The water-gas-shift reaction at short contact times

Article abstract of DOI:10.1016/j.jcat.2004.01.002

We have examined the water-gas-shift reaction over noble metals and metals with ceria for catalyst contact times between 0.008 and 0.05 s for temperatures from 300 to 1000°C. With reactants of CO, H₂, and H₂O at a 1/2/4 composition, the reaction approached equilibrium at these contact times for temperatures as low as 380°C, which corresponds to a 50/1 H ₂/CO ratio. All results follow the equilibrium conversion versus temperature curve at high temperatures and then the rate decreases rapidly as temperature decreases. All metals show considerable promotion upon ceria addition, with Pt/ceria showing the most promise in terms of promotion, stability, and lack of methanation activity. A simple kinetic model that assumes first-order reversible reaction in all species fits all data at all temperatures from equilibrium to low conversions. The rate is nearly independent of surface area over a wide variation of parameters, and rates indicate that mass-transfer limitations are not dominant.

Full text of DOI:10.1016/j.jcat.2004.01.002

Journal of Catalysis 223 (2004) 191–199
www.elsevier.com/locate/jcat
The water–gas-shift reaction at short contact times
C. Wheeler, A. Jhalani, E.J. Klein, S. Tummala, and L.D. Schmidt ∗
Department of Chemical Engineering and Materials Science, University of Minnesota, 421 Washington Ave SE, Minneapolis, MN 55455, USA
Received 3 October 2003; revised 24 December 2003; accepted 6 January 2004
Abstract
We have examined the water–gas-shift reaction over noble metals and metals with ceria for catalyst contact times between 0.008 and 0.05 s
◦
for temperatures from 300 to 1000 C. With reactants of CO, H , and H O at a 1/2/4 composition, the reaction approached equilibrium
2
2
◦
at these contact times for temperatures as low as 380 C, which corresponds to a 50/1 H /CO ratio. All results follow the equilibrium
2
conversion versus temperature curve at high temperatures and then the rate decreases rapidly as temperature decreases. All metals show
considerable promotion upon ceria addition, with Pt/ceria showing the most promise in terms of promotion, stability, and lack of methanation
activity. A simple kinetic model that assumes first-order reversible reaction in all species fits all data at all temperatures from equilibrium
to low conversions. The rate is nearly independent of surface area over a wide variation of parameters, and rates indicate that mass-transfer
limitations are not dominant.
2004 Elsevier Inc. All rights reserved.
Keywords: Water–gas shift; Short contact times; Pt; Ceria; Elementary-step mechanism; Noble metals; Methanation; First-order kinetics
1. Introduction
The need for H2 for fuel cells and the hydrogen economy
has caused a renewed interest in these processes because
neither steam reforming nor conventional water–gas-shift is
easily scaled down to smaller applications. This limits their
applicability to large industrial systems and rules out their
use in smaller systems such as transportation and portable
power.
The research in our laboratory over the past decade has
focused on understanding partial oxidation reactions at short
contact times on noble metal catalysts and in this paper we
examine the water–gas-shift reaction at short contact times.
The partial oxidation of methane
The water–gas-shift (WGS) reaction is crucial in produc-
ing H2 from hydrocarbons, and it has been used industrially
for nearly a century [1]. The conventional process for gener-
ating H2 uses steam reforming
CH4 + H2O ꢀ CO + 3H2 ꢀH = 205.8 kJ/mol
(1)
on Ni catalysts at ∼ 800 ◦C. This reaction is endothermic
and reversible, so high temperatures are required for high
conversions.
If pure H2 is desired, the CO must be further reacted with
H2O to form CO2 and an additional mole of H2 in the water–
gas-shift reaction
1
2
CH4 + O2 → CO + 2H2 ꢀH = −36 kJ/mol
(3)
can be carried out with high conversions of methane and
high selectivities to CO and H2 on Rh catalysts at contact
times as short as 10−3 s [2,3]. This process can be run au-
tothermally, thus eliminating the need for external heat as
required in steam reforming. Partial oxidation can be easily
scaled up or down by large factors for different applications,
and it responds rapidly to transient operation at response
times < 5 s [4].
In order to develop a compact reactor for H2 production
using a combination of partial oxidation, water–gas-shift and
preferential oxidation, one needs to improve the water–gas-
shift reactor or reactors to shorten the overall residence time
CO + H2O ꢀ CO2 + H2 ꢀH = −41.2 kJ/mol.
(2)
This reaction is mildly exothermic and reversible and re-
quires low temperatures for high equilibrium conversions.
The conventional water–gas-shift catalysts which have been
used in industry for nearly a century are Fe–Cr2O3 at
∼ 500 ◦C followed by Cu–ZnO at ∼ 200 ◦C. Both of these
processes occur in packed-bed reactors with residence times
of ∼ 1 s, and they attain nearly equilibrium compositions [1].
*
Corresponding author.
E-mail address: schmidt@cems.umn.edu (L.D. Schmidt).
