Catalytic activation and reforming of methane on supported palladium clusters

Article abstract of DOI:10.1016/j.jcat.2010.06.001

The effects of reactant and product concentrations on turnover rates and isotopic tracing and kinetic isotope effects have led to a sequence of elementary steps for CH₄ reactions with CO₂ and H ₂O on supported Pd catalysts. Rate constants for kinetically-relevant C-H bond activation steps are much larger on Pd than on other metals (Ni, Ru, Rh, Ir, Pt). As a result, these steps become reversible during catalysis, because the products of CH₄ dissociation rapidly deplete the required oxygen co-reactant formed from CO₂ or H₂O and co-reactant activation, and water-gas shift reactions remain irreversible in the time scale required for CH₄ conversion. H₂ and CO products inhibit CH₄ reactions via their respective effects on CH₄ and CO dissociation steps. These mechanistic conclusions are consistent with the kinetic effects of reactants and products on turnover rates, with the similar and normal CH₄/CD₄ kinetic isotope effects measured with H₂O and CO₂ co-reactants, with the absence of H ₂O/D₂O isotope effects, and with the rate of isotopic scrambling between CH₄ and CD₄, ¹²C ¹⁶O and ¹³C¹⁸O, and ¹³CO and ¹²CO during CH₄ reforming catalysis. This catalytic sequence, but not the reversibility of its elementary steps, is identical to that reported on other Group VIII metals. Turnover rates are similar on Pd clusters on various supports (Al₂O₃, ZrO₂, ZrO₂-La₂O₃) and independent of Pd dispersion over the narrow range accessible at reforming conditions, because kinetically-relevant C-H bond activation steps occur predominantly on Pd surfaces. ZrO₂ and ZrO₂-La₂O₃ supports, with detectable reactivity for CO₂ and H₂O activation, can reverse the infrequent formation of carbon overlayers and inhibit deactivation, but do not contribute to steady-state catalytic reforming rates. The high reactivity of Pd surfaces in C-H bond activation reflects their strong binding for C and H and the concomitant stabilization of the transition state for kinetically-relevant C-H activation steps and causes the observed kinetic inhibition by chemisorbed carbon species formed in CH₄ and CO dissociation steps.

Full text of DOI:10.1016/j.jcat.2010.06.001

Journal of Catalysis 274 (2010) 52–63
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Catalytic activation and reforming of methane on supported palladium clusters
Aritomo Yamaguchi, Enrique Iglesia *
Department of Chemical Engineering, University of California at Berkeley, Berkeley, CA 94720, United States
a r t i c l e i n f o
a b s t r a c t
Article history:
The effects of reactant and product concentrations on turnover rates and isotopic tracing and kinetic iso-
Received 29 March 2010
Revised 25 May 2010
Accepted 2 June 2010
Available online 8 July 2010
4 2 2
tope effects have led to a sequence of elementary steps for CH reactions with CO and H O on supported
Pd catalysts. Rate constants for kinetically-relevant C–H bond activation steps are much larger on Pd than
on other metals (Ni, Ru, Rh, Ir, Pt). As a result, these steps become reversible during catalysis, because the
products of CH
and co-reactant activation, and water–gas shift reactions remain irreversible in the time scale required
for CH conversion. H and CO products inhibit CH reactions via their respective effects on CH and
CO dissociation steps. These mechanistic conclusions are consistent with the kinetic effects of reactants
and products on turnover rates, with the similar and normal CH /CD kinetic isotope effects measured
with H O and CO co-reactants, with the absence of H O/D O isotope effects, and with the rate of isotopic
scrambling between CH
This catalytic sequence, but not the reversibility of its elementary steps, is identical to that reported on
other Group VIII metals. Turnover rates are similar on Pd clusters on various supports (Al , ZrO
ZrO ) and independent of Pd dispersion over the narrow range accessible at reforming conditions,
ꢀLa
because kinetically-relevant C–H bond activation steps occur predominantly on Pd surfaces. ZrO and
ZrO supports, with detectable reactivity for CO and H O activation, can reverse the infrequent
4 2 2
dissociation rapidly deplete the required oxygen co-reactant formed from CO or H O
Keywords:
4
2
4
4
Methane reforming
Palladium catalysts
CAH bond activation
4
4
4
Elementary steps for CH reactions
Isotopic tracer
2
2
2
2
1
2
16
13 18
13
12
4
and CD
4
,
C
O and
4
C O, and CO and CO during CH reforming catalysis.
2
O
3
2
,
2
2 3
O
2
2
ꢀLa
2
O
3
2
2
formation of carbon overlayers and inhibit deactivation, but do not contribute to steady-state catalytic
reforming rates. The high reactivity of Pd surfaces in C–H bond activation reflects their strong binding
for C and H and the concomitant stabilization of the transition state for kinetically-relevant C–H activa-
tion steps and causes the observed kinetic inhibition by chemisorbed carbon species formed in CH
CO dissociation steps.
4
and
Ó 2010 Elsevier Inc. All rights reserved.
1
. Introduction
ters [5,6]. Reaction rates, after being corrected for approach to
4
equilibrium, depend linearly on CH pressure but are insensitive
CH
reactants provide attractive routes to synthesis gas (H
streams [1,2] and to carbon nanostructures [3] from natural gas.
