Autothermal methanol reforming for hydrogen production in fuel cell applications

Article abstract of DOI:10.1039/b004881j

The catalytic autothermal reforming process of methanol is a promising option for hydrogen production. High-efficiency fuel cells and methanol reformers were developed to exceed the well-to-wheel efficiencies of 17% for the gasoline internal combustion engines (ICE). Microreactor testing of three different commercial Cu-containing catalysts was conducted at 250° and 330°C showing nearly complete methanol conversion at 85% hydrogen yield. All three catalysts were active for the autothermal reforming of methanol. For the overall process, a simplified model of the reaction network, comprising the total oxidation of methanol, the reverse water-gas shift reaction, and the steam-reforming of methanol, was proposed. Individual kinetic measurements for the latter two reactions on a commercial Cu/Zn/Al₂O₃ catalyst were presented.

Full text of DOI:10.1039/b004881j

View Article Online / Journal Homepage / Table of Contents for this issue
Autothermal methanol reforming for hydrogen production in fuel cell
applications
Konrad Geissler, Esmond Newson, Frederic Vogel, Thanh-Binh Truong, Peter Hottinger and
Alexander Wokaun
L aboratory for Energy and Material Cycles, Paul Scherrer Institute, CH-5232 V illigen-PSI,
Switzerland. E-mail: konrad.geissler=psi.ch
Received 19th June 2000, Accepted 15th August 2000
First published as an Advance Article on the web 3rd October 2000
Fuel cell powered electric cars using on-board methanol reforming to produce a hydrogen-rich gas represent a
low-emissions alternative to gasoline internal combustion engines (ICE). In order to exceed the well-to-wheel
efficiencies of 17% for the gasoline ICE, high-efficiency fuel cells and methanol reformers must be developed.
Catalytic autothermal reforming of methanol o†ers advantages over endothermic steam-reforming and
exothermic partial oxidation. Microreactor testing of copper-containing catalysts was carried out in the
temperature range between 250 and 330 ¡C showing nearly complete methanol conversion at 85% hydrogen
yield. For the overall process a simpliÐed model of the reaction network, consisting of the total oxidation of
methanol, the reverse water-gas shift reaction, and the steam-reforming of methanol, is proposed. Individual
kinetic measurements for the latter two reactions on a commercial Cu/ZnO/Al O catalyst are presented.
2
3
Of all the considered reactions, it o†ers the highest maximum
hydrogen content in the product gas (75%). In addition, since
no gases need to be compressed in the feed, the reaction can
easily be carried out at higher pressure, thus keeping mem-
branes as an option for the successive gas clean-up step. The
endothermicity of the reaction, however, requires permanent
external heating of the steam-reforming reactor which makes
short start-up times, as desired for mobile applications, and
fast transient behaviour difficult to achieve.5h7
1. Introduction
Hydrogen is considered to play an important role in future
energy systems.1 If produced from renewable sources, it is a
clean and carbon-free energy vector which can be converted
to electrical energy either by conventional combustion engines
or by fuel cells, the latter not being limited in theoretical effi-
ciency by the maximum Carnot efficiency. However, the idea
of using fuel cell propulsion systems for mobile applications
requires a safe and efficient means of hydrogen storage. Up to
now, none of the physical or chemical storage options being
considered as technically feasible has turned out to be the
method of choice.2 Among the liquid carriers under dis-
cussion, methanol o†ers a high energy density and the possi-
bility of using similar infrastructure for distribution. It is
considered to be safer than the currently used petrol3 with
respect to ignition temperature and other factors and it can be
produced from biological, i.e. renewable, sources as well as
from natural gas.4 In order to serve as a hydrogen source, it
must be converted by an on-board reforming process.
Even if the methanol for the above-mentioned process is
produced from non-renewables (such as natural gas), the
reformerÈfuel cell combination is, due to its low emissions, a
serious alternative to ICE if its full fuel cycle efficiency is
higher or at least the same. System efficiency analyses can
reveal whether this goal can be reached.