0021-9517/$ – see front matter
doi:10.1016/j.jcat.2004.01.002
2004 Elsevier Inc. All rights reserved.

192
C. Wheeler et al. / Journal of Catalysis 223 (2004) 191–199
while maintaining high conversions and assuring stability
under transient operating conditions. If thermodynamic con-
version can be achieved in water–gas-shift at these short
contact times at temperatures ∼ 350 ◦C, then the product
stream can be fed directly to a preferential oxidation reac-
tor for further CO reduction to ppm levels. Fe and Cu are
unsuitable for these applications because Fe readily forms
coke in excess fuel and Cu is irreversibly deactivated if ex-
posed to O2 [1]. Long reactor residence times also imply
long response times to startup and transient load conditions
because of the high thermal mass of conventional water–gas-
shift catalysts.
There has been considerable interest in noble metal cat-
alysts to overcome these problems [5], and promoters such
as ceria have been shown to increase the activity and low
temperature performance of noble metals [6]. Gorte and
co-workers have examined many noble metals under vari-
ous conditions and have characterized their kinetics [7–9].
They examined mostly low-temperature, low-conversion sit-
uations and found that the reaction was nearly zero order
in PCO. There have been few published studies of high-
conversion situations near reaction equilibrium which is the
situation used in practical applications [5,6].
It is the intent of this paper to examine the WGS reaction
at shorter times and higher temperatures to determine how
kinetics change with temperature and how different flow
conditions and catalyst flow patterns affect the process. We
also examine kinetics of these reactions which are applicable
over a wide range of temperature. In particular, we search for
indications of mass-transfer and pore-diffusion limitations
which indicate limits on surface area and on metal and ceria
loadings.
A possible integrated short contact time H2 generation re-
actor, starting with a catalytic partial oxidation reactor oper-
ating at 1100 to 900 ◦C followed by a water–gas-shift reactor
in which temperature decreases progressively is sketched in
Fig. 1a. In these experiments, we simulate this situation by
examining the water–gas-shift reaction in a nearly isother-
mal reactor whose inlet is H2/CO/H2O. We did not use CH4
and O2 reactants because we also wanted to study methana-
tion on the catalysts at these temperatures and contact times.
Fig. 1. (a) A possible combined catalytic partial oxidation (CPO) and wa-
ter–gas-shift (WGS) reactor with a possible temperature profile along the
reactor. (b) Apparatus used for water–gas-shift experiments. The reactor is
a quartz tube with an inner diameter of 18 mm. Catalyst supports are ei-
ther 10 mm alumina foam monoliths or alumina spheres of 1 mm diameter.
Both types of support are held in place between two blank foam monoliths.
Temperatures are measured at the inlet and outlet of the catalyst bed.
ports, while the activity of Pt with ceria was also examined
on high and low area spheres.
Monoliths had a diameter of 17 mm, 10 mm length, and
a pore density of 80 pores per linear inch with a void frac-
tion of 0.8. The monoliths were wash-coated using 3 wt%
slurry of γ -aluminum oxide in distilled water followed by
drying and heating at 600 ◦C for 6 h in a closed furnace.
Wash coats were applied to increase the surface area of the
monoliths. Typical wash coats were 5% by weight of mono-
lith, for which we estimate an average alumina film thickness
of 10 µm.
Metals were then coated on the supports using an aqueous
solution of metal salts (H2PtCl6, Rh(NO3)3, RuCl3, PdCl3,
Ni(NO3)3). The solution was dripped onto the support re-
peatedly, allowing the water to evaporate between applica-
tions. The monolith was then heated at 400 ◦C for 6 h in a
closed furnace. A calculated amount of metal salt was used
to ensure a 5 wt% metal loading based on the mass of the
monolith.
2. Experimental
2.1. Catalyst preparation
For the catalysts combining metal with ceria, ceria was
added before the metals, using a solution of Ce(NO3)3, fol-
lowed by heating for 6 h at 400 ◦C in a closed furnace.
Unless otherwise noted, a ceria loading of 5% by weight was
maintained on the monolith.
For Pt and Pt/ceria we also used high and low surface
area, 1-mm-diameter alumina spheres as supports. The cat-
alyst preparation and experimental procedure for the sphere
bed were analogous to those described above for monoliths.
No wash coat was applied to the spheres unless otherwise
We used three types of catalyst supports in these ex-
periments: (1) alumina ceramic foam monoliths containing
8% silica from Vesuvius Hi Tech Ceramics which had a
nominal surface area of ∼ 1.0 m2/g, (2) high surface area,
1-mm-diameter alumina spheres which had a nominal BET
surface area of ∼ 170 m2/g, and (3) low surface area, 1-mm-
diameter alumina spheres which had a nominal BET surface
area of ∼ 5 m2/g. The behavior of various metals, both with
and without ceria, was compared on the foam monolith sup-

C. Wheeler et al. / Journal of Catalysis 223 (2004) 191–199
193
indicated. Again, the metal loading on spheres was typically
5 wt% for both Pt and ceria.
predicted by the ideal gas law, and thus the contact time de-
creases with increasing temperature. For most experiments,
temperatures were measured at the outlet of the catalyst bed.
Due to the exothermicity of the reaction and possible heating
and cooling of gas mixture with position, the outlet tem-
perature was usually higher by 20 to 120 K than the inlet,
depending on conversions and temperature. Reported tem-
peratures in figures are those at the outlet of the catalyst bed.
Many data points shown lie slightly above the equilibrium
curve, and we attribute this discrepancy to recording conver-
sions at the exit temperature.
