary to
solve Eqs. (22)(24), and (27) and (28) with the respective
boundary conditions. These equations were transformed into
pseudo-transient ones, by adding a time derivative term to their
right-hand side, avoiding so numerical instability problems
(Sousa et al., 2001). The partial differential equations were
spatially discretized using the nite volumes method (Sa et al.,
2009) and the resulting ODEs were integrated in time by LSODA, a
package written by Petzold and Hindmarsh (1997), until an error
criterion was achieved (time derivative of each dependent vari-
able and for each of the spatial coordinate smaller than
110
12
). All physical properties of the gas mixture (m
g
and
r
g
) were evaluated at local conditions.
4. Results and discussion
4.1. Dimensional analysis: model validation
In this section, the effect of temperature, feed space velocity and
reaction pressure on the performance of the MR operating in
vacuum and sweep gas modes was studied. Moreover, for all the
operating conditions considered, the catalyst activity (evaluated
based on CO conversion) and the membrane separation ability
(evaluated in terms of H
2
recovery) were quantied and the results
compared with the ones from the proposed model. The experimen-
tal conditions and the simulation variables are shown in Table 2.
As described in Section 3.1, c and K
p
CO
are adjustable para-
meters, obtained by tting the H
2
recovery from the theoretical
model to the respective experimental results. The adjusted values of
c and K
p
CO
were in the ranges 0.61.0 and (0.51.5) 10
4
Pa
1
,
respectively, considering the values for all temperatures and ow
D. Mendes et al. / Chemical Engineering Science 66 (2011) 23562367 2361
rates. Similar results were reported by Mejdell et al. (2009). As
can be realized from Fig. 3A, the adherence of the simulated
hydrogen recovery to the experimental results is very good. The
comparison between the experimental and the calculated results
for the CO conversion is shown in Fig. 3B. Again, the model shows
a good agreement with the experimental data, with a few
exceptions at the lowest carbon monoxide conversions, obtained
for temperatures of 200 1C (remember that the model was tuned
to t the hydrogen recovery and the conversion was calculated
thereafter).
Figs. 4 and 5 show the inuence of the reaction temperature
and the feed space velocity (feed ow rate) on the MR perfor-
mance, either in terms of experimental data or in terms of
simulated results. The thermodynamic equilibrium conversion
of CO, X
e
, based on the inlet reformate gas composition, is also
included to show the conversion enhancement that is possible to
attain with the membrane reactor comparatively to the max-
imum value possible to obtain in a packed bed reactor, the
thermodynamic equilibrium value. Each gure refers to a differ-
ent operation mode for extraction of the permeated hydrogen:
sweep gas mode in Fig. 4 (Q
sweepN
2

1:0 L
N
min
1
, P
P
1.0 bar,
P
F
2.0 bar) and vacuum mode in Fig. 5 (P
P
$30 mbar,
P
F
1.1 bar).
Globally, both Figs. 4 and 5 show that the model describes
quite well the trend of the CO conversion and H
2
recovery as a
function of the temperature, for each feed ow rate. For any of the
Experimental Hydrogen Recovery, Re
H
2
S
i
m
u
l
a
t
e
d

H
y
d
r
o
g
e
n

R
e
c
o
v
e
r
y
,

R
e
H
2
0.0
0.0
0.2
0.4
0.6
0.8
1.0
MR - Sweep Gas Mode
MR - Vacuum Mode
Experimental CO Conversion, X
CO
S
i
m
u
l
a
t
e
d

C
O

C
o
n
v
e
r
s
i
o
n
,

X
C
O
0.75
0.75
0.80
0.85
0.90
0.95
1.00
PBR
MR - Vacuum Mode
MR - Sweep Gas Mode
0.2 0.4 0.6 0.8 1.0 0.80 0.85 0.90 0.95 1.00
Fig. 3. Parity plots of calculated and experimental results for H
2
recovery (A) and CO conversion (B).
200
C
O