For the conversion of methanol to hydrogen di†erent
chemical reactions can be applied. The simplest option is the
decomposition reaction (1)
This problem is less severe if exothermic partial oxidation,
reaction (3), is used.
CH OH ] 1 O ] CO ] 2 H ;
3
2
2
2
2
* H¡ \ [192.3 kJ mol~1
(3)
R
Due to the exothermicity of the reaction no external heating
of a partial oxidation reactor is required and start-up times of
less than 60 s for partial oxidation reactors are reported in the
literature.8 However, in addition to the low maximum hydro-
gen content in the product gas (40% with air operation), the
high exothermicity of the reaction is also the main drawback
of this reaction. It drastically lowers the efficiency since waste
heat is generated and temperature control of the reactor is
complicated.
Autothermal reforming of methanol, reaction (4), an idea
which was developed in the late 1980s by Johnson-Matthey,9
is a combination of reactions (2) and (3) having a net reaction
enthalpy change of zero.
CH OH ] CO ] 2 H ; * H¡ \ ]91 kJ mol~1
(1)
4 CH OH ] 3 H O ] 1 O ] 4 CO ] 11 H ;
3
2
R
3
2
2
2
2
2
yielding hydrogen and carbon monoxide. However, the high
content of the latter in the product gas makes this reaction
unsuitable for polymer electrolyte membrane (PEM) fuel cell
applications. Probably the most widely used way of producing
hydrogen from methanol is the endothermic steam-reforming
reaction (2).
* H573K \ 0
(4)
R
As a consequence, a reactor for this process does not require
external heating once having reached reaction temperature.
Depending on temperature, slightly di†erent stoichiometric
ratios of the feed can be calculated. The maximum obtainable
hydrogen content in the product gas is 65% using the stoichi-
ometry at 300 ¡C. For faster start-up or transient response the
CH OH ] H O ] CO ] 3 H ;
3
2
2
2
* H¡ \ ]49 kJ mol~1
(2)
R
DOI: 10.1039/b004881j
Phys. Chem. Chem. Phys., 2001, 3, 289È293
This journal is ( The Owner Societies 2001
289

View Article Online
methanol/oxygen ratio in the feed can be varied which was
shown in the compact Hot-Spot} reformer.10 As for partial
oxidation, the main problem for scale-up of the autothermal
reforming process is temperature control in the reactor. This is
due to the strongly di†ering reaction rates of the exothermic
oxygen conversion and the endothermic steam reforming reac-
tion. In order to facilitate reactor design it is highly desirable
to develop a complete kinetic model for the process.
3. Experimental and results
3.1. Microreactor set-up
Microreactor experiments were carried out using an electri-
cally heated tubular reactor (silica-coated stainless steel, inner
diameter 4 mm), equipped with an inner tube (diameter 1.5
mm) containing a moveable thermocouple. The length of the
isothermal zone of the reactor is 65 mm. Catalyst particle size
was always 0.25 to 0.5 mm. Feed gases were 99.995% or
higher purity, the used methanol was analytical grade
(P99.8%, Merck). All gas Ñows were controlled by electronic
mass-Ñow controllers (Bronkhorst), the methanolÈwater
mixture was dosed by a pump (RCT M16). Product gas
analysis was carried out using a HP 5890 gas chromatograph
equipped with both a thermal conductivity detector connected
to a two-column switching system (HP Plot Q and HP Plot
5 A), and a Ñame ionisation detector connected to an Alltech
AT 5 column. Additional measurements of the condensate
water contents were done using a 737 KF coulometer
(Metrohm). Activation of the catalyst at 400 ¡C consisted of a
1 h outgassing step under argon Ñow (30 ml min~1) and a 1 h
reduction with hydrogen (15 ml min~1) with careful tem-
perature control by slowly increasing the hydrogen content of
the gas stream.