Gas samples were taken through a septum at the exit of
the reactor. A gas-tight syringe was used to withdraw prod-
uct samples and inject it into a gas chromatograph equipped
with thermal conductivity detector. Mass balances on carbon
and hydrogen typically closed to within 5%.
We took no special precautions to assure uniform coat-
ings except for repeated loading and drying of the catalyst,
and we regard loadings reported to be accurate to within
20%. As will be shown and discussed later, results were
not strongly sensitive to details of preparation methods, such
as order of metal and ceria deposition, and heating tempera-
ture.
For some samples, areas were measured by BET analysis
at various steps in catalyst preparation and use in the reactor.
Total surface area did not change drastically with the addi-
tion of wash coat, ceria, or other metals. Areas varied by less
than 50% overall with or without the wash coat and catalyst
deposition, and before and after reaction.
All experiments were repeated several times on all cata-
lysts, with no significant differences between experiments
except where noted. All catalysts were heated to at least
800 ◦C repeatedly during use. All catalysts were used for
at least 10 h at different temperatures, and some catalysts
were used for as long as 30 h. Most experiments were also
repeated on several nominally identical samples with no sys-
tematic differences or deactivation noted.
2.2. Experimental procedure
All experiments were carried out in a tube furnace that
could be controlled at temperatures up to 1200 ◦C. As de-
picted in Fig. 1b, the coated monolith or spheres were placed
in the reactor between two blank monoliths. The monoliths
were sealed into the reactor with alumina cloth. Because it
was difficult to adjust the H2O vapor flow rate, all experi-
ments were carried out at 3 SLPM total flow rate with a feed
composition of CO/H2/H2O of 1/2/4 and 20% N2 dilution.
Reactant gases were a 2/1 H2/CO premixed syngas mix-
ture and steam was added from a vaporizer. N2 (20 mol%)
was also fed for use as an internal standard for GC analysis.
The syngas and N2 flow rates were controlled using mass
flow controllers. The gases were heated in a tube wrapped
in heating tape prior to being fed to the reactor. Liquid wa-
ter was delivered via an ISCO syringe pump, and steam was
generated in a heater at approximately 700 ◦C. The total flow
rate was set at 3 SLPM, and the molar composition of the re-
actant inlet stream was 22.9% H2, 11.4% CO, 45.7% H2O,
and 20% N2. The major cause of fluctuation in flow and tem-
perature was intermittent condensation and evaporation of
water in the system. So all experiments were conducted at a
single flow rate.
3. Results
3.1. Platinum and Pt/ceria
We first show typical results from Pt and Pt/ceria on
alumina foam monolith supports because they indicate the
general trend for all experiments. Fig. 2 shows CO conver-
sion versus catalyst temperature on Pt, Pt/ceria, and ceria. As
The experimental procedure consisted of heating the sys-
tem to 800–1000◦C such that equilibrium conversions were
achieved, and then reducing the temperature in steps un-
til negligible conversion of CO was observed. Gas samples
were taken at intervals as the temperature was lowered. For
each catalyst, this temperature variation procedure was car-
ried out at least three times to verify reproducibility. Any
variations in the data were largely attributed to inconsistency
in the steam flow rate due to condensation and subsequent
evaporation as noted previously. Blank experiments without
catalytic material on the support were occasionally run to
search for homogeneous reactions or reactions on impurities
in the reactor. In all the cases, CO conversions without cata-
lyst was no more than a few percent at 1000 ◦C.
Fig. 2. CO conversions in water–gas-shift reaction on Pt, Pt/ceria, and ceria
catalysts supported on an alumina foam monolith. Catalyst loading is 5%
for Pt and 5% Pt/5% ceria for Pt/ceria based on the weight of the support.
The upper dotted line represents equilibrium conversions for a feed compo-
sition of 1/2/4 for CO/H /H O with 20% N dilution. The lower dotted
2
2
2
line represents equilibrium conversions using the empirical expression of
Eq. (7). The solid lines are model fits assuming first-order reversible kinet-
Because the molar flow rate of reactants was held con-
stant, the volumetric flow rate varies with temperature as
ics with respect to CO and CO , Eq. (10), as discussed in the text. Arrows
indicate catalyst contact times at 300, 600, and 900 C for this flow rate.
2
◦

194
C. Wheeler et al. / Journal of Catalysis 223 (2004) 191–199
noted above, a feed mixture of CO/H2/H2O at 1/2/4 with
20% N2 dilution and 3 SLPM total flow was used for the ex-
periments, and temperatures were recorded at the back face
of the catalyst in the furnace.
have been reported for ceria-supported Pd, Pt, and Rh [7,8].
For any mechanism, detailed balance requires that kinetics
for this reaction approach first order in all species near ther-
modynamic equilibrium.
The catalyst support for these experiments was alumina
foam monolith 10 mm long. A 3 SLPM total flow corre-
sponds to catalyst contact times from 17 ms at 300 ◦C, 12 ms
at 600 ◦C, to 9 ms at 900 ◦C. Pt follows the equilibrium con-
version down to ∼ 1000 ◦C and then falls rapidly so that
the conversion is nearly zero by 500 ◦C. Pt/ceria follows
the equilibrium curve down to ∼ 600 ◦C and falls nearly to
zero by 350 ◦C. Ceria alone gave very low conversions at the
highest temperature as shown in the figure.
