Catalysis Today 25 ( 1995) 303-307

ELSEVIER

Steam reforming of methane in a membrane reactor

S.LaegsgaardJ~rgensen *, P.E.H@jlund Nielsen, P. Lehrmann
Haldor Tops& A/S, Nym)llevej 55. P.O. Box 213, DK-2800 Lyngby, Denmark

Abstract

Methane reforming in a Pd/Ag membrane reactor was carried out. The operational limits for the steam-to-methane ratio are
discussed. To avoid carbon formation, this ratio has to be higher in a Pd/Ag membrane reactor than in a conventional steam
reforming tubular reactor.

1. Introduction From this point of view, there will be a consid-

erable incitement to lower the reaction tempera-
ture. The equilibrium limitation then has to be
Steam reforming of methane is an endothermic overcome. Therefore, the reaction should be car-
process
ried out either at a very low pressure or, alterna-
tively, in a membrane reactor [ 21. If a palladium
CH4 + HZ0 = CO + 3H2 (1)
membrane reactor is used it provides the addi-
( -A@,, = - 206.2 kJ/mol) tional advantage of delivering very pure hydro-
P(C0) sZ’(H~)~ gen. This could be very important when a very
pure hydrogen is needed, such as for certain types
K1=P(CH,) *P(H20)
of fuel cells (PAFC, PEFC and AFC) .
CO+H20=C02+H, (2) The advantage of using a membrane reactor is
(-A&,=41.2 kJ/mol) demonstrated by the above equilibrium constant
where the hydrogen partial pressure enters the
The conversion of methane is limited by chem- third power. The removal of hydrogen from the
ical equilibrium. It is enhanced by high tempera- reaction zone increases the methane conversion
ture and low pressure and a high steam content [ 3-51. The reaction may consequently be carried
(oxygen-to-carbon ratio). In the majority of out at a lower temperature.
industrial ammonia and hydrogen plants, a high Besides an increased conversion of methane at
pressure in the order of 3-4 MPa is to be preferred. lower temperatures, the removal of hydrogen from
Thus, temperatures around 1100 K are applied. the reaction gas influences the potential for carbon
These conditions are quite strenuous as regards formation via
construction material [ 1] and, therefore, expen-
sive steel alloys are used. CH,=C(s) +2H, (3)
( -A@,, = - 75 kJ/mol)
* Corresponding author.

0920-5861/95/$09.50 0 1995 Elsevier Science B.V. All rights reserved

SSD10920-5861(95)00074-7
304 S. Lregsgaard J#rgensen et al. /Catalysis Today 25 (I 995) 303-307

side are being reviewed. We have further estab-

2CO=C(s) +co, (4) lished criteria for carbon-free operation.
( - A$??, = 173 kJ/mol)