Group VIII metals catalyze CH reforming and decomposition
1,2], but strong C–H bonds (439 kJ mol [4]) lead to endothermic
processes that require high temperatures for practical CH conver-
4
activation and reforming reactions using CO
2
or H
2
O co-
to the identity or concentration of the co-reactants on all these cat-
alysts. Measured turnover rates, isotopic tracing, and kinetic iso-
tope effects, obtained under conditions of strict kinetic control,
showed that C–H bond activation is the sole kinetically-relevant
step and that neither reactants nor products lead to significant sur-
face coverages of reactive intermediates during steady-state catal-
2
/CO) or H
2
4
ꢀ1
[
4
sions and, as a result, also catalytic materials that resist sintering
and carbon formation at severe operating conditions. The detailed
sequence of elementary steps and their kinetic relevance, as well
the effects of metal cluster size and supports, remain controversial,
at least in part because of ubiquitous thermodynamic and trans-
port corruptions of reaction rates, often measured at conditions
near thermodynamic equilibrium.
2 2
ysis. On all metals, turnover rates for H O and CO reforming
increased as clusters became smaller, because coordinatively
unsaturated surface atoms, prevalent on small clusters, tend to sta-
bilize C and H atoms and the transition states involved in C–H bond
activation more effectively than more highly coordinated surface
atoms on low-index surfaces. The extreme unsaturation of surface
atoms at corners or edges may lead, however, to unreactive carbon,
rendering such sites permanently unavailable at all reforming
reaction conditions. Supports did not influence C–H activation
turnover rates, except through their effect on metal dispersion,
as expected from the sole kinetic relevance of C–H bond activation
steps and from their exclusive occurrence on metal clusters. Co-
reactant activation steps and all elementary steps involved in
Our previous studies have addressed these mechanistic and
4 2 2
practical matters for CH reforming with CO and H O co-reactants
and for CH decomposition on supported Ni, Ru, Rh, Pt, and Ir clus-
4
*
Corresponding author. Fax: +1 510 642 4778.
E-mail address: iglesia@berkeley.edu (E. Iglesia).
0
021-9517/$ - see front matter Ó 2010 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2010.06.001

A. Yamaguchi, E. Iglesia / Journal of Catalysis 274 (2010) 52–63
53
3
ꢀ1 ꢀ1
water–gas shift reactions remained quasi-equilibrated on these
metals at all reaction conditions examined, consistent with isoto-
pic tracer data that showed carbon, oxygen, and hydrogen isotopic
flowing dry air (Praxair, UHP, 1.2 cm g
s
) at 923 K (0.167
ꢀ
1
3
ꢀ1 ꢀ1
K s ) for 3 h and then in H
temperature for 2 h.
2
(Praxair, UHP, 50 cm g
s ) at each
equilibration for CO–CO
catalysis.
We provide evidence here for a similar sequence of elementary
steps on Pd clusters. Pd atoms at clusters surfaces are much more
reactive for C–H bond activation than on surfaces of the other
2
and H
2
O–H
2
during CH
4
reforming
The dispersion of Pd clusters, defined as the ratio of all Pd atoms
exposed at surfaces, was measured by O adsorption and titration
of adsorbed oxygen by H at 373 K in a volumetric unit (Quanta-
chrome, Autosorb-1). Samples were treated in H (Praxair, UHP)
at 653 K (0.167 K s ) for 1 h and evacuated at 653 K for 1 h to re-
move chemisorbed hydrogen. Saturation monolayer uptakes were
estimated by extrapolating isotherms to zero pressure. The num-
ber of exposed Pd metal atom was estimated by assuming one
chemisorbed oxygen atom per exposed Pd atom [13].
2
2
2
ꢀ1
*
Group VIII metals. These fast reactions scavenge active O species
from co-reactants and prevent the equilibration of co-reactant acti-
vation steps and of water–gas shift on Pd metal surfaces, in contrast
with the equilibrated nature of these steps on less reactive metals.
In addition, the kinetic coupling of fast C–H activation steps on Pd
*
2.2. Methane reaction rate and isotopic measurements
with the steps that form O causes inhibition effects by products,
as a result of the reversibility of both C–H and C–O activation steps.
Sintering and carbon formation have impaired previous at-
tempts at kinetic measurements and at their rigorous mechanistic
Rates were measured in a packed-bed reactor with plug-flow
hydrodynamics. Pd catalysts (5 mg) were diluted with 25 mg of
-Al within catalyst pellets (250–425 m pellet diameter) and
then physically mixed with acid-washed quartz (500 mg, 250–
25 m). These samples were placed in a quartz tube (8 mm diam-
c
2
O
3
l
interpretations for CH
and additives (e.g., La, Ce oxides) have been used to inhibit carbon
formation [8,11], apparently because they can activate CO at con-
ditions (low CO /CH ratios) that prevent the effective removal of
chemisorbed carbon via reactions with CO
4
reforming on Pd catalysts [7–10]. Supports
4
l
2
eter) with a K-type thermocouple enclosed within a quartz sheath
in contact with the catalyst bed. Reactants consisted of mixtures of
2
4
*
2
-derived O on mono-
5
0% CH
tified standard), and He (Praxair, UHP) metered by electronic flow
controllers. H O (deionized water) was introduced using a syringe
4 2 2
/Ar (Praxair, Certified standard), 50% CO /N (Praxair, Cer-
functional catalysts, except through the assistance of co-reactant
activation on supports. These support effects have been claimed
as evidence for the kinetic relevance of CO
other studies conclude that only C–H bond activation steps control
the rates of CH reforming on Group VIII metals [2], consistent with
2
dissociation [7]. Yet
2
pump (Cole-Parmer, 74900 series) into the reactant stream by
vaporizing into a stainless line kept at 423 K. All transfer lines from
the injection point to the gas chromatograph were kept above
10 K to avoid condensation. The concentrations of reactants and
products were measured by on-line gas chromatography (Agilent,
4
the normal kinetic isotopic effects (rCH4/rCD4) observed on most
metal surfaces and with their similar values and rate equations
4
with H
2 2
O and CO co-reactants [5,6,12].