The upper dashed curve in Fig. 2 shows the exact equi-
librium curve calculated using Chemkin code EQUIL. The
lower dashed curve in Fig. 2 shows the equilibrium calcu-
lated for this reactant composition using the expression
We further simplify this rate expression by assuming that
H2O and H2 are nearly constant as CO and CO2 change
area
r =
rꢀꢀ = kfPCO − kbPCO
2
volume
ꢀ
ꢁ
X
= kfPCO 1 − X −
.
(9)
O
Keq
The pseudo-homogeneous rate r is related to the sur-
face reaction rate rꢀꢀ by the expression r = (area/volume)rꢀꢀ,
where (area/volume) is the area of active catalyst per unit
volume of reactor. This ratio is poorly defined for our system
because the total and metal areas are not known accurately
and because for ceria addition the reaction probably occurs
on both metal and ceria surfaces. For this simple rate expres-
sion, the mass balance equation can be integrated assuming
plug flow to give
ꢀ
ꢁ
ꢀHR
RT
[CO2][H2O]
[CO][H2O]
Keq = Ko exp −
=
,
(4)
where
ꢂ
ꢃ
kf
1 − e−(k +k )t
,
(10)
f
b
X(t) =
kf
Ko =
kb
(kf + kb)
(5)
(6)
(7)
where t is the residence time in the reactor, which varies
with temperature as predicted by the ideal gas law. This is
the expression used to fit all data shown in this paper. The
pseudo-homogeneous rate coefficients are defined as
and
ꢀHR = Ef − Eb.
For our reactant composition, this becomes
ꢀ
ꢁ
area
area
Ef
ꢀꢀ
kf =
kfꢀꢀ =
k exp −
(11)
of
X(2 + X)
volume
volume
RT
Keq
=
.
(1 − X)(4 − X)
so that the preexponentials kof and kob have units of s−1
.
The empirical equilibrium curve in Fig. 2 deviates from
the exact equilibrium curve because reactions forming CH4
are excluded from the empirical curve expression, Eq. (4). In
all subsequent figures, only the empirical equilibrium curve
will be shown since it gives a lower limit on CO conversion
that can be achieved in our reactor. Later in the paper we
show methanation equilibrium along with methanation data
for different catalysts to quantify the amount of CO that is
converted to CH4.
The reaction rate is a complex function of rate coefficients
of elementary steps (16 preexponential factors and 16 activa-
tion energies for a 8-step reversible process discussed later),
and the rate depends on concentrations of all reactants, prod-
ucts, and gas and surface intermediate species. In fitting the
data, we use a very simple rate expression approximating the
process as a single reversible surface reaction with rate ex-
pression
The curves shown along with the data points in Fig. 2
are calculated with this rate expression. A plot of lnX ver-
sus 1/T gave a straight line and so the activation energy
was set constant at all temperatures. The forward and the
backward rate constants were chosen to satisfy the equilib-
rium relation. The fits shown are for kof = 1.0 × 106 s−1
and Ef = 80 kJ/mol for Pt and kof = 2.5 × 107 s−1 and
Ef = 80 kJ/mol for Pt/ceria. A visual inspection is sufficient
to see that these curves give fits to the experimental points to
within the accuracy of the data.
For all data shown in this paper, experimental points can
be fit with this simple rate expression, Eq. (10). We will dis-
cuss the surprising agreement of all data on all catalysts with
this expression in connection with the reaction mechanism.
3.2. Comparison of metals
In addition to Pt, we examined Rh, Ru, Pd, and Ni under
identical conditions. A plot showing results for each of these
metals is shown in Fig. 3. The graph shows the conversion
of CO to produce CO2. As before, the total flow rate for
these experiments is 3 SLPM, resulting in a contact time of
∼ 13 ms at 500 ◦C.
rꢀꢀ = kfꢀꢀPCOPH O − kbꢀꢀPCO PH
.
2
(8)
2
2
This expression assumes an elementary reaction with first-
order kinetics with respect to all species. There have been
many studies of kinetics of water–gas-shift reaction [7–11],
and many of them report kinetics that are approximately first
order in all species, either measured or assumed. For CO at
lower temperatures and conversions, kinetics near zero order
As shown on the plot, the effectiveness of the metal cat-
alysts follows the order Ni > Ru > Rh > Pt > Pd. The

C. Wheeler et al. / Journal of Catalysis 223 (2004) 191–199
195
Fig. 4. CO conversions in water–gas-shift reaction on Ru/ceria, Ni/ceria,
Pt/ceria, Rh/ceria, and Pd/ceria supported on alumina foam monolith. Cat-
alyst loading is 5% metal/5% ceria based on the weight of the support. The
dotted line represents equilibrium conversions for a feed composition of
1/2/4 for CO/H /H O with 20% N dilution using Eq. (7). The solid lines
Fig. 3. CO conversions in water–gas-shift reaction on Ru, Ni, Rh, Pt, and Pd
supported on alumina foam monolith. Catalyst loading is 5% based on the
weight of the support. The dotted line represents equilibrium conversions
for a feed composition of 1/2/4 for CO/H /H O with 20% N dilution
2
2
2
2
2
2
using Eq. (7). The solid lines are model fits assuming first-order reversible
are model fits assuming first-order reversible kinetics with respect to CO
kinetics with respect to CO and CO , Eq. (10).
