ELSEVIER Applied Catalysis A: General 119 ( 1994) 305-325

Methane steam reforming in asymmetric Pd- and

Pd-Ag/porous SS membrane reactors
Jun Shu, Bernard P.A. Grandjean, Serge Kaliaguine *
Department of Chemical Engineering and CERPIC, Lava1 University, Ste-Foy, Quebec GlK 7P4, Canada
Received 31 May 1994; revised 2 August 1994; accepted 2 August 1994
Abstract
This work is devoted to applying electrolessly deposited Pd- and Pd-Ag/porous stainless steel
composite membranes in methane steam reforming. The methane conversion is significantly enhanced
by the partial removal of hydrogen from the reaction location as a result of diffusion through the Pd-
based membranes. For example, at a total pressure of 136 kPa, a temperature of 5OOC, a molar steam-
to-methane ratio of 3, and in the presence of a commercial Ni/A1203 catalyst together with continuous
pumping on the permeation side, a methane conversion twice as high as that in a non-membrane
reactor was reached by using a Pd/SS membrane. These effects were examined under a variety of
experimental conditions. A computer model of the membrane reactor was also developed to predict
the effects of membrane separation on methane conversion.
Keywords: Hydrogen permeation; Kinetic pcrtneation model; Membrane reactor; Methane stcatu reforming;
Palladium film: Palladium-silver film
1. Introduction
Methane steam reforming is one of the most important chemical processes for
the production of hydrogen or synthesis gas [ 11. This technology has been proved
more economic than other processes such as coal vaporization, hydrocarbon partial
oxidation, water electrolysis to produce hydrogen [ 21 for the reduction of iron
ores, for the use in fuel cells and for hydrocracking, etc. In recent years, the abundant
availability of natural gas and the increasing demand of hydrogen have led to more
and more interest to develop this process further [ 3,4].
Methane steam reforming involves two reversible reactions: the reforming ( 1)
and water-gas shift reaction (2) :
* Corresponding author. E-mail: kaliagui@gch.ulaval.ca, tel. ( + 1-418) 6562708, fax. ( + 1-418) 6567763.
0926-860X/94/$07.00 0 1994 Elsevier Science B.V. All rights reserved
SSD10926-860X(94)00178-2



306 J . Shu et al. /Applied Catalysis A: General 119 (1994) 305-325
CH,+H,O+CO+3H, A@!!,= -206kJ/mol
(1)
CO + HZ0 + COZ + H2 AH& = + 41 k.I/mol
(2)
CI-I, + 2Hz0 + COZ + 4Hz A&,, = - 165 kJ/mol
(3)
The reforming reaction ( 1) is endothermic and involves an increase in the
number of gaseous molecules so that it is thermodynamically favored by high
temperature and low pressure. In the conventional technology, this reaction is
conducted on supported nickel catalysts in multitubular reactors operated at tem-
peratures up to 85OC, pressures ranging from 1.6 to 4.1 MPa and steam-to-methane
ratios between 2 and 4. The elevated pressure is used to improve the energy
efficiency of the overall process [ 51. The endothermic reaction heat is supplied by
burning fuel in a radiant furnace. Methane conversion is usually around 78%,
limited by the thermodynamic equilibrium.
The development of palladium-based membrane separation processes [ 6-91 has
opened up a new possibility to enhance the methane steam reforming conversion.
With tbe continuous withdrawal of hydrogen by a palladium-based membrane from
the reaction location, the equilibrium of the reforming reaction ( 1) could be sig-
nificantly shifted toward the right hand side, resulting in an increase of the methane
conversion. In the meantime, high purity hydrogen could be obtained directly from
the permeation side of the membrane reactor. Oertel et al. [lo] performed a
calculation for a modified methane steam reforming process with integrated hydro-
gen separation. They demonstrated that almost 90% of methane conversion could
be achieved below 850C by using a palladium membrane with a thickness of 50
pm. They also fitted a Pd-membrane in a reformer to separate the produced hydro-
gen and obtained a methane conversion as high as 96% [ 10,111. The behavior of
a palladium membrane reactor used in methane steam reforming was simulated by
Prokopiev et al. [ 121. A computer simulation on methane steam reforming in a
fluidized bed membrane reactor was also performed by Adris et al. [ 131. Near
complete methane conversion was expected.
However, the high cost of palladium and the relatively low hydrogen permeation
capacity do not allow massive palladium sheets to be commercially used on a large
scale. Certain efforts have been reported in the literature either in the selection of
new membrane materials or in the improvement of membrane morphology. Besides
the exploration using catalytically modified microporous alumina membranes
[ 14,151, an alternative choice is to develop composite thin palladium membranes.
Kikuchi and coworkers [ 16-251 developed thin Pd and Pd-Ag membranes sup-
ported on either porous glass or porous A1203 ceramic substrates. These membranes
showed much higher hydrogen permeability than commercial palladium based
sheets. Using these membranes combined with a low-temperature reforming nickel
catalyst, these authors [20-241 showed that methane steam reforming can be
promoted far beyond the equilibrium. The problem with these substrates lies in
their relatively poor mechanical and/or thermal properties. The present work deals