6
890 series) using a HayeSep A (3.2 mm ꢂ 10 m) column and a
Here, we probe the elementary steps involved in CH
4
reactions
thermal conductivity detector.
on Pd-based catalysts and provide evidence for a sequence of ele-
mentary steps consistent with rate and isotopic data obtained un-
der conditions of strict kinetic control. These data show that C–H
bond activation steps on Pd are kinetically-relevant but reversible,
even at conditions for which the overall reforming reaction is
essentially irreversible, in contrast with their irreversible nature
on other Group VIII metals [5,6]. Pd clusters provide the most ac-
tive surfaces for C–H bond activation among Group VIII metals,
but the kinetic coupling of these steps with co-reactant activation
4
Kinetic isotopic effects were measured for CH reforming on a
1
.6% wt. Pd/ZrO
CH and CD
ꢀCO
CD O, and CD
ꢀH
9 atom % deuterium) and D
2
catalyst using the same reactor system using
ꢀCO reactants for CO reforming and CH ꢀH O,
ꢀD O reactants for H O reforming. CD (Isotec,
O (Isotec, 99.9 atom % deuterium)
4
2
4
2
2
4
2
4
2
4
2
2
4
9
2
were used as reactants without further purification. Isotopic ex-
change data were also measured on this catalyst using on-line
mass spectrometry (Inficon, Transpector series). The reactants con-
sisted of 50% CH
Praxair, Certified standard), Ar (Praxair, UHP), and He (Praxair,
UHP) all metered using electronic flow controllers. CD (Isotec,
(Matheson, 99.7%), CO (Isotec, 99 atom
4 2
/Ar (Praxair, Certified standard), 50% CO /He
2
steps leads to inhibition by the H and CO products of reforming
(
reactions. Supports do contribute to steady-state turnover rates,
but their ability to activate co-reactants can be used to reverse
the occasional blockage of Pd cluster surfaces during steady-state
4
1
3
9
9 atom % deuterium), D
2
1
3
13 18
13
%
C), and
C
O (Isotec–Sigma–Aldrich, 99 atom % C, 95 atom %
O) were used as reactants without further purification. Concen-
trations of CH , CO, CO isotopomers were estimated from mass
spectra using matrix deconvolution methods [14].
CH
. Experimental methods
.1. Catalysts synthesis and characterization
4
reforming.
1
8
2
4
2
2
3
. Results and discussion
ZrO
ZrO was prepared by hydrolysis of an aqueous solution of
ZrOCl O (Aldrich, >98% wt.) at a constant pH of 10 and subse-
ꢁ8H
quent filtration, drying, and treatment in flowing dry air (Praxair,
2 2 3
and c-Al O were prepared as described previously [6].
2
3.1. Catalyst characterization
2
2
All catalysts used in this study are listed in Table 1. These sam-
3
ꢀ1 ꢀ1
ꢀ1
UHP, 1.2 cm g
s
) at 923 K (0.167 K s ) for 5 h.
c
-Al
2
O
3
was
2
ples were treated in H at temperatures higher than those used in
prepared by treating Al(OH)
3
(Aldrich) in flowing dry air (Praxair,
CH reforming reactions to prevent further structural changes dur-
4
3
ꢀ1 ꢀ1
ꢀ1
UHP, 1.2 cm g
Pd/Al
tion of ZrO
Aesar, 99.9%) and treatment in ambient air at 393 K. These samples
s
) at 923 K (0.167 K s ) for 5 h. Pd/ZrO
2
and
ing catalysis. The mean cluster size in each sample was estimated
from Pd dispersions measured from chemisorption uptakes by
assuming hemispherical crystallites and the atomic density of bulk
2
O
3
samples were prepared by incipient wetness impregna-
or -Al with an aqueous solution of Pd(NO (Alfa
2
c
2
O
3
3 2
)
Pd. The Pd cluster diameter for the 1.6% wt. Pd/ZrO
2
sample treated
(1023))
3
ꢀ1 ꢀ1
were treated in flowing dry air (Praxair, UHP, 1.2 cm g
s
) at
in H at 1023 K (denoted hereinafter as 1.6% wt. Pd/ZrO
2
2
ꢀ1
9
23 K (heating rate 0.167 K s ) for 5 h and then in flowing H
2
ꢀ1
was 12.5 nm. The Pd cluster size increased with increasing H
treatment temperature and reached a value of 32 nm after H
2
3
ꢀ1 ꢀ1
(
Praxair, UHP, 50 cm g
for 2 h. Pd/ZrO was prepared by incipient wetness impreg-
ꢀLa
nation of Pd/ZrO (treated in H as described above) with an aque-
ous solution of La(NO O (Aldrich, 99.999%), treatment in
ꢁ6H
s ) at 1023 K (heating rate 0.167 K s )
2
2
2
O
3
2
treatment at 1123 K (1.6% wt. Pd/ZrO (1123)), a value consistent
with the Pd cluster size derived from transmission electron micros-
copy (31 nm mean cluster diameter, Fig. S1).