C
o
n
v
e
r
s
i
o
n
,

X
C
O
0.80
0.85
0.90
0.95
1.00
Temperature, T / C
200
H
y
d
r
o
g
e
n

R
e
c
o
v
e
r
y
,

R
e
H
2
0.0
0.2
0.4
0.6
0.8
1.0
220 240 260 280 300 220 240 260 280 300
Temperature, T / C
Fig. 4. Effect of the reaction temperature and feed gas space velocity on the COconversion (A) and H
2
recovery (B) as a function of the reaction temperature in the MR, operating in
counter-current mode. Qsweep 1:0 L
N
min
1
, P
F
2:0 bar, and P
P
1:0 bar. Error bars are based on t-student distribution and 95% condence limit (with 3 replicates).
Table 2
Experimental conditions for the PdAg MR runs.
Variable Value/range Variable Value/range
T 200300 1C r
m
0.50 cm
P
F
12 bar r
shell
1.75 cm
P
P
0.0301 bar E
p
10.72 kJ mol
1
W
cat
1.5 g L
0
H2
5.44510
8
mol m m
2
s
1
Pa
0.5
GHSV 120010,800 L
N
kg
cat
1
h
1
d 50 mm
Q
sweep
1000 mL
N
min
1
d
p
300 mm
5 cm e
b
0.40
D. Mendes et al. / Chemical Engineering Science 66 (2011) 23562367 2362
operating temperatures, both CO conversion and H
2
recovery
increase with the residence time, that is, they decrease with an
increase of the gas hourly space velocityGHSV (ratio between
total feed ow rate and catalyst mass). However, the agreement
between the experimental conversion of CO and the simulated
values for the vacuum operation mode is poorer than for the
sweep gas mode operation. This may be related to the difculty in
controlling accurately the vacuum pressure in the permeate side
of the membrane reactor.
Figs. 4A and 5A also show that the CO conversion attained in
the MR may surpass the thermodynamic equilibrium conversion.
This conversion enhancement is achieved at the lower feed ow
rates for the lower temperatures, but as the temperature
increases, such conversion enhancement is attained for higher
and higher feed ow rates. For the lower temperatures, the
thermodynamic equilibrium conversion of CO is high, due to the
exothermal nature of the reaction, but the effective conversion is
low, because of the relatively slow reaction rate, and, conse-
quently, the amount of H
2
produced is also small. Besides, the
permeation of H
2
is also not favoured in this temperature region,
so the equilibrium shift promoted by the hydrogen removal from
the reaction medium is small. As a result, such enhancement is
achieved only for low feed ow rates (high residence times). As
the temperature increases, the equilibrium shift promoted by the
hydrogen removal from the reaction medium becomes signicant
for higher and higher feed ow rates, as a result of two main
factors. On one hand, the increase of the reaction rate with the
temperature leads to an increase of the CO conversion and,
consequently, to a corresponding change of the H
2
concentration.
On the other hand, the permeation rate increases also with the
temperature, at least in a range of temperatures (see the next
paragraph) leading so to a greater capacity of removing H
2
from
the reaction medium. From the balance of these two factors (and
also because the thermodynamic limit is lower at higher tem-
peratures), results that the conversion enhancement is attained
for increasingly higher feed ow rates. A similar effect has been
recently reported by Tang et al. (2010), although the membranes
and operating temperature ranges are different.
Figs. 4B and 5B show also that H
2
recovery may tend to a
plateau, for certain operating conditions. This behavior can be
inferred from Fig. 4B for low feed ow rates, but it is clear
in Fig. 5B for all the feed ow rates. This hydrogen recovery
plateau (or even decrease, if we consider also the simulation
results) is attained for the high temperature region considered in
the study and results from the combined effect of reaction and
permeation rates.
Assuming that only the chemical reaction (no permeation)
takes place, the conversion of CO increases with the temperature
in the region below the thermodynamic equilibrium value, for a
given feed ow rate. However, for conversions close to the
thermodynamic equilibrium, the CO conversion should decrease
with the temperature since the thermodynamic equilibrium also
decreases. It can then be concluded that the hydrogen concentra-
tion as a function of the temperature reaches a maximum, for a
given feed ow rate. For a membrane reactor, the hydrogen is
removed from the reaction medium as it is being formed. For a
temperature range where the retentate hydrogen concentration
increases due to the chemical reaction, the hydrogen permeation
should also increase since the driving force increases, as well as
the membrane permeability (cf. Eq. (18)). If the hydrogen per-
meation increases at a higher rate than its formation, the recovery
would increase with the temperature; otherwise, a decrease in
the recovery should be observed. In addition, after a certain
threshold temperature (normally around 280300 1C) the hydro-
gen production decreases with the temperature. For this region,
the membrane permeability increase with temperature might not
be enough to compensate the driving force decrease and, as a
result, the hydrogen permeation would decrease as well as the
recovery.
Fig. 6 shows the simulated axial composition proles in the
retentate side for each species (dry basis) in terms of the
concerning molar fraction, for GHSV1200 L
N
kg
cat
1
h
1
and for
the two operating modes of the MR, vacuum and sweep gas, both
with 2 bar in the feed. The experimental compositions obtained at
the exit of the reactor are also included.
The comparison between the predicted compositions and the
experimental results at the exit of the reactor puts also in
evidence the validity of the model (cf. Fig. 6). As it can be realized
from Fig. 6, there are clearly two different regions inside the
reaction medium. In the rst one, for about 10% of the initial
length of the reactor, the molar fraction of CO decreases rapidly,
while the molar fraction of H
2
increases slightly. Since the
reaction temperature is high in both cases (T300 1C), the
reaction kinetics and H
2
permeation are favoured. As the reaction
condition for the feed mixture is far away from the thermody-
namic equilibrium condition, the chemical reaction shifts quickly
towards the reaction products, decreasing rapidly the CO con-
centration. However, the rate of H
2
permeation does not follow
the rate of production from the WGS reaction, and thus the
concentration of H
2
increases, though slowly. After this initial
region, the shift of the reaction condition towards the reaction
products depends only on the H
2
removal due to permeation.
Temperature, T / C
200
C
O