.I. Shu et al. /Applied Catalysis A: General 119 (1994) 305-325 307
with the preparation of thin Pd-based membranes supported on a porous stainless
steel (SS) substrate and their uses in performing methane steam reforming in a
membrane reactor. In order to further understand the experimental results, a com-
puter model will be described to give some general predictions of the membrane
reactor behavior.
2. Experimental
2. I. Membrane preparation
A porous seamless 3 16L stainless steel tube having an outside diameter of 0.95
cm and a nominal particle retention size of 0.2 pm was purchased from Mott
Metallurgical. The tube was cut to a length of 3.6 cm, and welded to a dense
stainless steel tube with the same diameter. The effective membrane area for hydro-
gen separation was about 10.7 cm*. This tube was then cleaned in an ultrasonic
bath of carbon tetrachloride for one hour. Asymmetric Pd or Pd-Ag membranes
were formed on the outer side of the porous tube by the electroless plating technique
described previously [ 261. Composition and structural information about the two
membranes are listed in Table 1. The impervious Pd/SS membrane was used as
deposited. The deposited Pd-Ag membrane was thermally treated over their Tam-
man temperature (550C) for five hours in the presence of hydrogen. This treatment
can lead to the formation of an impervious Pd-Ag alloy film [ 271.
2.2. Reaction testing
Methane steam reforming was carried out in the reaction system illustrated in
Fig. 1. The detailed design of the membrane reactor used is shown in Fig. 2. The
outer stainless steel tube had an I.D. of 1.7 cm and a length of 5.1 cm. The inner
tube was the membrane part just described. This reactor assembly could be easily
configured for the reaction test with or without membrane by a simple replacement
of the inner tube. The non-membrane reactor configuration could be made by using
a dense stainless steel inner tube. The shell side of the reactor served as the reaction
chamber and the tube side as the permeation chamber.
Most of the parts were arranged in a modified autoclave reaction apparatus
( BTRS-JrTM , Autoclave Engineers Group). The assembly had two heating systems
Table 1
List of Pd-based membranes used in methane steam reforming
Membrane
Pdlporous SS
Pd-Aglporous SS
Composition Geometric dimension O.D. X h (cm) Estimated film thickness (pm)
Pd 0.95 x 3.6 19.8
Pd-5.1 %Ag 0.95 x 3.6 10.3



308 J . Shu et al. /Applied Catalysis A: General 119 (1994) 305-325
4
LC-600
nl\rir
Y
q-
HZ0
vent
Plorrc*r
l-2(off) Puq~
l-S(on) Reaction
Fig. 1. Experimental apparatus for methane steam reforming.
with separate temperature controllers: a furnace and a surrounding reactor cabinet.
The reactor was fixed in the middle of the furnace to maintain a temperature
distribution as uniform as possible. The temperature was controlled by a Eurotherm
Fig. 2. Tubular membrane reactor design.



J . Shu et al. /Applied Catalysis A: General 119 (1994) 305-325 309
847 digital controller. The preheating cabinet allows mixers, switch valves and
transfer lines to be maintained at 200C without cold spots.
The dry reactant introduced to the reaction side was a mixture of methane with
5% helium. The inlet stream to the permeation side was pure helium used as a purge
gas. Another helium stream was also employed to sweep the reaction side during
start up and shut down of the reaction system. These gaseous streams were sepa-
rately controlled by an MKS type 1259C mass flow controller combined with an
MKS type 247C 4 channel readout. Before operation, the flow controller was
calibrated with He, HZ, and CH4 + S%He gases using a digital soap bubble flow-
meter (Wheaton Scientific). The reactant stream and the purge gas stream passed
separately through the reaction side and the permeation side in a cocurrent flow
mode (from bottom to top).
Deionized water was injected into the methane-5% helium mixture stream using
an LC-600 Shimadzu Liquid Chromatograph injector. This module was a solvent
delivery unit with a microvolume double-plunger reciprocating pump, which pro-
vided stable delivery with little flow pulsation compared with a conventional solvent
delivery system. Its accuracy of flow-rate was within + 2%.
To enhance the reforming conversion, a commercial 12%Ni/A1203 catalyst
(C 1 l-9-02) provided by United Catalysts was used. Before packing, catalyst pellets
were crushed to 18-30 mesh size. About 11 g of catalyst were packed inside the
reaction chamber of the reactor with glass fiber filling up small empty volumes of
both the inlet and the outlet. The catalyst completely covered the Pd-based
membrane surface. The permeation side contained no catalyst.
2.3. Operating procedures
Since the original steam reforming catalyst was provided in the form of nickel
oxide supported on alumina, an in-situ activation of the catalyst was necessary after
packing. This was done by first heating the catalyst bed to 300C in a helium
atmosphere to prevent the Pd-based membrane from developing cracks or pinholes,
followed by introducing a hydrogen stream to the reactor at atmospheric pressure.
Then the temperature was increased stepwise to 400C and 550C each step lasted
for one hour. After activation, the system was conditioned for either hydrogen
permeation measurement or steam reforming.
For methane steam reforming, the activated system was first heated to 300C in
a helium atmosphere. Once the conditions were achieved, a preheated reactant
mixture of steam and CI& + 5%He, in a steam-to-methane ratio of 3, was fed into
the catalyst bed through an air-actuated valve. Inert helium gas at a flow-rate of 40
ml ( STP) /min (abbreviated as SCCM) passed through the permeation side to purge
the permeated hydrogen. The steam reforming was conducted in the temperature
range of 300 to 550C and a typical pressure of 136 kPa. A CH, + 5%He gas flow-
rate of 42 SCCM was used corresponding to a gas hourly space velocity (GHSV)
of about 1067 h- . After condensation of the unreacted steam by passing the