A major drawback in the use of dense palladium 2. Experimental set-up

membranes in large-scale reforming plants is the
availability and the price of palladium. The hydro- The reactor was installed in a furnace and con-
gen permeability of palladium (alloys) is low sists of a Pd-23 wt.-.% Ag tube (0.1 mm thick,
compared to the very high catalytic reaction rates 380 mm long, outer diameter 1.55 mm) centred
and the high industrial space velocities. The devel- within an outer reactor tube (inner diameter 4.47
opment of membrane reactors with a high per- mm) made of stainless steel. The applied catalyst
meation rate is, therefore, very desirable. was a commercially available (Haldor Topsoe A/
The permeation of hydrogen through palladium S) steam reforming catalyst. The catalyst was
alloy membranes is described by the following based on Ni and MgO and had a high activity at
relation that is valid as long as the diffusion of low temperatures. 4.65 g of the catalyst was
hydrogen through the metal is considered the rate crushed and packed uniformly in the reaction zone
limiting step [ 61: between the two tubes. The length of the catalyst
bed was 320 mm. This means that the hydrogen
D.A diffusion is not solely restricted to the catalyst
F=- I Cc- 6) [Nm3/hl (5)
zone. About 30 mm of space was available for
hydrogen diffusion before the inlet and after the
D is the apparent diffusion coefficient (m* exit of the catalyst bed.
h-‘&-i), wh’ich’1s dependent on temperature A feed gas of 99.99% CH4 was supplied from
and the nature of the Pd alloy present. A is the area a bottle and was led over a Cu/ZnO catalyst at
of the membrane (m*) ; 1 is the membrane thick- 275°C for final purification. The gas flow of CH,
ness (m) ; PH2.rand PH2,$are the partial pressures was controlled by a mass flow-meter. The water
of hydrogen (bar) on the reaction and separation was purified by ion exchangers, supplied by a
side of the membrane. pump and evaporated at 200°C.
The only way to drastically increase the per- The reaction gas was mixed to a steam-to-meth-
meation rate is to decrease the membrane thick- ane ratio of 2.9 and preheated before entering the
ness. Research in preparation of ultrathin reactor. The space velocity was 0.5 Nl CHJg
supported Pd membranes has therefore been car- catalyst/h. The sweep gas ( 10 Nl/h of N2) flowed
ried out [ 7,8]. Gains with respect to a higher dif- countercurrently to the reaction gas.
fusivity, improved mechanical properties and At the exit of the reaction side of the membrane
improved resistance to poisoning may be obtained the unreacted water was condensed and measured.
by modifying the alloying composition [ 9,101. The hydrocarbon containing dry gas was analysed
Minor additional gains with respect to mini- on a gas chromatograph. The dry exit gas flows
mizing the needed palladium membrane area can were measured by use of mass flow-meters.
be achieved by performing the reforming reaction The temperature used is the oven temperature
at conditions with a high partial pressure of hydro- and the system pressure was measured at the reac-
gen. tor system exit. The reforming experiments were
In this study we have investigated the steam performed at temperatures between 350-500°C.
reforming of methane in a Pd/23% Ag membrane The pressure on the reaction side was in the range
reactor at temperatures from 350 to 500°C and at of 6.1-10 bar g. The pressure on the sweep side
pressures from 0.6 to 1 MPa. Conditions favouring was 0.5 bar g.
a high partial pressure of hydrogen on the reaction The conversion of methane was calculated as:
 S. Lqsgaard Jmrgensen et al. /Catalysis Today 25 (I 995) 303-307 305

yco* + yco
Conversion of CH, = (6)
yctr4 + yco + yco2
where Y is the mole fraction in the exit gas from
the reaction side.
The hydrogen-to-carbon ratio (H/C) of the exit
gas from the reaction side was calculated as:

.,,=2.~,,+2.y,,,+4.y,,, Temperature [Cl

(7)
YCH4+ yco + yco*
- 6 bar 9, WC Equil -- 10 bar g Calc Equil ?? 6 bar g. No Sweep
+ 6 bar g on Sweep + 10 bar g on Sweep

Fig. 1. Plots of measured conversion of methane as a function of

3. Results and discussion temperature. The fully drawn lines are the calculated equilibrium
curves. Steam-to-carbon ratio is 2.9.

3.1. Experimental results

The measured conversion of methane in the

membrane reactor as a function of temperature is
summarized in Fig. 1. As expected it is verified
that the conversion of methane increases when H2
is removed from the reaction system.
At 500°C and 6 bar g the conversion of methane
10 1 I_ --T. -___~-__ 1
was measured without gas flow on the sweep-side. I
I
‘_
0
The measured conversion corresponds very well 2 3 4 5 6 7 a 9 10
H/C-Ratio
with the expected equilibrium conversion of 2 1%
at these conditions.
I-5OOC,6barg -350C,6barg ---5OOC,lObarg ’
At 500°C 6 bar g and with a gas flow on the - 500C.6barg + 350 C, 6 bar g i* 5OOC, 10barg

sweep-side, 51% of the methane is converted. A

Fig. 2. Plots of measured conversion of methane as a tunctron ot the
5 1% conversion corresponds to an equilibrium hydrogen-to-carbon ratio of the exit gas from the reaction side. The
temperature of 639°C calculated from the inlet gas fully drawn lines are the calculated equilibrium curves.
( 1). At 500°C and 10 bar g the conversion is
increased to 6 1% corresponding to an equilibrium
temperature of 707°C. The highest conversion of
methane is obtained at the highest temperature and
F
pressure level. 0