2. Systems analysis
Current gasoline ICE technology reaches a full fuel cycle
(““well-to-wheelÏÏ) efficiency of about 17È18%.11 This Ðgure
can be broken down into an overall vehicle (““tank-to-wheelÏÏ)
efficiency of 19È20% and a fuel efficiency (““well-to-tankÏÏ) of
90%. Fuel efficiencies for methanol production from natural
gas are in the range 67È71%. The overall vehicle efficiency for
a methanol powered car must therefore reach at least 25È27%
in order to equal the efficiency of the full gasoline ICE cycle.
Steady-state system analyses, carried out at our institute,
including an autothermal methanol reformer linked to a PEM
fuel cell yielded overall vehicle efficiencies of 27È30%
assuming anode rejected hydrogen of 17%. Hohlein11 investi-
gated other options for on-board methanol reforming and
obtained overall vehicle efficiencies in the range 25È34%.
The efficiency cascade for on-board autothermal methanol
reforming, starting with methanol and ending with the net
electric power delivered to the motor, is represented graphi-
cally in Fig. 1. The values on the left represent the cumulative
efficiency while those on the right represent the efficiencies of
the individual sub-processes. Only 4% of the methanolÏs
energy content is lost as heat during the reforming and gas
clean-up process (preferential CO oxidation). Most of the
losses occur in the fuel cell itself. Current fuel cell technology
does not allow complete utilisation of hydrogen from the
reformer if delivered to the anode as a gas mixture as com-
pared to full utilisation of pure hydrogen stored on-board.
Furthermore, only about 50% of the hydrogen converted in
the fuel cell produces electricity. About 20% of the electric
power generated by the fuel cell has to be used for driving
auxiliary equipment (compressors, pumps) which reduces the
net electricity available for the motor to 33% of the energy
content in the methanol fed. Any improvements along the fuel
processing chain will increase the overall efficiency. This
analysis suggests that the fuel cell itself has the biggest poten-
tial for increasing the overall efficiency by increasing the
hydrogen utilisation and reducing losses from hydrogen con-
version to electricity.
3.2. Results of catalyst testing
Three di†erent commercial copper-containing catalysts were
tested for their activity for the autothermal methanol reform-
ing reaction. Catalyst A and B were of the formulation
CuO/ZnO/Al O whereas Catalyst C was a Cu/Al O cata-
2
3
2 3
lyst. Feed Ñows were adjusted according to the stoichiometry
of reaction (4) with additional argon to model the nitrogen
content of air.
Results are summarised in Table 1, showing that all three
catalysts are active for the autothermal reforming of meth-
anol, with catalyst A and B being more active and selective
than catalyst C. Methanol conversions for these two catalysts
reached almost 100% and hydrogen yields of 85%
(corresponding to 90% lower heating value (LHV) efficiency)
could be measured. Non-converted oxygen could never be
detected in the product gas. While catalyst A showed almost
no deactivation in terms of hydrogen production after 100 h, a
slight activity loss of 9% was observed for catalyst B. Carbon
monoxide concentrations in the range 0.3 to 3.5% were mea-
sured, with a strong dependence on temperature and space
velocity. The extrapolated numbers for spaceÈtime yield of
around 20 000 l
thermal power density of 60 kW per litre of catalyst.
(h l
)~1 can be correlated to a
H2
REACTOR
th
3.3. Model of the reaction network
For further experiments, eventually aiming at the develop-
ment of a kinetic model for the autothermal reforming of
methanol, catalyst B was chosen. From the species being
present in either the feed or the product gas, a system of at
least 7 components and 8 possible reactions can be set up (see
Fig. 2). The reaction scheme is strongly crosslinked since most
of the species occur in more than one reaction and the pro-
ducts of one reaction might act as the reactant for other reac-
tions. From these considerations it is highly desirable to
simplify the reaction network.
It is known that all reactions incorporating molecular
oxygen ((5), (6), (11) and (12); reaction numbers relating to Fig.
2) are fast and highly exothermic. On the contrary, the reverse
water-gas shift reaction (10) is very slow and equilibrium
limited. From measurements showing lower carbon monoxide
concentrations than determined by equilibrium calculations it
can be concluded that the steam-reforming reaction (9), being
the major source of hydrogen in the system, is a single-step
Fig. 1 Calculated steady-state ““tank-to-motorÏÏ efficiency for auto-
thermal reforming of methanol and subsequent gas clean-up by pref-
erential oxidation of CO. The values on the left represent the
cumulative efficiency while the ones on the right represent the effi-
ciencies of the individual sub-processes.