310 J. Shu et al. /Applied Catalysis A: General 119 (1994) 305-325
reaction product through a water-ice mixture trap, the effluent was analyzed using
an on-line MS (SMD-I-100 Mass Spectrometer, VG Gas Analysis System) and a
TCD (HP 5890 Series II Gas Chromatograph) .
In the present case of a dense Pd-based membrane through which only hydrogen
was permeated, methane conversion was calculated using the following expression
based on a carbon balance:
X
cH.+=
G0.r + CCO2.r
c
CH4.r + G0.r + Go2.r
(4)
where the index r stands for reaction side.
3. Mathematical model
In comparison with analyses for common chemical reactors, the membrane
reactor analysis must consider the additional contribution of mass transfer through
the membrane, that is, the permeation rate should be involved along with the
reaction rate expression, To investigate the concentration profile of methane steam
reforming inside Pd-based membrane reactors, a kinetic permeation model was
considered. A schematic illustration of the model is shown in Fig. 3.
3. I. Basic assumptions
( 1) Isothermal and isobaric reaction conditions;
(2) Steady-state operation;
(3) Plug flow on both reaction and permeation sides;
(4) Intrinsic kinetics for methane steam reforming (MSR) and water-gas shift
(WCS) reactions;
(5) No boundary layer on membrane surfaces.
Fig. 3. Illustration of the kinetic permeation model.



J . Shu et al. /Applied Catalysis A: General 119 (1994) 305-325
3.2. Reaction rates
311
Xu and Froment [28] studied carefully the intrinsic MSR and WGS reaction
kinetics on a Ni/MgAl,O, catalyst in a classical tubular reactor. The developed
model was proved to be more general than previous literature models for nickel
catalysts [ 291. This model was adopted in the present study. The model, based on
a Langmuir-Hinshelwood reaction mechanism, contains 13 steps. The rate expres-
sions for reactions ( 1) to (3) were given as:
R, =h(pCH4pH20 -9) /DEN2
Pk: 1
R2 =~(pCOpHzO -y) /DEN2
2
k3
4
R3 =--&pm P&O -
2
py) /DEN2
3
DEN = 1+ Kco PCO +KHZ PHI +Kcm PCH~ +ho ~~201~~2
with the partial pressures given by the relations:
pcH4= (1 -&LA/~ PHzO= (m-XCH4 -&oz) /a
PCO= (XCH4-XCOz)/a PHz = @I? + 3xCH, + Go*
- YH)/(T
PC02 =&02/u
(T=( 1+m+p&+2XoHq-Yn)/Pr
(5)
(f-3)
(7)
(8)
where, P, is the pressure on the reaction side of the reactor; Xc, is the total methane
converston; Xco2 the methane conversion to CO,; YH is the permeated hydrogen
fraction with respect to the integrated hydrogen on the reaction side; and m the
steam-to-methane ratio. The meaning of p& will be discussed below.
To estimate the model parameters, a kinetic study of methane steam reforming
over C 1 l-9-02 Ni/Al,O, catalyst was performed in a Betty reactor [ 271. The study
indicated that the specific model developed by Xu and Froment [ 281 fitted essen-
tially our methane conversion results over Cl l-9-02 catalyst. Thus those model
parameters obtained by Xu and Froment were also adopted in the present study, as
listed in Table 2. Reaction rates for the disappearance (i.e., total conversion) of
Table 2
Kinetic input parameters [ 281
Paramete r Pre-exponential factor b Eor AH (kJ/mol )
k, (km01 MPa./kg h)
k2 (kmol/kg h MPa)
ka (km01 MPa./kg h)
f&W'-')
KH2 (MPaC)
KcH4 (ma-)
KHZ0
1.336. lOI 240.10
1.955 10 67.13
3.226. lOI 243.90
8.23. 1O-4 - 70.65
6.12.1OF - 82.90
6.65 10-S - 38.28
1.77.1@ 88.68



312 J . Shu ef al. /Applied Catalysis A: General 119 (1994) 305-325
Table 3
Influence of p&/pcti value on predicted methane conversion (conditions: 5WC, 136 kpa, H,O/CI-& = 3 and
GHSV= 1067 h-l)
Required step size AL Predicted Xc, at the reactor exit
1
0.8
0.5
0.2
0.1
0.08
0.05
0.02
0.01
Equilibrium
0.0001
0.0001
0.0001
0.0001
0.0001
0.0001
0.00005
0.00001
o.ooooO5
0.264
0.288
0.326
0.366
0.381
0.383
0.387
0.392
0.393
0.394
methane and for the formation of carbon dioxide in methane steam reforming were
thus given as:
R
dx
CH4
CH4 -
--=R1 +Rg
R
ax
co2
co2 -
--=R2+R3
(9)
(10)
where W is the catalyst weight, and F&, is the flow-rate of methane in the reactor
entrance.
Eqs. (9) and ( 10) can be further expressed by the introduction of a dimensionless
reactor length L, dL* =dLIL:
dx,H4_Pm-h3(R +R )
dL* F& 3
(11)
.o*_Pwe3
dx
dL*
F&4
(RI + 4) (12)
In Xu and Froments kinetic model, a critical parameter is the hydrogen partial
pressure. The present study used a hydrogen-free methane-steam feedstock, thus
the entrance hydrogen partial pressure was zero which made the initial rate of
methane conversion infinite. The problem was solved by setting a very small inlet
hydrogen partial pressure introduced as an arbitrary value of &,/pcn,. An evalu-
ation of the p&/p cti parameter setting upon the predicted methane conversion is
shown in Table 3. Although a smallerp~,/po,, value results in a higher calculation
precision, it needs much a smaller calculation step size, thus causing a longer
calculation time. The computation showed that at high residence time ( W/F), the