The reaction gas from the catalyst bed exit is in 2

a.
chemical equilibrium. Fig. 2 shows the conver-
sion of methane as a function of the H/C ratio of
the process gas. The fully drawn line is calculated
from an equilibrated gas. The measured points are 2 3 4 5 6 7 a 9 10
H/C-Ratio
generally placed a little to the left of the calculated
Fig. 3. Calculated equilibrium partial pressure of hydrogen for a gas
equilibrium curve because the exit gas looses with a constant oxygen-to-carbon ratio of 2.9 as a function of the
some hydrogen to the sweep side before entering hydrogen-to-carbon ratio of the gas. ( x ) 5OO”C, 10 bar g; (m)
the measuring system. 500°C 6 bar g; ( + ) 450°C 6 bar g; (second line from bottom)
400°C 6 bar g; (0) 350°C 6 bar g.
Fig. 3 shows the calculated equilibrium partial
pressure of hydrogen at the experimental condi-
306 S. hgsgaard Jqrgensen et al. /Catalysis Today 25 (1995) 303-307

al. obtained about 80% methane conversion. The

a$ 700
membrane thickness in Uemiya’s experiments is
0
only 20 micron which is a possible explanation
$ 600
1
?a
for Uemiya’s higher conversion. Uemiya also
2g 500
observed an increase in the methane conversion
with an increase of the pressure on the reaction
side.
i40. /
8 300. /
0 o/c=1
3.2. Criteria for carbon-free operation
200-l """'
0 12 3 4 5 6 7 6 9 10
H/C - RATIO
In addition to the reforming and shift reactions
600 ( 1) and (2)) carbon formation can take place by
o/c=1 (b)
the endothermic decomposition of methane (3)
8 750
I\IllII! II
k? or by the exothermic decomposition of carbon
0 700- '
\ '
/

c
monoxide, i.e. the Boudouard reaction (4).
s ffio- In a catalytic system the risk of carbon forma-
% \ , tion is normally evaluated by means of the so-
E 600-
z \
called ‘Principle of Equilibrated Gas’ stating that
:g 550
carbon formation is to be expected if the gas shows
F 500 affinity for carbon formation after the establish-
H
cl 450 ment of the reforming ( 1) and shift equilibria (2)
400
[Il.
0 12 3 4 5 6 7 6 9 10
H/C - RATIO
The principle is developed in the light of the
Fig. 4. (a) Upper carbon limit temperatures (TM) based on graphite fact that the reforming and water-gas shift reaction
data. T=XKY’C. P=6 bar g. (b) Lower carbon limit temperature is very fast. This means that the feed gas rapidly
( TB) based on graphite data. T= 500°C. P = 6 bar g. reaches equilibrium in the catalyst bed.
As the removal of hydrogen through a dense
tions of temperature, pressure and steam-to-car- Pd/Ag membrane is slow compared to the cata-
bon ratio as a function of the H/C ratio of the lytic reaction rate the gas composition near the
process gas. Maximum partial pressure of hydro- membrane equals the gas composition in the cat-
gen is reached at high temperature and pressure. alytic bed. On the assumption that the effect of
As the catalyst is considered active enough to film diffusion is negligible the gas near the
establish equilibrium at the actual process condi- membrane surface is thus in chemical equilibrium
tions the effect of increasing the pressure from 6 and the principle can be extended to be valid for
to 10 bar g on the reaction side (Fig. 1) is easily the membrane as well as the catalyst.
understood. An increase of the total pressure Calculations of the lower carbon temperature
results in an increase of the partial pressure of limit for the methane decomposition ( TM) and the
hydrogen, thereby helping the membrane to upper temperature limit for the Boudouard reac-
remove hydrogen down to a lower H/C ratio of tion (Tu) are carried out as follows:
the process gas, which results in a higher methane For a given feed gas composition, temperature
conversion. and pressure, the gas composition after the estab-
Uemiya et al. [ 111 have earlier demonstrated lishment of the reforming and shift equilibria was
the benefit of performing the steam reforming calculated.
reaction in a palladium membrane reactor. At Then the reaction coefficient for the methane
500°C and 10 bar g at the reaction side Uemiya et decomposition is calculated as:
 S. hgsgaard &rgensen et al. /Catalysis Today 25 (1995) 303-307 301

theoretical temperature range of carbon forma-

Kac

9 7
3= (yH*)**p (8) tion, even if hydrogen is removed down to the
YCHd least possible value of the H/C ratio.
and the reaction coefficient for the carbon mon-
oxide decomposition is calculated as 4. Conclusion