2
2
3
)
3
2

5
4
A. Yamaguchi, E. Iglesia / Journal of Catalysis 274 (2010) 52–63
Table 1
15
Reduction temperature, Pd dispersion, and mean particle size of supported Pd
catalysts.
Catalyst
Reduction
temperature (K)
Dispersion
(%)
Mean particle
size (nm)b
a
1
1
1
1
.6% wt. Pd/ZrO
.6% wt. Pd/ZrO
2
2
1023
1023
1023
1123
8.9
5.3
7.5
3.5
12.5
21.1
14.9
32.0
1
0
5
0
ꢀLa
2 3
O
.6% wt. Pd/Al
2
O
3
.6% wt. Pd/ZrO
2
a
The fraction of the surface Pd atoms from O
chemisorbed oxygen by H
2
chemisorption and titration of
2
.
b
The mean particle size (D) was estimated from Pd dispersion (d) using D = 1.1/d.
CH
ZrO
ꢀLa
over rates were stable with time on Pd/ZrO
Pd/Al deactivated during reaction, indicating that ZrO
ZrO supports stabilize Pd clusters against deactivation.
ꢀLa
Pd clusters on Al appeared to grow or become covered with
unreactive species during CH reforming [15]. Initial CH turnover
4
ꢀCO
2
turnover rates were measured on Pd/ZrO
(1023), and Pd/Al (1023) (1.6% wt.; Fig. 1). CH
and Pd/ZrO
ꢀLa
2
(1023), Pd/
2
2
O
3
2
O
3
4
turn-
O ;
2 3
2
and
2
2
2
O
3
2
O
2 3
0.0
0.5
1.0
1.5
2
O
3
6
3
(
10 cm /g-h)/Gas hourly space velocity
4
4
rates were similar on Pd clusters dispersed on all three supports.
The low and similar dispersion (5.3–8.9%, Table 1) of these samples
causes reactions to occur predominantly at low-index planes,
which prevail on large clusters. The fraction of surface atoms ex-
posed at edges and corners decreases monotonically with increas-
ing cluster size [16], but their high reactivity tends to be dampened
Fig. 2. Measured CH
reaction on 1.6% wt. Pd/ZrO
of catalyst diluted with 25 mg of Al
00 mg of ground quartz (250ꢀ425
Al m), then diluted with 500 mg of ground quartz (250ꢀ425
(250ꢀ425
h) 5 mg of catalyst diluted with 25 mg of Al
(63ꢀ106
500 mg of ground quartz (250ꢀ425 m)).
4
reaction rates as a function of residence time for CH
(1023) at 823 K (PCH4 = 10 kPa, PCO2 = 40 kPa, (d) 5 mg
(pellet size 250ꢀ425 m), then diluted with
4
ꢀCO
2
2
2
O
3
l
5
lm), (N) 5 mg of catalyst diluted with 50 mg of
2
O
3
l
lm),
(
2
O
3
lm), then diluted with
l
by their tendency to form strongly-bound CH
17,18] that are removed too slowly for significant contributions
to reforming turnovers.
4
-derived species
[
Pd) catalysts with different pellet size (63–106
and extent of dilution within pellets ( -Al
dilution ratios = 5:1, 10:1) and within the catalyst bed (acid-
washed quartz (250–425 m); quartz:catalyst dilution ra-
tio = 100:1). Neither pellet size nor intrapellet dilution influenced
measured CH turnover rates, indicating that rate data are unaf-
l
m, 250–425
l
m)
c
2 3
O :catalyst intrapellet
We conclude from these data that these supports do not influ-
ence turnover rates on Pd clusters because of their low reactivity
in activation of H O and CO co-reactants. Stochastic processes
2 2
lead to the occasional deactivation of a given Pd cluster by the for-
mation of a carbon overlayer. In such cases, supports with low but
l
4
fected by transport artifacts that cause ubiquitous temperature
and concentration gradients; therefore, all measured rates strictly
reflect the rates of chemical reactions at cluster surfaces.
detectable rates of co-reactant activation, such as ZrO
ZrO , allow the ultimate removal of these overlayers and inhibit
deactivation.
2 2 3
or La O –
2
4 f
Forward CH turnover rates (r ) were obtained from measured
turnover rates (rnet) by correcting for the approach to equilibrium
3
4
.2. Effects of residence time and dilution on CH reaction rates
(
g
i
) for CH
4
ꢀCO
2
(Eq. (2)) and CH
ꢀH
4 2
O (Eq. (3)) reactions [19]:
ð1Þ
Fig. 2 shows CH
4
reforming turnover rates (per surface Pd atom)
(1023) (1.6% wt.
rnet;i
r
f;i
¼
at 823 K as a function of residence time on Pd/ZrO
2
1 ꢀ
g
i
2
2
½
P
COꢃ ½PH2
ꢃ
1
ꢂ K
g1
g2
¼
¼
ð2Þ
ð3Þ
½
P
CH4 ꢃ½PCO2
ꢃ
1
3
½
P
COꢃ½PH2
ꢃ
1
ꢂ K
2
½
P
CH4 ꢃ½PH2
O
ꢃ
[
P
j
] is the average pressure (in atm) of species j in the reactor, and K
and K are equilibrium constants for CO and H O reactions with
CH , respectively [20]. The average pressures of reactants and prod-
1
2
2
2
4
ucts were used in all rate and equilibrium equations to correct for
the small changes that occurred as conversion changed along the
catalyst bed. At all conditions used in our experiments,
were much smaller than unity for reactions with either CO
O, and CH conversion levels were maintained below 9% in all
experiments.