290
Phys. Chem. Chem. Phys., 2001, 3, 289È293

View Article Online
Table 1 Catalyst screening results for the autothermal reforming of methanol
Furnace
temperature
/¡C
WHSV
/h~1 a
Conversion
(%)
Yield H
(%)
c
c
H -Production
MeOH
2
CO
DME
2
Hot-Spot/K
(%)
(%)
/lH2(h l
)~1
REACTOR
Commercial Cat Ab
280
6
12
18
18
12
12
2
8
20
25
6
99.9
99.3
89
70
99.9
93
84
83
80
61
87
84
2.7
1.3
0.6
0.3
3.6
0.6
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
9700
19 500
27 800
21 000
20 100
19 400
250
330
280c
12
Commercial Cat Bd
280
6
12
18
18
12
12
2
8
21
25
6
99.5
94
82.5
70
99.9
91
83
85
58
40
84
78
1.3
0.6
0.4
0.3
2
0.02
0.03
0.16
0.1
0.2
0.1
10 900
22 500
23 000
15 900
22 100
20 500
250
330
280c
11
0.5
Commercial Cat Ce
330
300
12
12
3
3
96
93
78
72
3.5
3.5
n.m.
n.m.
11 600
10 600
a WHSV: weight-hourly space velocity in mass methanol per time and mass of catalyst. b m
CAT
\ 100 mg, l \ 14 mm.
\ 162.4 mg. c Deactivation check-point after 100
h run time. n.d.: none detected, n.m.: not measured. d m
\ 185.5 mg. e m
CAT
CAT
reaction, i.e. not a consecutive reaction consisting of methanol
decomposition (7) and water-gas shift (10). The dimethyl ether
(DME) formation (8) can be treated independently from the
other reactions if DME is considered not to react further since
the active sites for this reaction are unique (acidic sites of
Al O ). Additionally, it is assumed that the catalyst can be
modiÐed by cationic ion exchange or support composition,
resulting in decreased or even zero DME formation activity.
Generally, the autothermal reactor can be divided into two
parts. The Ðrst part with oxygen present in the gas has an
exothermic overall reaction whereas the second part has an
endothermic overall reaction. It is obvious that in the second
part the steam-reforming reaction is mainly dominant. For the
Ðrst part of the reactor two simpliÐed models for the mecha-
nism of oxygen consumption can be discussed: the pure
partial oxidation (POX) case and the total oxidation (TOX)
case. Both reactions can be combined with the steam-
reforming (SR) reaction to give the autothermal reforming
(ATR) reaction. The coefficients are shown in Table 2.
In order to distinguish between the models, experimental
conditions, i.e. temperature and space velocity, were chosen to
ensure oxygen conversion less than 100%. From the observed
hydrogen to carbon dioxide ratio close to zero in the product
gas, as shown in Fig. 3, it can clearly be concluded that, in the
presence of oxygen, the methanol conversion occurs mainly
via the TOX reaction, the hydrogen being produced by sub-
sequent steam reforming. This observation, made at 223È
232 ¡C was veriÐed later at a reactor temperature of 250 ¡C.
In comparison to this, it is reported for the reaction of
water-free mixtures of methanol with oxygen over
2
3
Fig. 3 H /CO ratio in the product gas for di†erent oxygen conver-
2
2
sions. The lines show the expected H /CO ratio for TOX
2
2
(nH2/nCO2 \ 0), POX (nH2/nCO2 \ 2) and SR (nH2/nCO2 \ 3). Conditions:
53 mg of Catalyst B; stoichiometric feed for ATR with 1.5 ml h~1
methanolÈwater mixture; temperatures 222, 228 and 232 ¡C.
Fig. 2 Considered reactions for the autothermal methanol
reforming.