C
o
n
v
e
r
s
i
o
n
,

X
C
O
0.70
0.75
0.80
0.85
0.90
0.95
1.00
200
H
y
d
r
o
g
e
n

R
e
c
o
v
e
r
y
,

R
e
H
2
0.0
0.1
0.2
0.3
0.4
0.5
220 240 260 280 300 220 240 260 280 300
Temperature, T / C
Fig. 5. Effect of the reaction temperature and feed gas space velocity on the CO conversion (A) and H
2
recovery (B) in the MR, operating in vacuum mode. P
F
1:1 bar and
P
P
$30 mbar. Error bars are based on t-student distribution and 95% condence limit (with 3 replicates).
D. Mendes et al. / Chemical Engineering Science 66 (2011) 23562367 2363
As a consequence, the concentrations for both reactants (water
not presented) decrease slowly and in a more or less linear way.
The concentration proles of CO
2
and N
2
exhibit an increasing
trend along the reactor axis, because the mixture becomes richer
in the non-permeating species and, moreover, CO
2
is a reaction
product. The reactor can then be divided into a region of chemical
control and a region of permeation control, this accounting for the
last nearly 90% of the reactor length, for these operating
conditions.
The membrane reactor can only withstand a pressure differ-
ence of about 2 bar. Higher feed pressures can only be applied if a
sweep gas is used at a pressure such that the pressure threshold
difference (2 bar) is not exceeded. High feed pressures coupled
with high sweep ow rates results in a high H
2
partial pressure
difference between the two chambers of the membrane reactor.
On the other hand, operating at high feed pressures results in an
enhancement of CO conversion as a consequence of the reaction
rate increase. Following, lower contents of CO were obtained at
the exit of the retentate chamber when the sweep mode was used
(cf. Fig. 6(A) and (B)); the content of H
2
also decreased and its
recovery improved because of the higher driving force for hydro-
gen permeation.
4.2. Inuence of the Damk ohler number and contact time
From the results presented above, it is clear that there are
regions where the CO conversion is almost complete, as well as
the H
2
recovery is maximum. However, it would be of interest to
dene the parametric regions where such variables could be
optimized. In order to do so, it is presented a simulation result
of the CO conversion and H
2
recovery as a function of the two
model dimensionless parameters, Damk ohler number (Da) and
contact time (G)Figs. 7 and 8.
The analysis of the MR operating with vacuum pumping and
sweep gas mode is briey accessed in terms of dimensionless
parameters. The values for c and K
p
CO
were assumed to be
Fig. 7. Effect of Da and G on the CO conversion (A) and H
2
recovery (B) in the MR operating in vacuum mode. T300 1C, P
F
1:1 bar and P
P
$30 mbar. The white lines
describe the parametric region of the operating conditions used in this work.
x
0.2
C
O