2
and CO , Eq. (10).
2
low that even with the addition of ceria, equilibrium was
achieved only down to ∼ 770 ◦C, yielding a CO conversion
of ∼ 62%.
Table 1
Preexponential factors, k , and activation energies, E , for metal and
metal/ceria combinations
of
f
−1
7
Although the effect of adding ceria to Ru is not as sig-
nificant as that on Pt and Pd, a comparison of Figs. 3 and 4
shows that it lowers the temperature of deviation from equi-
librium from 620 to 525 ◦C. This results in a CO conversion
of ∼ 82%, and a H2 to CO ratio approaching 16. The effects
of ceria on Rh and Ni are the smallest, allowing the conver-
sion to follow the equilibrium curve to only ∼ 50 ◦C lower
as compared to pure metal.
Metal or metal/ceria
k
(s
)
E (kJ/mol)
k
/k
of
f
of,metal/ceria of,metal
Ru
Ru/ceria
Ni
Ni/ceria
Rh
Rh/ceria
Pt
Pt/ceria
Pd
Pd/ceria
1.6 × 10
5.0 × 10
8.0 × 10
1.7 × 10
3.0 × 10
1.5 × 10
1.0 × 10
2.5 × 10
4.0 × 10
4.0 × 10
80
80
85
7
3.12
2.12
5
7
8
85
9
130
130
80
80
100
100
10
6
7
25
6
The kinetic fits for metal/ceria combination, also shown
in Fig. 4, agree reasonably well with the experimental data.
The constants for the model are tabulated in Table 1. It
should be noted that the model uses the same activation en-
ergy for metal and metal/ceria combination. The only effect
of ceria is to enhance the rate of the reaction by a factor equal
to the ratio of the forward rate constants for the metal/ceria
and metal catalysts.
7
10
maximum CO conversions are ∼ 80% with Ni, 75% with
Ru, 70% with Rh, and 45% with Pt. The results with Pd
were well below the equilibrium curve even at the highest
temperature measured, ∼ 35% at 900 ◦C.
As discussed above, the results shown in Fig. 3 were all fit
with a first-order, reversible kinetic model of Eq. (10). The
constants for the model are tabulated in Table 1.
3.4. Methanation
Methane can also be formed in the reactor from the re-
verse steam reforming reaction, Eq. (1). Not only is CH4
undesired in the product stream, but its formation also re-
sults in the loss of two H2 molecules for every CH4 molecule
produced, thus lowering the H2 to CO ratio.
3.3. Addition of ceria to metals
We also investigated the effect of adding ceria to each
of these metals, as shown in Fig. 4. The order is almost the
same as that on the metals alone, Ni ≈ Ru > Pt > Rh >
Pd, except that Pt and Rh are now reversed. The effect of
adding ceria to the Pt catalyst, shown in Fig. 2, is the most
pronounced. The equilibrium conversion is maintained down
to ∼ 600 ◦C with Pt/ceria, versus ∼ 1000 ◦C with Pt alone.
This results in an increase in CO conversion from 45% to
about 75%. The next largest effect of adding ceria was ob-
served with Pd. However, the activity of Pd alone was so
Fig. 5 shows the fraction of the CO that is converted
to methane, SCH , for each metal versus temperature. The
4
dashed line in the Fig. 5 represents the calculated equilib-
rium selectivity to CH4. The order in which the metals pro-
mote methanation is Ru > Rh > Ni > Pt > Pd. Methane
formation on Pt and Pd is less than 1%, while on the other
metals it is large, exceeding 10% on Ru. Thus, the maximum
H2 to CO ratio on catalysts with high methanation activity

196
C. Wheeler et al. / Journal of Catalysis 223 (2004) 191–199
Fig. 6. The effect of catalyst support for water–gas-shift reaction on Pt/ceria.
The catalyst loading in each case is 5% of the total support weight for
each metal. Type and weights of catalyst support and approximate residence
times at 400 C are indicated.
Fig. 5. Methane formation during water–gas-shift experiments on Ru, Ni,
Pt, Rh, Pd, and Ru/ceria. The dashed line represents the equilibrium CH
4
◦
selectivity calculated for a feed composition of 1/2/4 for CO/H /H O
2
2
with 20% N dilution.
2
is considerably lower than what the water–gas-shift equilib-
rium indicates.
We found that the addition of ceria had little effect on
the amount of methane formed at a particular temperature,
and results for Ru/ceria are shown in Fig. 5. One benefit
of Pt/ceria catalysts was that the methanation reaction was
found to be negligible at temperatures where the CO conver-
sion was the maximum.
For most of the metals examined, no effects of carbon
formation were noted from measured conversions. However,
while Ni gave the highest CO conversion, it also resulted
in the undesired promotion of carbon formation, which was
visible both on the surface of the reactor tube downstream of
the catalyst and on the foam monolith support itself after the
reaction.
Fig. 7. The effect of increasing the catalyst amount and contact time for wa-
ter–gas-shift reaction on Pt/ceria supported on high area alumina spheres.