J , Shu et al. /Applied Catalysis A: General 119 (1994) 305-325 313
predicted methane conversion was in good agreement with the equilibrium one
when a ph21pCti value less than 0.1 was used. Thus the value 0.1 for p&/pcH4 was
chosen in the present kinetic permeation model. This treatment is physically realistic
considering the rapid conversion of methane to hydrogen and carbon dioxide, as
pointed out by Xu and Froment [ 281.
3.3. Permeationjux
The hydrogen permeation rate through the palladium membrane with a thickness
S has the following differential equation form:
-=
(13)
with pH2,p = H p
Y P / ( YH + I), where Z is the ratio of inert gas molar flow-rate with
respect to methane flow-rate, Z= Fy/F CH4; Q. and Ep are the pre-exponential factor
and the activation energy of hydrogen permeation, respectively.
This equation is de facto derived from the extended Sieve&s relation used for
the a-hydride phase regime [ 271. Since methane steam reforming needs high
temperature operating conditions (e.g., > 500C)) this a-hydride approach would
not lead to a significant error.
3.4. Boundary conditions
At L* =0: X,,,=X,,,=Y,=O
3.5. Numerical simulation
The set of differential Eqs. ( 11-13) was solved numerically by a 4th order
Runge-Kutta method [ 301.
4. Results and discussion
4.1. Methane steam reforming in a non-membrane reactor
As a reference, methane steam reforming was first tested in a reactor without
hydrogen-permeable membrane. For this purpose, a dense stainless steel tube was
employed in the reactor assembly in place of the membrane. After carefully starting
up, following the experimental procedures described previously, methane conver-
sion was measured as a function of temperature. A reaction pressure of 136 kPa
was set considering the favorable reaction equilibrium at low pressure and the
requirement for further membrane effect test. The steam-to-methane ratio used was



314 J . Shu et al. /Applied Catalysis A: General 119 (1994) 305-325
0 1 2 3 4
Reaction time, II
Fig. 4. Product distribution of methane steam reforming at various temperatures in the non-membrane reactor
(conditions refer to Fig. 5).
3. Fig. 4 shows an example of the steam reforming product distribution at various
reaction temperatures recorded using the mass spectrometer. The measured total
methane conversion data are plotted in Fig. 5 (Curve 1) . Repeated experiments
gave almost the same methane conversion results. The catalytic reaction was in a
steady state over the activated catalyst. Furthermore, raising the GHSV from 1067
h- to 1600 h- resulted in no change in methane conversion. It can be seen that
the methane conversion increases monotonically with increasing temperature in
this temperature range. The main reforming product is carbon dioxide at low
reaction temperature. For example, the methane selectivity to carbon dioxide is
96.6% at 400C and 86.5% at 500C.
For comparison, the equilibrium methane conversion was calculated using a
QuickBasic program. In the calculation, equilibrium constants for MSR and WGS
reactions were taken from ref. [2]. The calculated equilibrium values were also
plotted in Fig. 5 (Curve 2). As can be seen, the measured methane conversion was
slightly lower than the equilibrium value, and the difference increases with increas-
ing temperature and thereby methane conversion.
A prediction from Xu and Froments model indicated that near equilibrium
methane conversion could be achieved under the above reaction conditions. An X-
ray diffraction (XRD) measurement (Fig. 6) of the reforming catalyst showed that
metallic nickel, which is the active species in methane steam reforming, is present
on the used catalyst after activation even though the mild reduction temperature of
550C was used. In the original C 1 l-9-02 reforming catalyst, the nickel component
was supported on a-A1203 mostly in the form of NiO. Unlike some other steam
reforming catalysts, no NiA1204 spine1 was found in this original catalyst. This



J . Shu et al. /Applied Catalysis A: General 119 (1994) 305-325 315
40
I
30 -
20 -
250 300 350 400 450 500 550
Temperature,C
Fig. 5. Methane conversion in a non-membrane reactor: ( 1) non-membrane; (2) calculated equilibrium.
might be the characteristics of the low-temperature reforming catalyst as the reduc-
tion of NiO is much easier than that of NiA1204. After the activation at 550C in
the hydrogen atmosphere, NiO was entirely reduced, as revealed by XRD.
4000
5
d
^
2 3000
a
B
s
2000
loo0
20 30 40 50 60 70
2theta
Fig. 6. X-ray diffraction patterns of the steam reforming catalyst: (a) original catalyst; (b) after reduction at
550C.