As the catalytic reaction rate of the reforming

(9) reaction is fast compared to the rate of hydrogen
removal through a dense palladium membrane the
The calculated reaction coefficients are com- gas in the membrane reformer is very near chem-
pared to the equilibrium constants obtained from ical equilibrium.
literature data based on carbon as graphite [ 1 ] and As the equilibrium partial pressure of hydrogen
the corresponding temperatures (TM and 7’n) increases with temperature and pressure, the
above or below which there will be affinity for obtained recovery of hydrogen through the
carbon formation is found. membrane and the obtained conversion of CH,
Fig. 4 shows the upper temperature limit of the increase with increasing temperature and pressure
exothermic methane decomposition reaction (3) Due to the decline in the H/C ratio of the proc-
and the lower temperature limit of the exotbermic ess gas the membrane reformer has to be operated
carbon monoxide decomposition reaction (4) at at a higher steam-to-carbon ratio compared to an
various steam-to-carbon ratios of 500°C and 6 bar ordinary, low-temperature steam reformer.
g, as a function of the H/C ratio of the process
gas. Acknowledgements
An ordinary, low-temperature steam reformer
is operated at a constant steam-to-carbon ratio and The authors thank J.-H. Bak Hansen for valu-
at a constant hydrogen-to-carbon ratio. A rela- able discussions and S. E. Mikkelsen for his assis-
tively low steam-to-carbon of about 1 and a cor- tance in the experimental work.
responding hydrogen-to-carbon ratio of 6 is thus
sufficient to avoid carbon forming conditions.
When the reforming reaction is performed in a References
Pd/Ag membrane reactor hydrogen is selectively [ 11 J.R. Rostrup Nielsen, Catalytic Steam Reforming, Springer
removed from the reaction gas, which results in a Verlag, Berlin, 1984.
decline of the H/C ratio of the process gas. The [2] R.G. Minet, Catalytic Ceramic Membrane Steam-
Hydrocarbons Reformer, US Patent No. 5 229 102 (1993).
oxygen-to-carbon ratio remains, however, con-
[31 A.M. Adris, S.S.E.H. Elnashaie and R. Hughes, Can. J. Chem.
stant. Eng., 69 (1991) 1061-1070.
Fig. 4 illustrates the approached temperature [4] M. Oertel, J. Schmitz, W. Weirich, D.J. Neumann and R.
limits of carbon formation with the decline of the Schulten, Chem. Eng. Technol., 10 ( 1987) 248-255.
[5] M. Chai, M. Machida, K. Eguchi and H. Arai, Appl. CataJ. A:
H/C ratio. General, 110 (1994) 239-250.
If a gas with a steam-to-carbon ratio of 1 is [6] G. Saracco and V. Specchia, Catal. Rev. Sci. Eng., 36 (2)
reformed in a membrane reactor the temperature ( 1994) 305-384.
[7] S. Uemiya, T. Matsuda and E. Kikuchi, J. Membr. Sci., 56
limits of carbon formation are exceeded at an H/ (1991) 315-325.
C ratio of 5.5 . Recovery of further hydrogen from 181 V.M. Ciryaznov, Plat. Met. Rev., 36 (2) (1992) 70-79.
the gas results in operation at possible carbon [91 H.P. Hsieh, Catal. Rev. Sci. Eng., 33 (1991) l-70.
forming conditions. [ 101 V.M. Gryaznov, O.S. Serebryannikova and Y.M. Serov, Appl.
Catal. A: General, 96 (1993) 15-23.
By keeping the steam-to-carbon ratio higher [ 111 S. Uemiya, N. Sato, H. Ando, T. Matsuda and E. Kikuchi,
than about 2.5, it is possible entirely to avoid the Appl. Catal., 67 ( 1991) 223-230.