f
r
a
c
t
i
o
n

(
m
o
l
/
m
o
l
)
0.00
0.01
0.02
0.03
0.04
H
N
CO
CO
x
0.0
H
2
,

C
O
2

a
n
d

N
2

f
r
a
c
t
i
o
n

(
m
o
l
/
m
o
l
)
0.0
0.1
0.2
0.3
0.4
0.5
0.6
H
N
CO
CO
0.2 0.4 0.6 0.8 1.0 0.4 0.6 0.8 1.0
Fig. 6. Simulated proles on the reaction side as a function of the dimensionless reactor length: vacuum mode, P
F
2:0 bar and P
P
$30 mbar (A) and sweep gas mode,
P
F
2:0 bar, P
P
1:0 bar and Q
sweep
1.0 L
N
min
1
(B). Other experimental conditions: T300 1C, GHSV1200 L
N
kg
cat
1
h
1
. Symbols represent the experimental molar
fraction (dry basis) of each species measured at the exit of the reactor.
D. Mendes et al. / Chemical Engineering Science 66 (2011) 23562367 2364
0.93 and 9.5310
5
Pa
1
, respectively, as reported by Mejdell
et al. (2009).
Fig. 7A shows that there is a region in the parametric space
contact time/Damk ohler number where an almost complete
conversion of CO can be achieved. This happens for Da4E7
and G4E0.4. For lower values of G (keeping the same Damk ohler),
the CO conversion decreases slightly, but for lower values of Da
(keeping the same contact time) the decrease is abrupt. Fig. 7B, on it
turns, shows that the H
2
recovery is maximized in the same
parametric region. The maximum value attained for vacuum opera-
tion was about 0.93, which is related to the hydrogen partial
pressure at the permeate side $30 mbar. For lower values of G
and in the same region of Da, the H
2
recovery decreases quickly,
with a still high CO conversion. This is a consequence of the quick
decrease of the stage cut, that is, the fraction of H
2
in the retentate
increases. Fig. 7B shows again that the H
2
recovery is much more
sensitive to the contact time than CO conversion.
The experimental results obtained for vacuum mode are
presented in Fig. 7 (white line). As it can be inferred, both CO
conversion and H
2
recovery can be improved increasing essen-
tially the contact time parameter (that is, decreasing the GHSV).
These results show that, for a given membrane reactor (with a
xed amount of catalystxed Damk ohler) and operating in
vacuum mode, the CO conversion is mostly controlled by the
feed ow rate, while the H
2
recovery depends also strongly on the
hydrogen permeation rate (which depends on the hydrogen
partial pressure at the permeate side and total pressure at feed-
side). If the membrane reactor operates in sweep mode and with
2 bar in the feed, there is a region of the parametric space contact
time/Damk ohler number where it is possible to achieve almost
100% of CO conversion with almost 100% of H
2
recovery, as shown
in Fig. 8.
5. Conclusions
The present study deals with model analysis and experimental
assessment of a self-supported nger-like membrane reactor
particularly conceived for ultra-pure hydrogen production. WGS
reaction runs were carried out on a PdAg MR under a broad
range of operating conditions such as temperature (200300 1C)
and gas hourly space velocity (120010,800 L
N
kg
cat
1
h
1
), using
different operating modes (vacuum and sweep gas). The metallic
membrane permeability and the kinetics of the reaction were also
accessed experimentally, whose equations were introduced in the
phenomenological model. The simulation and experimental car-
bon monoxide conversion results showed a suitable agreement
for the MR working in both sweep gas and vacuum operating
modes. The comparison between the predicted compositions and
the experimental results, at the exit of the reactor, conrm also
the validity of the proposed MR model. Apart from two tting
parameters (related with the decline of permeability due to the
presence of CO in the reaction mixture), all other model para-
meters were determined from independent studies, namely reac-