On other Group VIII metals (Ni, Ru, Rh, Pt, Ir) [5,6], forward CH
turnover rates (from Eq. (1)) did not depend on residence time or
and CO product concentrations. In contrast, both measured
Fig. 2) and forward (Fig. S2) CH reforming rates decreased with
increasing residence time on Pd-based catalysts, even though
CH and co-reactant concentrations were essentially unaffected
g
values
2
or
H
2
4
4
H
(
2
4
4
Fig. 1. Measured CH
stream on (N) 1.6% wt. Pd/ZrO
.6% wt. Pd/Al (1023) at 823 K (PCH4 = 10 kPa, PCO2 = 40 kPa, residence time 0.83
10 cm /g h)/Gas hourly space velocity).
4
turnover rates for CH
4
ꢀCO
2
reaction as a function of time on
by residence time at the low conversions prevalent in these exper-
iments. We conclude that these effects must reflect the inhibition
of CH reforming reactions by H and/or CO reaction products.
4 2
2
(1023), (s) 1.6% wt. Pd/ZrO
2
ꢀLa (1023), and (j)
2 3
O
1
(
2 3
O
6
3

A. Yamaguchi, E. Iglesia / Journal of Catalysis 274 (2010) 52–63
55
The next sections provide evidence for these conclusions by prob-
ing the effects of added H and CO and of reactant concentrations
on reforming turnover rates.
15
2
3.3. Effect of reactant and product concentrations on CH
4
turnover
rates
1
0
5
0
CH
ZrO (1023) at 823 K are shown as a function of CH
Fig. 3. Forward CH turnover rates increased with increasing CH
pressure, but non-linearly, for both reactions; turnover rates were
also slightly higher with H O than with CO co-reactants. On other
metals (Ni, Ru, Rh, Pt, Ir) [5,6], forward CH turnover rates were
strictly proportional to CH pressure but unaffected by the identity
or concentration of the co-reactant. As we discuss below, these
small differences in CH and CH ꢀH O turnover rates on
ꢀCO
Pd-based catalysts reflect differences in the prevalent concentra-
tions of CO and H with these two co-reactants and their respective
inhibitory effects on C–H bond activation rates.
Fig. 4 shows the effects of CO and H O pressures on CH
over rates on 1.6% wt. Pd/ZrO (1023) at 823 K. CH turnover rates
O pressures. The small differences be-
O turnover rates again reflect differ-
pressures and their inhibitory
4
ꢀCO
2
and CH
4
ꢀH
2
O
reaction rates on 1.6% wt. Pd/
2
4
pressure in
4
4
2
2
4
4
4
2
4
2
0
10
20
30
40
2
Average CO or H O pressure (kPa)
2
2
2
2
4
turn-
Fig. 4. Forward CH
4
turnover rates as a function of average (d) CO
ꢀCO or CH ꢀH O reaction on 1.6% wt. Pd/ZrO
PCH4 = 10 kPa, residence time 0.83 (10 cm /g h)/Gas hourly space velocity).
2
2
or (N) H O
2
(1023) at 823 K
2
4
pressure for CH
4
2
4
2
did not depend on CO
tween CH and CH
2
or H
2
6
3
(
4
ꢀCO
2
4
ꢀH
2
ences in the prevalent CO and H
2
effects on C–H bond activation, as shown below (Sections 3.3 and
.4). These data indicate that co-reactant activation steps are faster
3.4. Kinetic isotope effects and isotopic tracing experiments
3
than C–H bond activation and thus kinetically-irrelevant and that
adsorbed species derived from these co-reactants are not present
at any significant surface concentrations during steady-state
catalysis.
Table 2 shows H/D kinetic isotopic effects measured by compar-
ing turnover rates with CH
and CD O reactant mixtures (10 kPa CH
ꢀD
O or D O) on 1.6% wt. Pd/ZrO at 823 K. The measured normal
CH /CD kinetic isotopic effects were identical for CO or H O co-
4
ꢀCO
2
, CD
4
ꢀCO
2
, CH
4
ꢀH
2
O, CD
, 40 kPa CO ,
4
ꢀH
2
O,
4
2
4
or CD
4
2
H
2
2
2
Forward CH
sure for CH
ꢀCO
from 10.8 to 2.7 s as H
.8 kPa, consistent with the observed effects of residence time on
reforming rates (Fig. 2). The effects of CO pressure on CH reform-
ing turnover rates were measured using CH ꢀCO reactants
ꢀCO
Fig. 6). Turnover rates decreased from 10.3 to 3.9 s as CO pres-
sures increased from 0.98 kPa to 5.2 kPa. CH turnover rates
4
turnover rates are shown as a function of H
2
pres-
4
4
2
2
4
2
ꢀH reactants in Fig. 5. Turnover rates decreased
2
reactants (1.39–1.41). These data indicate that C–H bond activation
is the only kinetically-relevant step on Pd clusters with both co-
ꢀ1
2
pressures increased from 0.33 kPa to
7
reactants. These similar kinetic isotope effects for CH
4 4
/CD with
4
CO and H O co-reactants also show that the extent to which this
2
2
4
2
step controls overall reforming rates is independent of the identity
of the co-reactant. No kinetic isotopic effects were observed for
ꢀ1
(
4
ꢀCO
2
CD
ꢀD
O and CD
ꢀH
2
O reactants (0.97), consistent with the lack
4
2
4
before and after CO co-feed were identical, indicating that these ef-
fects reflect reversible inhibition instead of irreversible deactiva-
tion of Pd surface sites by reaction products.
of kinetic relevance of water activation steps or of any steps involv-
ing water-derived intermediates. These kinetic isotopic effects on
Pd catalysts are slightly smaller than typical values for other
1
1
5
0
5
0
15
10
5
0
0
5
10
15
0
2
4
6
8
10
Average CH pressure (kPa)
Average H pressure (kPa)
4
2
Fig. 3. Forward CH
CH and (N) CH
ꢀCO
H2O = 40 kPa, residence time 0.83 (10 cm /g h)/Gas hourly space velocity).