316 J. Shu et al. /Applied Catalysis A: General 119 (1994) 305-325
Table 4
Possible carbon deposition reactions in methane steam reforming
No. Reaction
A@%?
(kJ/mol)
KP at 773 K (zzFqi
(units as K,)
(ZIP) .lK
I I
1 C&+C+2H, - 74.8 51.7 kPa 91.2 1.76
2 2co+c+c4 172.5 2.56 kPa_ 4.5 1 1.76
3 CO+H2+C+HZ0 131.3 0.46 Wa- 1.14 2.47
4 CO* + 2H2 90.1 1.87. 10d4 kPa_ 0.29 1.57. lo3
+C+2H,O
5 cH4+2co 187.8 11.2 Wa- 120.4 10.8
+ 3C + 2H20
6 CH, + co* 15.3 4.35 26.7 6.14
-+ 2C + 2H20
The possibility of the catalyst deactivation caused by carbon deposition over the
catalyst was examined. Table 4 lists possible carbon deposition reactions in the
methane steam reforming process. Equilibrium constants of each reaction at 500C
were calculated based on thermodynamic data taken from ref. [ 3 11. Under our
typical reaction conditions of XWC, 136 kPa and a steam-to-methane ratio of 3,
the equilibrium criteria (nP,Yi)J& (here vj are the stoichiometric coefficients,
with sign plus for productsand minus for reactants) for each reaction exceed 1 in
accordance with the calculation based on the corresponding reaction product dis-
tribution. That is to say, no carbon species would deposit over the catalyst surface.
This was confirmed by the fact that no catalyst deactivation evidence was found
after ca. 24 h of operation.
After a careful analysis of the experimental conditions, the difference between
the measured methane conversions and the equilibrium values observed in Fig. 5
is believed to result from the temperature gradient inside the integral reactor. The
reactor was made of stainless steel with a thermocouple tube welded to the reactor
body. The thermocouple measuring point was set at the exit of the reactor in order
to measure the exit temperature. However, because of the integrated reactor form
and the poor thermal conductivity of the catalyst, the strongly endothermic steam
reforming reaction might result in a temperature difference between the catalyst
bed and the reactor wall. Temperature gradients as high as 10C do exist inside
reformers [ 51. The present temperature is measured close to the reactor wall while
the catalyst bed temperature tends to decrease toward the inner tube. Since the
purge gas (helium) was also used in this non-membrane system, it constituted
another important cooling factor. The helium stream was preheated to 200C before
entering into the tube side of the reactor. This effect would also enlarge the tem-
perature gradient across the catalyst bed. Unfortunately, the temperature profile is
hard to calculate since the heat transfer from the environment is unknown. For the
sake of comparison, the measured conversion in the non-membrane system was
used as a reference in the present study.



J . Shu et al. /Applied Catalysis A: General 119 (1994) 305-325 317
0 30 40 60 80
Hz removal percent, %
100
Fig. 7. Influence of hydrogen removal on the equilibrium methane conversion (conditions: P, =136 kPa; Pp =101
kPa; H,O/CH.,= 3; FcH4 = 40 SCCM; FHe = 40 SCCM)
4.2. Temperature effect in the membrane reactor
The major advantage of the membrane reactor is the conversion enhancement of
an equilibrium-limited reversible reaction by selective hydrogen removal. In the
case of methane steam reforming, the use of a membrane reactor can lead to a
significantly enhanced methane conversion at a moderate temperature. Fig. 7 shows
such a thermodynamic equilibrium calculation of the methane conversion as a
function of the hydrogen removal ratio. This was done by simply setting a residual
hydrogen partial pressure in the equilibrium calculation for methane steam reform-
ing.
As can be seen from Fig. 7, when the temperature is below 4OOC, the reforming
reaction could not reach a satisfactory conversion level until a high hydrogen
separation efficiency, say 90%. Over 7OOC, the hydrogen removal through the
membrane can only slightly shift the reforming equilibrium due to the intrinsically
favorable temperature effect. At a moderate temperature of 500 to 6OOC, membrane
separation can result in a great improvement on the methane steam reforming
equilibrium. This is why moderate temperature conditions were used in the present
study.
The temperature effect of membrane separation on MSR was studied under the
same conditions as in the case of the non-membrane reactor. Both Pd/SS and Pd-
Ag/SS membranes were examined. When methane steam reforming was performed
in the palladium membrane reactor without a catalyst, no detectable conversion



318 J . Shu et al. /Applied Catalysis A: General 119 (1994) 305-325
136kPa, H~O/CH4=3
GHSV= 1067li
01 I I
250 300 350 400 450 500 550
Temperature,%
Fig. 8. Temperature effect in Pd-based membrane reactor: ( 1) non-membrane; (2) calculated equilibrium; (3)
Pd/SS; (4) calculated from the model for Pd with 6= 20 pm; (5) Pd-Ag/SS.
occurred at XMYC, which might be due to the low specific surface area of the
membrane as well as its limited intrinsic catalytic activity. Thus a catalyst was
necessary to perform effectively the reforming reaction. Fig. 8 illustrates the total
methane conversion against reaction temperature. In comparison with the methane
conversion in the non-membrane system (Curve 1) , the membrane effect is clear
in the Pd/porous SS membrane case (Curve 3). This conversion curve is also
beyond the calculated equilibrium methane conversion (Curve 2). No deactivation
evidence of the membrane was observed after about 47 h of reaction tests. The
membrane remained still dense showing no cracks or pinholes, and permselective
toward hydrogen separation. No carbon-based reactant or products were detected
in the permeation side. It is believed that the close thermal expansion coefficients
of palladium ( 11.8 pm/m C at 20C) and 3 16L stainless steel ( 14.7 pm/m C at
20C)) compared to the value for alumina ceramics (5.4 pm/m C at 20C) make
the Pd/SS membrane more resistant to thermal cycling than Pd/ceramic mem-
branes [24]. Curve 5 is the methane conversion performed in the Pd-Ag/SS
membrane reactor. Evidently, methane steam reforming was greatly enhanced by
using this Pd-Ag membrane. For example, at a total pressure of 136 kPa, a reaction
temperature of XKYC, and a molar steam-to-methane ratio of 3, a methane conver-
sion of 50.9% was achieved, much higher than that in the non-membrane system
(36.7%). The methane conversion was reproducible after a series of reaction tests.