Table 2 Coefficients of linear combination for two models of autothermal reforming
POXÈSR
TOXÈSR
SR
CH OH ] H O ] CO ] 3 H
3
1
0
32
3
3
2
2
2
2
POX
TOX
&: ATR
CH OH ] 1 O ] CO ] 2 H
0
3
2
2
2
CH OH ] 3 O ] CO ] 2 H O
1
3
2
2
2
2
3
4 CH OH ] 3 H O ] 1 O ] 2 N ] 4 CO ] 11H ] 2 N
3
2
2
2
2
2
2
2
Phys. Chem. Chem. Phys., 2001, 3, 289È293
291

View Article Online
Table 3 Results of the kinetic measurements on Catalyst B
RWGS
SR
T /K
513É É É553
503É É É548
m Catalyst/mg
k(513 K)
511
197.5/77.6
4.38 ] 10~8 mol (s g kPa)~1 (^5%)
1.52 ] 10~5 mol (s g)~1 kPa~0.4 (^2%)
E /kJ mol~1
123(^5%)
83(^3.5%)
A
Cu/ZnO/Al O that hydrogen is produced from methanol at
free sites was considered to be a constant. This leads to the
following rate eqn. (13)
2 3
a
maximum for substoichiometric oxygen to methanol
ratios.12 Further increase of the oxygen partial pressure to the
stoichiometric value causes hydrogen production to decrease,
probably because of surface oxidation of the catalyst.
Reduction of the oxygen partial pressure could not recover
the hydrogen production activity of the catalyst.
C
E
R
A
1
1
BD
A
r
(T ) \ k(513K) exp
RWGS
T 513K
A
p
p
B
H2O CO
] p
1 [
(13)
CO2
KP
(T )pCO2 pH2
In order to answer the question, whether the small detected
amount of hydrogen shown in Fig. 3 is produced by POX or
SR, the experiment was repeated using the same conditions
but exchanging H O in the feed by D O. The mass spectrum
RWGS
The measured activation energy of 123 kJ mol~1 for this reac-
tion is higher than the literature value of 95 kJ mol~1.13
For the SR reaction a simple power law Ðt was chosen since
the LangmuirÈHinshelwood models described in ref. 14 and
15 consider the catalyst surface hydrogen concentration to
always be in the equilibrium state, which means that those
models cannot be applied for measurements with zero hydro-
gen concentration in the feed. Formal orders in methanol and
water were determined to be 0.4 and 0. This is in good agree-
ment with the formal orders described in ref. 16. The full rate
equation is given in eqn. (14).
2
2
of the product gas clearly revealed that hydrogen production
in the presence of oxygen is at least partially occurring via the
SR reaction. However, the general non-existence of the POX
reaction could not be concluded from this experiment. Never-
theless, it could be shown that the combination of the TOX
and the SR reactions is an appropriate model for the meth-
anol conversion by autothermal reforming of methanol. Con-
sidering the reverse water-gas shift reaction allows the carbon
monoxide formation to be modelled.
C
E
R
A
1
1
BD
A
r (T ) \ k(513K) exp
3.4. Kinetic measurements
SR
T 513K
Based on the proposed simpliÐed model for the autothermal
reforming of methanol over catalyst B, kinetic measurements
of methanol SR and the reverse water-gas shift (RWGS) reac-
tion have been performed separately in the microreactor
system with feed components being methanolÈwater for the
SR reaction and hydrogenÈcarbon dioxide for the RWGS
reaction. Other conditions and results of the measurements
are listed in Table 3. For evaluation of the data obtained for
both reactions the software package SIMUSOLV and a one-
dimensional plug-Ñow model were used. Possible equilibrium
limitation was included by introducing temperature-
dependent equilibrium terms.
A
p
CO2 pH32
B
] p0.4 1 [
(14)
MeOH
KP (T )p
p
SR
MeOH H2O
The parity plot (Fig. 4) for the SR reaction shows that the
model describes the measured data reasonably well.