4
turnover rates as a function of average CH
4
pressure for (d)
Fig. 5. Forward CH
CH
ꢀCO
residence time 0.83 (10 cm /g h)/Gas hourly space velocity).
4
turnover rates as a function of average H
2
pressure for
4
2
4
ꢀH O reaction on 1.6% wt. Pd/ZrO (1023) at 823 K (PCO2 or
2
2
4
2 2
reaction on 1.6% wt. Pd/ZrO (1023) at 823 K (PCH4 = 10 kPa, PCO2 = 40 kPa,
6
3
6
3
P

5
6
A. Yamaguchi, E. Iglesia / Journal of Catalysis 274 (2010) 52–63
1
1
5
0
5
0
other Group VIII metals [5,6]. These differences among metals re-
flect, as we discuss in greater detail below, the larger C–H bond
activation rate constants on Pd clusters compared with those on
the other metals. CH
2
D
2
and CHD
3
isotopomers formed at much
lower rates (ꢅ20 times) than CH
3
D, suggesting that subsequent
C–H bond activation of CH species is essentially irreversible. The
ꢄ
3
ratio of H/D scrambling to CH
r(CH D)/r(CO)} increased with increasing D
suggesting that CH D formation rates (r(CH
4
chemical conversion rate
{
3
2
pressure (Table 3),
3
D)), via the micro-
3
scopic reverse of the initial C–H bond activation step, increase with
increasing concentrations of chemisorbed deuterium. This conclu-
3
sion was confirmed by the linear dependence of the {r(CH D)/
1
/2
r(CO)} ratio on PD2 (Fig. S3); the latter dependence gives, in turn,
*
the surface coverage of D for equilibrated adsorption–desorption
steps.
The various dihydrogen isotopomers were present in binomial
distributions at all conditions (e.g., H
0.044:0.33:0.63 for binomial; 10 kPa D
dissociation–recombination steps are indeed at quasi-equilibrium
during CH reforming, even at the highest D pressures and for
CH conversions far from thermodynamic equilibrium. The H cov-
erages expected during reforming reactions were estimated from
2 2
:HD:D , 0.046:0.30:0.65 vs.
0
2
4
6
8
), indicating that hydrogen
2
Average CO pressure (kPa)
4
2
*
Fig. 6. Forward CH
CH
4
turnover rates as a function of average CO pressure for
(1023) at 823 K (PCH4 = 10 kPa, PCO2 = 40 kPa,
4
4
ꢀCO
2
reaction on 1.6% wt. Pd/ZrO
2
6
3
residence time 0.83 (10 cm /g h)/Gas hourly space velocity).
ꢀ
1
2
the heat of H adsorption on Pd surfaces, reported to be 87 kJ mol
ꢀ1
on Pd(1 1 1) [22] and 102 kJ mol on Pd(1 1 0) [22]. These values
would lead to H surface coverages much smaller than 0.01 on Pd
*
Table 2
Kinetic isotope effects for CH
CH or CD , 40 kPa CO , H O, or D
CD , 25 kPa CO , H O, or D O).
4
reactions on 1.6% wt. Pd/ZrO
2
(1023) at 823 K (10 kPa
at 873 K (25 kPa CH or
2
surfaces at typical reaction conditions (823 K; 0.3–2 kPa H ) and
4
4
2
2
2
O) and on 1.6% wt. Pt/ZrO
2
4
to desorption activation barriers (typically equal to adsorption
heats for non-activated adsorption events) that would allow rapid
adsorption–desorption equilibrium during reforming catalysis.
We conclude that initial C–H bond activation steps are revers-
ible even when overall reforming reactions are essentially irrevers-
ible. The reversible nature of C–H bond activation elementary steps
4
2
2
2
Catalysts
r(CD
Pd/ZrO
2
[This work]
2
Pt/ZrO [6]
4
ꢀH
ꢀCO
ꢀH O)/r(CD
2
O)/r(CD
4
ꢀD
ꢀCO
ꢀH
2
O)
0.97
1.41
1.39
1.07
1.77
1.69
r(CH
r(CH
4
2
)/r(CD
4
2
)
4
2
4
2
O)
is responsible for the observed H
2
inhibition effects, which reflect
the more frequent reversal of C–H activation events with increas-
*
metals (Pt (1.69–1.77 at 873 K) [6], Ir (1.75–1.81 at 873 K) [5], Ru
ing H
2
pressures and H concentrations.
(
(
1.40–1.42 at 873 K) [6], Rh (1.54–1.56 at 873 K) [5], and Ni
1.62–1.66 at 873 K) [5]). These differences appear to reflect earlier
3.5. Effect of CO on CH
4
reforming turnover rates
transition states with shorter C–H bonds on the more reactive sur-
faces of Pd clusters, as suggested by theoretical treatments of the
transition states involved in C–H bond activation steps on various
metal surfaces [21].