J . Shu et al. /Applied Catalysis A: General 119 (1994) 305-325
319
When compared with the result from Pd/SS membrane, the enhancement in the
case of Pd-Ag/SS should be attributed to the combination of the use of a thinner
film and a possibly higher hydrogen permeability through the Pd-Ag membrane.
The temperature effect was also calculated using the kinetic permeation model
for the Pd/SS membrane. Under the above mentioned reaction conditions, methane
conversions at various temperatures were predicted, as shown in Fig. 8 (Curve 4).
Additional input parameters include a membrane thickness of 20 pm, a purge gas
flow-rate of 40 SCCM, a hydrogen permeation activation energy of 15.7 kJ/mol
and a pre-exponential factor Q, of 1.56 - lo-* m3/m s Pao.5 [ 321. Comparing the
calculated data with Curve 3, it can be seen that the two lines fit essentially at
temperatures below 350C. With raising reaction temperature, the measured con-
versions become lower than the predicted ones. The overestimated calculation is
likely attributed to the temperature gradient inside the membrane reactor as well as
our simplified model assumption. Both the endothermal behavior of the steam
reforming and the cooling effect of the sweep gas yield a temperature decreasing
toward the inner tube. A boundary layer might exist on the membrane surface
toward the porous support. Besides the axial hydrogen concentration gradient, there
is a radial distribution of permeated hydrogen through the porous support when the
driving force is given by use of a sweep gas.
4.3. Influence of the H,O/CH, ratio
Methane steam reforming usually proceeds in the presence of an excess of steam
to prevent the carbon deposition over the catalyst surface and to enhance the steam
reforming. The effect of molar steam-to-methane (H,O/CI&) ratio on methane
conversion was examined by varying this ratio from 2 to 4. The results obtained at
500C are drawn in Fig. 9. Curve 1 is the experimental data obtained from the non-
membrane system. Curve 2 is the calculated equilibrium conversion, exhibiting
monotonic increase with the steam-to-methane ratio. It can be seen that these points
are slightly lower than the equilibrium conversion, showing the same trend as in
Fig. 8. In the case of the Pd/SS membrane, an enhancement of the membrane
separation efficiency was realized by continually pumping the permeation side
(tube side) with a rotary vacuum pump. The pumping eliminated the radial distri-
bution of the hydrogen concentration inside the support pores and kept the per-
meation side under a low pressure, resulting in a relatively high driving force of
hydrogen permeation through the palladium membrane. The measured methane
conversion against the steam-to-methane ratio is shown as Curve 4 in Fig. 9. It is
clear that the methane conversion was greatly enhanced in this manner. Under our
typical operating conditions (500C 136 kPa and a molar steam-to-methane ratio
of 3), the achieved methane conversion was 63%, almost twice as high as the one
in the non-membrane system. For the Pd-Ag/SS membrane, the effect of HZ01
CH, is shown in Curve 3. In this case, helium passed through the permeation side
at a flow-rate of 40 SCCM. The promotion of the reforming conversion is clearly



320 J . Shu et al. /Applied Catalysis A: General I 19 (1994) 305-325
loo1
60
201
!
1.5 2 2.5 3 3.5 4 4.5
Steam/methane ratio
SO0 %. 136kPa
GHSV~1067ti
Fig. 9. Steam-to-methane ratio effect in membrane reactor: ( 1) non-membrane; (2) calculated equilibrium; (3)
Pd-Ag/SS; (4) Pd/SS; (5) calculated from the model for Pd with 6= 20 pm.
exhibited. It should be noticed that the lower conversion obtained with the Pd-Ag/
SS membrane is due to the fact that in this experiment a smaller difference in
hydrogen partial pressure in the two compartments was implemented which induced
a lower hydrogen permeation flux.
It is somewhat difficult to model the steam-to-methane ratio effect in the present
Pd/SS membrane case for lack of the exact pressure value on the permeation side.
A calculation using the kinetic permeation model indicates that a very low pressure,
as low as 1.3 Pa, on the permeation side would result in a complete methane
conversion. However, it is possible for a certain amount of residual gas to exist on
the permeation side considering the continual hydrogen permeation from the reac-
tion side and a long transfer tubing between the membrane reactor and the vacuum
pump. If a residual gas pressure of 30 kl?a is supposed, the calculated methane
conversion is approaching the measured conversion for the Pd/SS membrane, as
shown in Curve 5 of Fig. 9.
4.4. Pressure efSect in the membrane reactor
In a conventional reactor, the increase of pressure results in a decrease of the
methane conversion. However, the incorporation of a hydrogen permeable
membrane would change this general trend. Although a high reaction pressure