tion kinetics and membrane permeability towards hydrogen.
The model was also used to simulate the performance of the
MR in a wide range of the parametric space, described by
Damk ohlers number and contact time. This allowed us to dene
the optimal operating regions in terms of CO conversion and H
2
recovery.
Nomenclature
A cross-sectional area (m
2
)
d
p
catalyst particle diameter (m
2
)
D
ax
axial dispersion coefcient (m
2
s
1
)
Da Damk ohler number
E
k
activation energy for the WGS reaction (kJ mol
1
)
E
p
activation energy for hydrogen permeation (kJ mol
1
)
f variable related with the operation mode (f 1 for co-
current; f 1 for counter-current)
H
a
enthalpy of adsorption (J mol
1
)
J ux through the membrane (mol m
2
s
1
)
k rate constant for the WGS reaction (lower temperature
regime mol g
1
cat
h
1
Pa
2
; higher temperature regime
mol g
1
cat
h
1
Pa
1
)
Fig. 8. Effect of Da and G on the CO conversion (A) and H
2
recovery (B) in the MR operating in sweep gas (counter-currently) mode. T300 1C, P
F
2:0 bar, P
P
1:0 bar,
and Q
sweep
1 L
N
min
1
. The white line describes the parametric region of the operating conditions used in this work.
D. Mendes et al. / Chemical Engineering Science 66 (2011) 23562367 2365
k
0
pre-exponential factor of the rate constant (lower tem-
perature regime mol g
1
cat
h
1
Pa
2
; higher temperature
regime mol g
1
cat
h
1
Pa
1
)
K
e
equilibrium constant for the WGS reaction
K
p
CO
equilibrium adsorption constant of CO for Sieverts
Langmuir formulation, adjustable parameter (Eq. (19))
(Pa
1
)
K
a
i
equilibrium adsorption constant of species i for the
reaction rate equation (lower temperature regime
Pa
1
; higher temperature regime dimensionless)
K
a,0
i
pre-exponential equilibrium adsorption constant of
species i (lower temperature regime Pa
1
; higher
temperature regime dimensionless)
reactors length (m)
L
H
2
hydrogen permeability (mol m m
2
s
1
Pa
0.5
)
L
0
H
2
pre-exponential factor for hydrogen permeation
(mol m m
2
s
1
Pa
0.5
)
M molar mass (g mol
1
)
p partial pressure (Pa or bar)
P total pressure (Pa or bar)
Pe Pe clet number for mass transfer
r internal radius (m)
R ideal gas constant (8.314 J mol
1
K
1
)
R rate of consumption or formation (mol g
1
cat
h
1
)
Re
H
2
hydrogen recovery
T temperature (K or 1C)
u interstitial velocity (m s
1
)
V
R
volume of the retentate chamber (m
3
)
W
cat
mass of catalyst (g)
x dimensionless axial coordinate
X
CO
conversion of carbon monoxide
X
e
thermodynamic equilibrium conversion
z axial coordinate (m)
Subscripts
i species involved in the reaction experiments (CO, H
2
O,
CO
2
, H
2
, or N
2
)
ref reference component (H
2
) or conditions
Superscripts
n dimensionless variable
in inlet of the reactor
k relative to the reaction kinetics
m membrane
out outlet of the reactor
p relative to the H
2
permeation
F feed-side
P permeate-side
R retentate side
Greek letters
a Ergun equation coefcient (Eq. (24))
b Ergun equation coefcient (Eq. (24))
d PdAg membrane thickness (m)
e
b
void bed fraction
g Arrhenius number
G dimensionless contact time
m
g
(dynamic) gas mixture viscosity (kg m
1
s
1
)
r
g
gas mixture density (kg m
3
)
s dimensionless parameter: ratio between the cross-
sectional areas of the retentate and the permeate
chambers
n
i
stoichiometric coefcient of species i in the WGS
reaction
C adjustable parameter (Eq. (19))
Acronyms
GHSV gas hourly space velocity
HT higher temperature regime
LT lower temperature regime
MR membrane reactor
PBR packed-bed reactor
PEMFC polymer electrolyte membrane fuel cell
WGS water-gas shift
Acknowledgments
Diogo Mendes and Sandra Sa are grateful to the Portuguese
Foundation for Science and Technology (FCT) for their doctoral
Grants (Ref. nos. SFRH/BD/22463/2005 and SFRH/BD/30385/2006,