The total amount of catalyst-coated spheres is varied from 2 to 10 g, cor-
3.5. Comparison of supports
◦
responding to a contact time at 400 C of 10 to 50 ms. The solid lines are
model fits assuming first-order reversible kinetics with respect to CO and
CO with k s proportional to catalyst amount.
Three types of catalyst supports were used in this study,
all alumina but covering a wide range of surface area, 1, 5,
and 170 m2/g for foam monolith, and low and high area
spheres, respectively. A catalyst loading of 5 wt% Pt and
5 wt% ceria was used for these sets of experiments. Re-
sults for the various supports are compared in Fig. 6. Contact
times at 400 ◦C are indicated in the figure for the supports.
Though results differ considerably among supports, but there
is little correlation of CO conversion with total area.
2
of
fits with kof s adjusted appropriately (linear dependence on
catalyst amount).
A significant increase in CO conversions is achieved by
lengthening the bed, as expected. The longest bed results in a
CO conversion of about 95% at a temperature below 400 ◦C,
and methane was not detectable in the product stream. Thus
the highest H2 to CO ratio attained exceeded 50 to 1.
3.6. Comparison of residence times
3.7. Effect of Pt and ceria loading
To examine the effect of changing the catalyst contact
time, we used a high area sphere bed of various lengths,
holding the flow rate constant at 3 SLPM. Results with 5% Pt
and 5% ceria loading are plotted in Fig. 7 along with model
To ascertain the effect of Pt and ceria loadings, experi-
ments were performed using the low surface area alumina
spheres with 5% Pt/5% ceria, 1% Pt/5% ceria, and 0.2%
Pt/2% ceria catalysts. Results are plotted in Fig. 8 and show

C. Wheeler et al. / Journal of Catalysis 223 (2004) 191–199
197
Fig. 9. Plot showing CO conversions on Rh for several models as described
in the text. All models can be made to fit the experimental results to within
the accuracy of the data.
Fig. 8. The effect of different Pt and ceria loadings on CO conversions. The
s vary by a factor of 4.4 as Pt loading varies by a factor of 25.
k
o
that a higher loading of metal and ceria gives slightly higher
CO conversions. The interesting point to be noted here is that
the kof’s vary by a factor of 4.4 as we move from 5% Pt/5%
ceria to 0.2% Pt/2% ceria, with 5% Pt/5% ceria having a
higher kof than 0.2% Pt/2% ceria. Thus we observe a non-
linear dependence of kof on catalyst loading as compared to
a linear dependence of kof on catalyst amount as discussed
earlier.
5. Mechanism
In this section we examine the mechanisms that may be
responsible for these results and attempt to fit the results with
different rate expressions. Several of these fits for water–gas
shift on Rh catalysts are shown in Fig. 9.
5.1. Noble metals
The mechanism on a noble metal is assumed to consist of
elementary steps such as the following sequence:
4. Discussion
To summarize these results, we have measured rates of
water–gas-shift reaction on five noble metal catalysts with
and without ceria on monoliths and on low and high sur-
face area alumina spheres from 300 to 1000 ◦C giving CO
conversions from 0 to 95%. All catalysts were prepared by
comparable methods and were heated to > 800 ◦C for at
least 1 h during use. Results were highly reproducible for
identically prepared samples and for repeated experiments
on a given sample over 10 h. No activation or deactivation
was observed except for the activation associated with the
reduction of metal oxides on oxidizable metals such as Ni.
All data that we have obtained in this study can be fit
to within the accuracy of the experiments with a simple
reversible reaction, which assumes first-order kinetics with
respect to CO and CO2. Results on each metal can be fit with
the same activation energy with and without ceria addition.
All experiments with Pt and Pt/ceria can also be fit with
the same activation energy for all three supports, with and
without wash coats, and for a factor of 5 variation in resi-
dence times.
(1) CO + M ꢀ CO–M,
(2) H2O + M ꢀ H2O–M,
(3) H2O–M + M ꢀ OH–M + H–M,
(4) OH–M + M ꢀ H–M + O–M,
(12)
(5) 2OH–M ꢀ H2O–M + O–M,
(6) CO–M + O–M ꢀ CO2–M + M,
(7) CO2 + M ꢀ CO2–M,
(8) H2 + 2M ꢀ 2H–M.
Here M is a vacant site on the noble metal surface, –M is
the occupied adsorption site, steps 1, 2, 7, and 8 are adsorp-
tion and desorption steps, and steps 3, 4, 5, and 6 are surface
reaction steps. We have written them all as reversible, and,
assuming that all are elementary, this mechanism suggests
that there are 16 rate coefficients consisting of 16 preexpo-
nential factors and 16 activation energies.
In the following sections we discuss the mechanisms con-
sistent with these experiments. We will consider detailed
chemistry and simplified Langmuir–Hinshelwood kinetics,
and we will discuss the effects of ceria on these reactions.
Since rates appear to be nearly independent of surface areas,
we will consider the effects of external mass transfer and
pore diffusion through wash coats.
5.2. Langmuir–Hinshelwood kinetics
This is the simplest form of solution to these steps. If
adsorption-desorption steps are fast and reversible and we
simplify the surface reaction steps into a single step
CO–M + H2O–M ꢀ CO2–M + H2–M
(13)

198
C. Wheeler et al. / Journal of Catalysis 223 (2004) 191–199
and assume it to be rate controlling, then the coverage θj of
species j is given by the Langmuir isotherm
5.4. Coverage-dependent activation energy
A fourth fit comes from assuming that ECO decreases
with θCO. If we assume
Kj Pj
ꢄ
θj =
.