J . Shu et al. /Applied Catalysis A: General 119 (1994) 305-325
321
SOOC.H,0/C&=3
GHSV=1067b"
I
6
5
. . . . . --.
100 120 140 160 180 200 220
Reaction pressure, kPa
Fig. 10. Pressure effect on methane conversion: ( 1) non-membrane; (2) calculated equilibrium; (3) Pd-Ag/SS,
FHc = 40 SCCM; (4,5,6) calculated from the model for Pd with 6= 20 pm and FHe = 40, 140, 200 SCCM,
respectively.
depletes the steam reforming conversion equilibrium, the increase of hydrogen
partial pressure on the reaction side would increase the driving force for hydrogen
permeation, resulting in an enhancement of the methane conversion under certain
conditions.
The effect of pressure on methane steam reforming conversion is shown in Fig.
10. As expected, in the case of the non-membrane system, the methane conversion
decreased with increasing total reaction pressure, in agreement with the equilibrium
calculation. However, in the case of Pd-Ag/SS membrane, a decreasing variation
in the methane conversion with the increase of pressure was also observed. This
seems to indicate a low membrane separation efficiency.
Uemiya et al. [ 241 noticed a conversion increase with increasing reaction pres-
sure. The sweep gas they used was an argon stream in a flow-rate of 400 SCCM,
ten times higher than that in our case. An increasing conversion upon pressure was
predicted by Prokopiev et al. [ 121 based on a mathematical model calculation. In
their calculation, the partial pressure of permeated hydrogen was taken as zero. We
also calculated the dependence of methane conversion on the total pressure under
various purging conditions for a palladium membrane of 20 pm thick. It was found
that the increasing tendency exists in the case where hydrogen separation is efficient.
If the separation is relatively weak, for example with a mild purging effect on the
permeation side, the decreasing methane conversion upon pressure is maintained



322 J . Shu et al. /Applied Catalysis A: General 119 (1994) 305-325
in the low pressure region. Large flow-rates of the purge gas bring in a direct
increase of methane conversion with the pressure. These results are reported in Fig.
10 where three levels of helium flow-rate are examined. There is thus no contra-
diction between our observation and the literature results. In our case, the use of a
mild purge flow-rate (40 SCCM) and a low reaction pressure yielded a decreasing
conversion dependence upon the pressure.
4.5. Purge effect
Purging the permeation side is important in maintaining a high hydrogen partial
pressure difference between the two sides of the membrane and moving out hydro-
gen product from the reaction system. As discussed above, the increase of the purge
gas flow-rate is expected to enhance the methane conversion owing to the decrease
of the hydrogen partial pressure on the permeation side. This effect was examined
in the Pd/SS membrane system, as shown in Fig. 11 (Curve 1). The methane
conversion was found slightly increased with raising the purge gas flow-rate.
The computer model analysis for the palladium membrane predicts a stronger
sweeping flow-rate effect on the methane conversion (Curve 2 in Fig. 11) than the
one determined in our experiment. The effect reported by Uemiya et al. [24] is
also lower than the calculated purge effect for the same reaction under similar
ioor
80 -
20 -
O-
20
-
sooc, H*O/CH,=3
136kPa. GHSV=1067li
60 100 140 180
He flow rate, SCCM
Fig. 11. Purge flow-rate effect on methane conversion: (1) Pd/SS; (2) calculated from the model for Pd with
S=2Opm.



J . Shu et al. /Applied Catalysis A: General 119 (1994) 305-325 323
conditions. This may imply that there is a great potential to improve the membrane
separation efficiency. In our experiment, the higher the flow-rate of the purge gas,
the more the measured methane conversion deviated from the predicted value. This
was quite possibly caused by the cooling effect of the purge gas stream as well as
the oversimplified boundary layer condition. In future work, the improvement of
the reactor setup to eliminate this cooling effect or measuring the temperature
profile in the reactor will be useful to completely understand the membrane effect-
iveness.
In general, increasing the purge gas flow-rate enhances the hydrogen separation
efficiency. However, the economy in a large scale industrial application may not
permit the use of large amounts of inert gas. Instead, in the case of methane steam
reforming, steam may be used as the purge gas. This would also make the subsequent
separation of hydrogen from the purge stream much easier since steam can be easily
condensed from the mixture. Another choice to increase the hydrogen permeation
driving force is the vacuum operation on the permeation side, as already demon-
strated in this study. It seems that operation under rough vacuum ranging from 10 1
kPa to ca. 100 Pa is industrially acceptable from economic considerations [ 331.
5. Conclusions
Electrolessly prepared Pd/SS and Pd-Ag/SS composite membranes were used
to perform methane steam reforming in the presence of a Ni/A1203 catalyst. The
methane conversion was significantly enhanced as a result of diffusion through the
Pd-based membrane by the partial hydrogen removal from the reaction location.
For example, at a total pressure of 136 kPa, a temperature of 500C and a molar
steam-to-methane ratio of 3 together with continuous pumping on the permeation
side, a methane conversion twice as high as the conversion in the non-membrane
system was reached by using a Pd/SS membrane. No deactivation evidence of this
membrane was observed after about 47 h of operation. These effects were examined
under a variety of experimental conditions. A computer model of the reactor was
developed and utilized in predicting the effects of membrane separation on methane
conversion.
Nomenclature
c molar fraction (%)
E activation energy (J/mol)
EP
activation energy of hydrogen permeation (J/mol)
F flow-rate ( m3( STP) /s)
GHSV gas hourly space velocity (h-i)
h height (m)