The expected CO coverages during reforming reactions were
estimated from the heat of CO adsorption on Pd surfaces to con-
sider whether inhibition could simply reflect competitive adsorp-
tion of CO or require its dissociative adsorption to form more
CH
4
ꢀCO
2
ꢀD
2
reactant mixtures were used to determine
*
*
whether inhibition reflects competitive co-adsorption leading to
strongly bound C and O species. The heat of CO molecular adsorp-
significant H* coverages or the involvement of H-adatoms in
ꢀ1
tion has been reported to be 140 kJ mol
on Pd(1 1 1) [23],
ꢀ1
ꢀ1
reversing C–H bond activation steps. CH
rates were compared with the rate of formation of CH4ꢀx
omers, measured by mass spectrometry (Table 3). CH D formation
rates (r(CH D)) were similar to those for chemical conversion of
CH to H and CO (r(CO)), indicating that CH and D (or H ) recom-
bine often to form CH
version turnovers. Thus, we conclude that CH
4
chemical conversion
170 kJ mol on Pd(1 1 0) [23], and 148 kJ mol on 1.5 to 9.5 nm
Pd clusters [24]. These values would lead to CO surface coverages
D
x
isotop-
smaller than 0.1 on Pd surfaces for CH
4
ꢀCO
2
or CH
4
ꢀH
2
O reactions
3
at 823 K and typical CO pressures in our study (0.5–1 kPa). We con-
clude that molecularly adsorbed CO is unlikely to block Pd surface
atoms to the extent required to cause the observed inhibition ef-
3
ꢄ
*
*
4
2
3
3
D (or CH
4
) in the time scale of chemical con-
activation steps
leading to CH and H are reversible; in contrast, these steps were
4
fects. Yet CH turnover rates decreased markedly with increasing
4
ꢄ
*
CO pressure (Fig. 6), apparently because of concurrent CO dissoci-
3
*
*
found to be irreversible and H
2
did not inhibit CH
4
reforming on
ation via the microscopic reverse of the recombination of C and O ,
which forms the CO products of CH reforming.
The reversible nature of CO dissociation steps was confirmed
4
Table 3
1
2
16
13 18
from the rate of isotopic scrambling of
C
O(1 kPa)ꢀ
(40 kPa) reactions at 823 K. Table 4
2
shows CO and CO isotopomer concentrations in the effluent
C O(1 kPa)
Ratios of CH
CH chemical conversion rates (r(CO)) during
reactions of CH
Pd/ZrO (1023)
3 3
D formation rates (r(CH D)) to
1
2
12 16
during CH
4
(10 kPa)ꢀ
C O
2
4
4
ꢀCO
2
ꢀD
2
mixture on 1.6% wt.
(PCH4 = 10 kPa,
2
at
823 K
13 16
O and 13C
16
stream.
C
2
O isotopomers formed (via CO dissocia-
tion to C and O and subsequent recombination) at rates compara-
6
3
P
CO2 = 40 kPa, residence time 0.83 (10 cm /
*
*
g h)/Gas hourly space velocity).
1
3
ble to those for chemical conversion of CH
CH
not equilibrated (
4
to CO ( C scrambling to
Inlet D
2
3
pressure (kPa) r(CH D)/r(CO)
4
chemical conversion rate ratio ꢅ0.17). CO isotopomers were
12 16 12 18 13 16 13 18
0
2
5
0
C O: C O: C O: C O, 0.78:0.013:0.070:0.13
1.4
2.6
3.8
vs. 0.70:0.12:0.17:0.030 for binomial distributions), indicating that
CO dissociation steps are reversible but not quasi-equilibrated
during CH reforming. We conclude that CO inhibition of CH
4 4
1
0

A. Yamaguchi, E. Iglesia / Journal of Catalysis 274 (2010) 52–63
57
Table 4
4 2 2
3.7. Elementary steps involved in CH reactions with H O and CO
Isotopic distribution during 12CH
(10 kPa)ꢀ12C16
O
(40 kPa)ꢀ13C18O(1 kPa)ꢀ C16
12
O
4
2
6
3
(1 kPa) reaction on 1.6% wt. Pd/ZrO
2
(1023) at 823 K (residence time 0.83 (10 cm /
We propose next a sequence of elementary steps (Scheme 1)
consistent with the rate and isotopic data described in detail
g h)/Gas hourly space velocity).
P (kPa) in reactants
P (kPa) in products
above. The reversibility of CH
was confirmed from the rate of formation of CH4ꢀx
during reactions of CH mixtures. The activation steps of
4
activation steps (steps (5)–(8))
1
1
1
1
2
2
3
3
16
18
16
C
C
C
C
O
O
O
1.0
0
0
3.13
0.051
0.28
0.53
D
x
isotopomers
4
ꢀCO
2
ꢀD
2
18O
co-reactants (step (9)–(11)) and the CO dissociation (step (12))
showed some reversibility but were not quasi-equilibrated during
1.0
1
1
1
1
1
1
2
2
2
3
3
3
16
C
C
C
C
C
C
O
2
18O
40.0
0
0
38.16
0.46
0.0026
0.31
0.013
0
16
18
16
CH
ation–recombination steps (step (13)) were quasi-equilibrated
during CH reforming, as shown by the binomial dihydrogen iso-
topomer distribution obtained from CH reactant mix-
4
ꢀCO
2
4
or CH ꢀH
2
O reactions on Pd clusters. Hydrogen dissoci-
O
O
O
2
2
0
4
16 18
O
O
0
4
ꢀCO
2
ꢀD
2
18O
0
2
tures. Literature values [23,24] for CO adsorption enthalpies and
CO desorption activation barriers (step (14)) indicate that molecu-
lar CO adsorption steps are quasi-equilibrated, but do not lead to
reforming reactions is caused by the formation of kinetically-
detectable concentrations of active C species via the microscopic
*
*
significant CO coverages, on Pd catalysts at the conditions of our
*
*
CH
4
reforming experiments.