(14)
1 + i KiPi
ECO = ECO − αθCO
,
(18)
Since hydrogen is dissociated, its isotherm should be pro-
O
portional to PH1/2, but for simplicity we will leave it as PH
.
we obtain good agreement between the detailed model
2
2
The rate of reaction (13) is given by
and the experiments using ECO = 125 kJ/mol and α =
O
30 kJ/(mol monolayer).
rꢀꢀ = kfꢀꢀθCO H O
θ
− kbꢀꢀθCO θH
.
2
(15)
2
2
With Langmuir isotherms for all species, this rate becomes
5.5. Carbon model
kꢀꢀKCOKH OPCOPH O − kꢀꢀKCO KH PCO PH
2
2
2
2
2
2
f
(1 +KCOPCO +KH OPH O +bKCO PCO +KH PH
)
2
rꢀꢀ =
.
Finally, we also fit these results using a more complete
partial oxidation model that includes reactions of CHx-
adsorbed species [16]. This involves 18 reversible elemen-
tary steps where the 8-step mechanism is a subset which
ignores all CHx species.
2
2
2
2
2
2
(16)
If all the coverages are small, the denominator is unity, and
the rate becomes our global expression, Eq. (8), with which
we have fit all of our data. If CO is the most strongly bound
compared to other species, the rate becomes
5.6. Noble metals/ceria
kꢀꢀKCOKH OPCOPH O − kꢀꢀKCO KH PCO PH
2
2
2
2
2
2
f
b
rꢀꢀ =
. (17)
In the presence of ceria, the rate is presumably greater
than on the noble metals alone because H2O can adsorb on
ceria rather than being blocked by a surface saturated with
CO. The mechanism with ceria might involve the steps:
2
(1 + KCOPCO
)
This allows the order in CO to be less than +1, and it
−1
becomes proportional to P as θCO approaches saturation.
CO
All of thꢀeꢀse ꢀeꢀ xpressions are consistent with equilibrium as
(1) CO + M ꢀ CO–M,
long as k /kb = Keq, the equilibrium constant for the water–
gas-shiftfreaction.
(2) H2O + Ce ꢀ H2O–Ce,
(3) H2O–Ce ꢀ O–Ce + H2,
(19)
5.3. Detailed chemistry
(4) CO–M + O − Ce ꢀ CO2–Ce + M,
(5) CO2–Ce ꢀ CO2 + Ce.
For partial oxidation reactions on Rh and Pt, we and
others have developed rate expressions for each of these
steps [12–14], again assuming that rate coefficients for each
species are independentof all coverages. All rate expressions
are obtained from surface science experiments [12,15], al-
though some of these have been adjusted to give optimum
fits to a wide range of partial oxidation experiments.
We have used these rate coefficients to predict the rates
of water–gas shift on Rh. We assumed a plug flow reactor
and solved the set of rate equations for 7 surface species
(CO, H2O, H, O, OH, CO, and CO2) and 4 gas phase species
(CO, H2O, CO2, and H2). These equations were integrated
versus time up to the residence time t in the reactor. Since
area/volume is not known, the time becomes an adjustable
parameter in the calculation.
Since the overall activation energy is observed to be the
same for each metal with or without ceria, this suggests that
the effect of ceria is to increase rate of the reaction between
adsorbed CO on the metal and O on ceria, step (4) above. If
the activation energy of this step is small, the overall activa-
tion energy of the process might be unaffected, as observed.
We suggest the possibility that, even without ceria, some
reaction steps might occur on the support. H2O and CO
could adsorb on alumina, similar to the steps sketched above,
and reaction between O and CO could be a reaction step at
high pressures and temperatures.
5.7. Preexponential factors
Fig. 9 shows results for water–gas shift on Rh along
with fits to these data using four different assumptions. The
first is the global expression that we have used to fit all
data, Eq. (10). The second is the fit assuming detailed ki-
netics as described above. The temperature dependence pre-
dicted is greater than observed using an activation energy
of CO, ECO, of 133.4 kJ/mol. However, reducing ECO to
125 kJ/mol (while simultaneously decreasing the activation
energy of the CO–M + O–M reaction to maintain thermo-
dynamic equilibrium) gives a curve that fits the experiment
to within the accuracy of the data.
The adjustable parameters in all fits were the preexponen-
tial factors kof and kob, whose ratio was constrained to agree
with equilibrium as summarized in Table 1. Since ko’s con-
tain an effective area/volume, these preexponentials have no
simple significance since area/volume is not known. How-
ever, the only effect of ceria is to increase kof’s, typically by
factors of 2 to 25 on the different metals.
The most extensive experiments were on Pt and Pt/ceria.
For these systems the variations in kof’s were small for dif-
ferent supports, typically by factors less than 6. Results are

C. Wheeler et al. / Journal of Catalysis 223 (2004) 191–199
199
nearly independent of total surface area of the support which
7. Summary
varied by a factor of ∼ 170. Of course, the Pt surface area
may not have varied by this amount because the metal parti-
cle size may compensate for substrate area variations.