324
AH
k
K
L
L*
m
O.D.
P
P
Qo
i?
T
W
X
Y
J . Shu et al. /Applied Catalysis A: General I 1 9 (1994) 305-325
heat of reaction (J/mol)
rate constant
equilibrium constant
reactor length (m)
dimensionless reactor length
molar steam/methane ratio
outer diameter (m)
partial pressure (Pa)
total pressure (Pa)
pre-exponential factor ( m3/m s Pa.5)
reactor radius (m)
reaction rate (kmol/kg-cat h) ; or gas constant (8.314 J/mol K)
temperature (K)
catalyst weight (kg)
conversion (%)
permeated hydrogen fraction with respect to the produced hydrogen
(%)
6. I. Greek letters
s membrane thickness (m)
P
catalyst density ( kg/m3)
6.2. Subscripts
r reaction side
P
permeation side
i inner
0 outer
vj
stoichiometric coefficient
6.3. Superscript
0 initial
Acknowledgements
This work was supported by the Natural Sciences and Engineering Research
Council (NSERC) of Canada. The authors thank CANMET-EDRL at Varennes
for providing facilities to perform methane steam reforming, and United Catalysts,
Inc. (Louisville, KY) for supplying Ni/A1203 steam reforming catalysts. They



J . Shu et al. /Applied Catalysis A: General 119 (1994) 305-325 325
also acknowledge Drs. Gilles Jean, Jean Paquette and Michel Pokier for valuable
experimental suggestions.
References
[ 1 ] J.P. Van Hook, Catal. Rev.-Sci. Eng., 21 ( 1980) 1.
[2] M.V. Twigg, Catalyst Handbook, 2nd edition, Wolfe, England, 1989, p. 225.
[ 31 E.L. Tollefson, in K.J. Smith and EC. Sanford (Editors), Progress in Catalysis (Studies in Surface Science
and Catalysis, Vol. 73), Elsevier, Amsterdam, 1992, p. 179.
[4] J.D. Fleshman, Chem. Eng. Prog., 89 (1993) 20.
[S] J.R. Rostrup-Nielsen, Catalytic Steam Reforming, Springer, Berlin, 1984.
[6] V.M. Gryaznov, Plat. Met. Rev., 30 ( 1986) 68.
[7] J.N. Armor, Appl. Catal., 49 (1989) 1.
[ 81 J.N. Armor, ChemTech, ( 1992) 557.
[9] J. Shu, B.P.A. Grandjean, A. Van Neste and S. Kaliaguine, Can. J. Chem. Eng., 69 (1991) 1036.
[lo] M.M. Oertel, J. Schmitz, W. Weirich, D. Jendryssek-Neumann and R. Schulten, Chem. Eng. Technol., 10
(1987) 248.
[ 1 l] J. Schmitz and H. Gerke, CLB, Chem. Labor. Betr., 39 ( 1988) 75.
[ 121 S.I. Prokopiev, Yu.1. Aristov, V.N. Parmon and N. Giordano, Int. J. Hydrogen Energ., 17 ( 1992) 275.
[ 131 A.M. Adris, S.S.E.H. Elnashaie and R. Hughes, Can. J. Chem. Eng., 69 (1991) 1061.
[ 141 M. Chai, M. Machida, K. Eguchi and H. Arai, Chem. Lett., (1993) 41.
[ 151 M. Chai, M. Machida, K. Eguchi and H. Arai, Appl. Catal. A, 110 ( 1994) 239.
[ 161 E. Kikuchi, S. Uemiya, N. Sato, H. Inoue, H. Ando and T. Matsuda, Chem. Len., (1988) 489.
[ 171 E. Kikuchi, Chem. Ind. (Jpn.), (1989) 37.
[ 181 E. Kikuchi, Chem. and Ind. (Jpn.), 42 (1989) 442.
[ 191 E. Kikuchi, Chem. Eng. (Jpn.), (1990) 31.
[20] E. Kikuchi, S. Uemiya and T. Matsuda, in A. Holmen, K.J. Jens and S. Kolboe (Editors), Natural Gas
Conversion (Studies in Surface Science and Catalysis, Vol. 61), Elsevier, Amsterdam, 1991, p. 509.
[21] E. Kikuchi, in Proc. of the 7th ROC and Japan Joint Workshop on Catalysis, Chungli, Taiwan, 1992, p. 43.
[22] S. Uemiya, Y. Kude, K. Sugino, N. Sato, T. Matsuda and E. Kikuchi, Chem. Lett., (1988) 1687.
[23] S. Uemiya, N. Sato, H. Ando, T. Matsuda and E. Kikuchi, Sekiyu Gakkaishi, 33 ( 1990) 418.
[24] S. Uemiya, N. Sato, H. Ando, T. Matsuda and E. Kikuchi, Appl. Catal., 67 ( 1991) 223.
[25] S. Uemiya, N. Sato, H. Ando and E. Kikuchi, Ind. Eng. Chem. Res., 30 (1991) 589.
[26] J. Shu, B.P.A. Grandjean, E. Ghali and S. Kaliaguine, J. Membr. Sci., 77 (1993) 181.
[27] J. Shu, PhD Thesis, Lava1 University, PQ, Canada, 1994.
[28] J. Xu and G.F. Froment, AIChE J., 35 ( 1989) 88.
[29] S.S.E.H. Elnashaie, A.M. Adris, AS. Al-Ubaid and M.A. Soliman, Chem. Eng. Sci., 45 (1990) 491.
[ 301 J.C. Sprott, Numerical Recipes: Routines and Examples in BASIC, Cambridge University Press, Cambridge,
1991, p. 360.
[31] W.J. Moore, Basic Physical Chemistry, Prentice Hall, New Jersey, 1983.
[32] S.A. Koffler, J.B. Hudson and G.S. Ansell, Trans. AIMS, 245 (1969) 1735.
[ 331 N. Milleron, in R.E. Kirk and D.F. Othmer (Editors), Encyclopedia of Chemical Technology, 3rd edn., Vol.
23, Wiley, New York, 1983, p. 644.