respectively). The authors also acknowledge nancing from FCT
through the projects PTDC/EQU/ERQ/66045/2006 and PTDC/EQU-
EQU/71617/2006.
References
Adrover, M.E., Lopez, E., Borio, D.O., Pedernera, M.N., 2009. Theoretical study of a
membrane reactor for the water gas shift reaction under nonisothermal
conditions. A.I.Ch.E. J. 55 (12), 32063213.
Aresta, M., Dibenedetto, A., 2007. Utilisation of CO
2
as a chemical feedstock:
opportunities and challenges. Dalton Trans. 28, 29752992.
Ayastuy, J.L., Gutierrez-Ortiz, M.A., Gonzalez-Marcos, J.A., Aranzabal, A., Gonzalez-
Velasco, J.R., 2005. Kinetics of the low-temperature WGS reaction over a CuO/
ZnO/Al
2
O
3
catalyst. Ind. Eng. Chem. Res. 44 (1), 4150.
Barbieri, G., Scura, F., Lentini, F., De Luca, G., Drioli, E., 2008. A novel model
equation for the permeation of hydrogen in mixture with carbon monoxide
through PdAg membranes. Sep. Purif. Technol. 61 (2), 217224.
Barbir, F., 2005. PEM Fuel Cells. Theory and Practice. Elsevier Academic Press,
New York.
Basile, A., 2008. Hydrogen production using Pd-based membrane reactors for fuel
cells. Top. Catal. 51 (14), 107122.
Basile, A., Chiappetta, G., Tosti, S., Violante, V., 2001. Experimental and simulation
of both Pd and Pd/Ag for a water gas shift membrane reactor. Sep. Purif.
Technol. 25 (13), 549571.
Brunetti, A., Caravella, A., Barbieri, G., Drioli, E., 2007. Simulation study of
water gas shift reaction in a membrane reactor. J. Membr. Sci. 306 (12),
329340.
Criscuoli, A., Basile, A., Drioli, E., 2000. An analysis of the performance of
membrane reactors for the water-gas shift reaction using gas feed mixtures.
Catal. Today 56 (13), 5364.
Dittmeyer, R., Hollein, V., Daub, K., 2001. Membrane reactors for hydrogenation
and dehydrogenation processes based on supported palladium. J. Mol. Catal.
A: Chem. 173 (12), 135184.
Froment, G.F., Bischoff, K.B., 1990. Chemical Reactor Analysis and Design, 2 ed.
John Wiley & Sons, New York.
Gosiewski, K., Warmuzinskia, K., Tanczyka, M., 2010. Mathematical simulation of
WGS membrane reactor for gas from coal gasication. Catal. Today.
doi:10.1016/j.cattod.2010.02.031.
ISO, 2008. Hydrogen fuel product specication Part 2: proton exchange
membrane (PEM) fuel cell applications for road vehicles, ISO/TS 14687-2.
International Organization for Standardization.
Koryabkina, N.A., Phatak, A.A., Ruettinger, W.F., Farrauto, R.J., Ribeiro, F.H., 2003.
Determination of kinetic parameters for the water-gas shift reaction on copper
catalysts under realistic conditions for fuel cell applications. J. Catal. 217 (1),
233239.
Koukou, M.K., Papayannakos, N., Markatos, N.C., 1996. Dispersion effects on
membrane reactor performance. A.I.Ch.E. J. 42 (9), 26072615.
Markatos, N.C., Vogiatzis, E., Koukou, M.K., Papayannakos, N., 2005. Membrane
reactor modellinga comparative study to evaluate the role of combined
mass and heat dispersion in large-scale adiabatic membrane modules. Chem.
Eng. Res. Des. 83 (A10), 11711178.
Mejdell, A.L., Jondahl, M., Peters, T.A., Bredesen, R., Venvik, H.J., 2009. Effects of CO
and CO
2
on hydrogen permeation through a similar to 3 um Pd/Ag 23 wt%
membrane employed in a microchannel membrane conguration. Sep. Purif.
Technol. 68 (2), 178184.
D. Mendes et al. / Chemical Engineering Science 66 (2011) 23562367 2366
Mendes, D., Chibante, V., Mendes, A., Madeira, L.M., 2010a. Determination of the
low-temperature water-gas shift reaction kinetics using a Cubased catalyst.
Ind. Eng. Chem. Res. 49 (22), 1126911279.
Mendes, D., Chibante, V., Zheng, J.-M., Tosti, S., Borgognoni, F., Mendes, A., Madeira,
L.M., 2010b. Enhancing the production of hydrogen via water-gas shift