The results in this paper show that noble metal catalysts
with ceria can give very high water–gas-shift activity at very
short contact times. Further, these catalysts are very stable,
withstanding reaction conditions above 800 ◦C with no de-
tectable deactivation over many hours of operation.
These systems are very different than conventionalwater–
gas-shift catalysts in that we use very high metal and ceria
loadings and very low area catalyst supports. Therefore,
these catalysts have large particle sizes initially, and they
are therefore resistant to activity loss due to sintering. Re-
sults are insensitive to total surface area, although this is
not correlated with evidence of either mass-transfer or pore-
diffusion limitations.
The kinetics of these reactions is also remarkably simple,
being consistent with an elementary reaction that is first or-
der in all species and single activation energy for each metal
with or without ceria. However, we have not tested this ex-
pression under wide ranges in partial pressures, so deviations
may of course occur.
6. Mass-transfer and pore-diffusion limitations
Since one is interested in minimizing contact times of
reactants over the catalysts, this requires that catalyst be op-
timized to maximize mass-transfer and limit pore-diffusion
influences.
Mass transfer of reactants to the catalyst surface and mass
transfer of products from the surface are described by noting
that at steady state the mass flux from a flowing channel to
the catalyst surface must be equal to the reaction at an exter-
nal surface if mass transfer is rate controlling [17],
flux of reactant A = kmA(CA − CAs) = rate of reaction of A
= rꢀꢀ = kꢀꢀCAs
,
(20)
Catalysts of this type could be useful in small applications
where robustness is essential and catalyst cost is not a pri-
mary concern. However, while catalyst loadings are higher
than in conventional catalysts, the very short contact times
required compensates for the total amount of metal required,
and we have not attempted to optimize the amount of metal
needed.
where kmA is the mass-transfer coefficient, CA and CAs are
the concentrations in the flowing stream and at the catalyst
surface, respectively, and kꢀꢀ is the surface reaction rate coef-
ficient. In this expression we assume first-order, irreversible
reaction in the reactant A, from which the reaction rate ver-
sus the reactant concentration in the flowing stream is given
by
rꢀꢀ = kꢀꢀCA
(21)
Acknowledgments
if reaction limited, and
This research was partially supported by grants from NSF
and DOE.
rꢀꢀ = kmACA
(22)
if mass-transfer limited.
References
If the catalyst is a porous support layer impregnated with
the catalyst, the reactant concentration will decline through
the catalyst pore because of diffusion resistance, and the con-
sequent effect on rate is described by multiplying the rate at
the external surface of the porous catalyst by an effective-
ness factor η(φ), where φ is the Thiele modulus, defined as
[1] M.V. Twigg, in: Catalyst Handbook, Manson, Frome, England, 1996,
p. 283.
[2] D.A. Hickman, L.D. Schmidt, Science 259 (1993) 343.
[3] P.M. Witt, L.D. Schmidt, J. Catal. 163 (1996) 465.
[4] C.A. Leclerc, J.M. Redenius, L.D. Schmidt, Catal. Lett. 79 (2002) 39.
[5] A.F. Ghenciu, Curr. Opin. Solid State Mater. Sci. 6 (2002) 389.
[6] B.I. Whittington, C.J. Jiang, D.L. Trimm, Catal. Today 26 (1995) 41.
[7] T. Bunluesin, R.J. Gorte, G.W. Graham, Appl. Catal. B 15 (1998) 107.
[8] S. Hilaire, X. Wang, T. Luo, R.J. Gorte, J. Wagner, Appl. Catal. A 215
(2001) 271.
ꢀ
ꢁ1/2
4kꢀꢀ
φ =
l,
(23)
dDA
[9] X. Wang, R.J. Gorte, Appl. Catal. A 247 (2003) 157.
[10] D.C. Grenoble, M.M. Estadt, D.F. Ollis, J. Catal. 67 (1981) 90.
[11] T. Shido, Y. Iwasawa, J. Catal. 141 (1993) 71.
[12] D.A. Hickman, L.D. Schmidt, AIChE J. 39 (1993) 1164.
[13] C.T. Goralski, R.P. O’Connor, L.D. Schmidt, Chem. Eng. Sci. 55
(2000) 1357.
[14] O. Deutschmann, L.D. Schmidt, AIChE J. 44 (1998) 2465.
[15] W.R. Williams, C.M. Marks, L.D. Schmidt, J. Phys. Chem. 96 (1992)
5922.
[16] R. Schwiedernoch, S. Ticher, C. Correa, O. Deutschmann, Chem. Eng.
Sci. 58 (2003) 633.
[17] L.D. Schmidt, in: The Engineering of Chemical Reactions, Oxford
Univ. Press, New York, 1998, p. 283.
where d is the average pore diameter, DA is the diffusion
coefficient of A in the porous catalyst layer, and l is the pore
length.
In the limiting situation where φ ꢁ 1, the rate is predicted
to be proportional to kꢀꢀ1/2 which predicts an activation en-
1
2
ergy that is of the actual reaction activation energy. Since
the results are consistent with constant rate parameters at all
temperatures, we conclude that there is no evidence of ei-
ther external mass transfer (rate nearly independent of T)
1
2
and pore-diffusion limitation (activation energy of actual
value).