Describing the TOX of methanol in the presence of water,
being the third reaction considered in the model is more diffi-
cult. From preliminary results it can be suspected that the
catalyst surface is changed in the presence of oxygen. The
number and/or adsorption properties of the active Cu sites
seem to be dependent on oxygen partial pressure. Further-
more, the high reaction enthalpy complicates isothermal mea-
surements at elevated temperatures in the tubular reactor.
The RWGS reaction was assumed to be Ðrst order in CO
as described in ref. 13. Due to the low pressure the number of
2
4. Conclusions
It has been shown that the autothermal reforming process of
methanol is a promising option for hydrogen production. If
used for non-stationary PEM fuel-cell-based propulsion, the
system can at least compete with the 17% full fuel cycle effi-
ciency of current state-of-the-art ICEs. Further optimisation
of single process steps might even improve this Ðgure,
however, the clear beneÐt of Ðrst generation fuel cell vehicles
will be in the Ðeld of emission reduction.
The microreactor results show that the reaction can be
carried out over commercially available copper catalysts at
temperatures around 300 ¡C, provided that deactivation is not
signiÐcant. The observed high conversions and selectivities are
very important for non-stationary reactor systems which, for
reasons of simplicity, must work in the once-through oper-
ation mode. A closer look at the reaction network suggests
that the total oxidation of methanol rather than the partial
oxidation is the major oxygen-converting reaction. Yet, in
contrast to the steam-reforming and reverse water-gas shift
reaction, isothermal kinetic measurements of this reaction are
difficult to perform under realistic conditions.
Fig. 4 Parity plot for methanol conversion in the SR reaction using
the model described above. Good agreement between measured and
Ðtted data is shown.
292
Phys. Chem. Chem. Phys., 2001, 3, 289È293

View Article Online
7
8
9
J. Amphlett, K. Creber, J. Davis, R. Mann, B. Peppley and D.
Stokes, Int. J. Hydrogen Energy, 1994, 19, 131.
5. Acknowledgements
S. Ahmed and R. Kumar, Annual Automotive T echnology Con-
tractorsÏ Meeting, Dearborn, Michigan, 1995.
The authors thank Johnson-Matthey plc (UK), Sudchemie
AG (D) and Haldor-Topsoe A/S (DK) for the supply of cata-
lyst samples under non-analysis agreements.
J. W. Jenkins and E. Shutt, Platinum Metals Rev., 1989, 33, 118.
10 N. Edwards, S. Ellis, J. Frost, S. Golunski, A. van Keulen, N.
Lindewald and J. Reinkingh, J. Power Sources, 1998, 71, 123.
11 B. Hohlein, Minutes of the 2nd IEA Advanced Fuel Cell W ork-
shop on ““Fuel Processing for Modular (O100 kW ) Fuel Cell
Power PackagesÏÏ, September 29ÈOctober 1, 1997, Wislikofen,
Switzerland.
References
12 L. Alejo, R. Lago, M. Pena and J. Fierro, Appl. Catal. A, 1997,
1
V. Ho†mann, W assersto†ÈEnergie mit Zukunft, Teubner-Verlag,
Stuttgart, 1994.
162, 281.
13 K. Vanden Bussche and G. Froment, J. Catal., 1996, 161, 1.
14 C. Jiang, D. Trimm and M. Wainwright, Appl. Catal., A, 1993, 97,
145.
2
3
4
5
G. Berry and S. Aceves, Energy Fuels, 1998, 12, 49.
K. Adamson and P. Pearson, J. Power Sources, 2000, 86, 548.
Y. Borodko and G. Somorjai, Appl. Catal. A, 1999, 186, 355.
B. Hohlein, M. Boe and J. Bogild-Hansen, J. Power Sources,
1996, 61, 143.
15 B. Peppley, J. Amphlett, L. Kearns and R. Mann, Appl. Catal. A,
1999, 179, 31.
16 C. Jiang, D. Trimm and M. Wainwright, Appl. Catal. A, 1993, 93,
245.
6
J. Yamaguchi, Automotive Eng. Int., October 1999, 74.
Phys. Chem. Chem. Phys., 2001, 3, 289È293
293