reaction using Pd-based membrane reactors. Int. J. Hydrogen Energy 35,
1259612608.
Mendes, D., Garcia, H., Silva, V.B., Mendes, A., Madeira, L.M., 2009. Comparison of
nanosized gold-based and copper-based catalysts for the low-temperature
water-gas shift reaction. Ind. Eng. Chem. Res. 48 (1), 430439.
Moe, J.M., 1962. Design of water-gas shift reactors. Chem. Eng. Prog. 58 (3), 3336.
Olah, G.A., Goeppert, A., Prakash, G.K.S., 2006. Beyond Oil and Gas: The Methanol
Economy. Wiley-VCH, Germany.
Pasel, J., Cremer, P., Wegner, B., Peters, R., Stolten, D., 2004. Combination of
autothermal reforming with water-gas-shift reactionsmall-scale testing of
different water-gas-shift catalysts. J. Power Sources 126, 112118.
Perinline, H.W., Luebke, D.R., Jones, K.L., Myers, C.R., Morsi, B.I., Heintz, Y.J.,
Ilconich, J.B., 2008. Progress in carbon dioxide capture and separation research
for gasication-based power generation point sources. Fuel Process. Technol.
89 (9), 897907.
Perry, R.H., Green, D.W., 1999. Perrys Chemical Engineers Handbook, 7 ed.
McGraw-Hill, New York, USA.
Petzold, L.R., Hindmarsh, A.C., 1997. LSODA, Computing and Mathematics Research
Division. Lawrence Livermore National Laboratory.
Poling, B.E., Prausnitz, J.M., OConnell, J.P., 2004. The Properties of Gases and
Liquids, 5 ed. McGraw-Hill, New York.
Ratnasamy, C., Wagner, J., 2009. Water gas shift catalysis. Catal. Rev.-Sci. Eng. 51
(3), 325440.
Sa , S., Silva, H., Sousa, J.M., Mendes, A., 2009. Hydrogen production by methanol
steam reforming in a membrane reactor: palladium vs carbon molecular sieve
membranes. J. Membr. Sci. 339 (12), 160170.
Salemme, L., Menna, L., Simeone, M., 2010. Thermodynamic analysis of ethanol
processorsPEM fuel cell systems. Int. J. Hydrogen Energy 35, 34803489.
Smith, J.M., Ness, H.C.V., Abbott, M., 1996. Introduction to Chemical Engineering
Thermodynamics. McGraw-Hill, Singapore.
Sousa, J.M., Cruz, P., Mendes, A., 2001. Modeling a catalytic polymeric non-porous
membrane reactor. J. Membr. Sci. 181, 241252.
Srinivasan, S., 2006. Fuel Cells: From Fundamentals to Applications, Chapter 9:
Status of Fuel Cell Technologies. Springer.
Tang, Z., Kim, S.-J., Reddy, G.K., Dong, J., Smirniotis, P., 2010. Modied zeolite
membrane reactor for high temperature water gas shift reaction. J. Membr. Sci.
354, 114122.
Tiemersma, T.P., Patil, C.S., Annaland, M.V., Kuipers, J.A.M., 2006. Modelling of
packed bed membrane reactors for autothermal production of ultrapure
hydrogen. Chem. Eng. Sci. 61 (5), 16021616.
Tosti, S., Basile, A., Bettinali, L., Borgognoni, F., Chiaravalloti, F., Gallucci, F., 2006. Long-
term tests of PdAg thin wall permeator tube. J. Membr. Sci. 284 (12), 393397.
Tosti, S., Bettinali, L., 2004. Diffusion bonding of PdAg rolled membranes. J. Mater.
Sci. 39 (9), 30413046.
Tosti, S., Bettinali, L., Lecci, D., Violante, V., Marini, F., 2001. Method of bonding thin
foils made of metal alloys selectively permeable to hydrogen, particularly
providing membrane devices, and apparatus for carrying out the same, Eur.
Patent EP 1184125.
Uemiya, S., Sato, N., Ando, H., Kikuchi, E., 1991. The Water Gas Shift reaction sssisted
by a palladium membrane reactor. Ind. Eng. Chem. Res. 30 (3), 585589.
Unemoto, A., Kaimai, A., Sato, K., Otake, T., Yashiro, K., Mizusaki, J., Kawada, T.,
Tsuneki, T., Shirasaki, Y., Yasuda, I., 2007. The effect of co-existing gases from
the process of steam reforming reaction on hydrogen permeability of palla-
dium alloy membrane at high temperatures. Int. J. Hydrogen Energy 32 (14),
28812887.
Zhang, J., 2008. PEM Fuel Cell Electrocatalysts and Catalyst Layers: Fundamentals
and Applications. Springer.
D. Mendes et al. / Chemical Engineering Science 66 (2011) 23562367 2367

You might also like