A rate equation for CH conversion turnover rates (r
rived from the sequence of elementary steps in Scheme 1 by apply-
reverse of the C –O recombination reactions that form CO during
CH reforming. These species decrease the number of unoccupied
4
f
) was de-
4
Pd sites (ꢄ) available for kinetically-relevant C–H activation steps
and, as a result, the rate of the overall catalytic sequence. We show
below that these conclusions are consistent with the form of the
rate equation required to describe the effects of CO pressure on
reaction rates.
ꢄ
*
*
ing the pseudo-steady-state hypothesis (PSSH) to [CH ], [C ], [O ],
and [OH ] adsorbed species (see Appendix I for details). PSSH for
3
*
ꢄ
[
CH ] gives:
3
k
1
½CH ꢃ
4
ꢄ
½
CH ꢃ ¼
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi½ꢄꢃ
ð15Þ
3
½
2
H ꢃ
k
ꢀ1
þ k
2
K
7
3.6. Rates and mechanism for activation of CO
2
and H
2
O co-reactants
(10 kPa)ꢀD
which taken together with the expression for the net rate of step (5)
and the quasi-equilibrium equation for hydrogen dissociation (step
H–D scrambling rates during reactions of CH
40 kPa) reactant mixtures were measured at 823 K from the efflu-
ent concentrations of CH4ꢀx and H2ꢀx O isotopomers during
catalysis. CH D/CH and HDO/D O ratios in the effluent were
.014 and 0.029, respectively, while other isotopomers were not
detected. Measured HDO/D O ratios (0.029) were significantly
smaller than predicted from equilibrium between D O and the H-
atoms (0.07) formed via CH conversion and D O activation. These
data indicate that water activation steps are not equilibrated dur-
ing CH O reactions on Pd cluster surfaces. On other Group VIII
ꢀH
metals (Ni, Ru, Rh, Pt, Ir) [5,6], H O activation steps were equili-
brated and binomial isotopomer mixtures were formed from
CH OꢀD reactant mixtures.
ꢀH
The reversibility of CO
4
2
O
(
4
13)) gives an equation for the forward rate of CH reforming:
(
ꢃ½ꢄꢃ ꢀ kꢀ1½CHꢄ
2
ꢃ½H ꢃ ¼ k
2
ꢄ
½CHꢄ
ꢃ½ꢄꢃ
D
x
D
x
r
f
¼ k
1
½CH
1
4
3
2
3
3
4
2
k
½CH ꢃ
4
2
¼
qffiffiffiffiffiffiffiffiffiffiffiffiffiffi½ꢄꢃ ¼
a½ꢄꢃ
ð16Þ
ð17Þ
0
k
k
ꢀ1
½H ꢃ
2
þ 1
K
7
2
2
2
k
1
½CH ꢃ
4
4
2
a
¼
qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
k
k
ꢀ1
½H ꢃ
2
þ 1
2
K
7
4
2
in terms of the concentration of unoccupied sites (ꢄ) present during
steady-state catalysis. The concentration of unoccupied sites is gi-
ven by the site balance:
2
4
2
2
2
activation steps was determined from
ꢄ
ꢄ
ꢄ
½
ꢄꢃ þ ½C ꢃ þ ½O ꢃ þ ½CO ꢃ ¼ L
ð18Þ
1
3
12
12
C exchange rates between CO
2
and CO in CH
4
(10 kPa)ꢀ CO
2
1
3
(
40 kPa)ꢀ CO(5 kPa) reactant mixtures. On other metals (Ni, Ru,
by evaluating the steady-state coverages of all other adsorbed inter-
13
*
Rh, Pt, and Ir) [5,6], these experiments led to identical C fractions
in the CO and CO molecules in the effluent even at very low CH
conversions; such rapid exchange indicates that CO
ꢀCO intercon-
version is much faster than chemical conversion of CH and thus oc-
mediates. The [C ] term in Eq. (18) is given by its PSSH expression:
2
4
2
½COꢃ
2
ꢄ
ꢄ
a
½ꢄꢃ þ k
6
½ꢄꢃ ¼ kꢀ6½C ꢃ½O ꢃ
ð19Þ
2
K
8
4
*
A similar treatment for [OH ] gives:
curs in quasi-equilibrated steps on these other metals. In
agreement with these data, the water–gas shift reaction, which in-
volves H
sffiffiffiffiffiffiffiffi
sffiffiffiffiffiffiffiffi
k
k
4
k
k
ꢀ5 ½H
2
ꢃ
k
k
4
k
k
ꢀ5 ½H
2
ꢃ
½
H
2
Oꢃ½ꢄꢃ þ
½Oꢄꢃ
½H
2
Oꢃ½ꢄꢃ þ
½Oꢄꢃ
2 2
O and CO activation steps, was also equilibrated at all
5
5
K
7
5
5
K
7
ꢄ
reaction conditions on Ni, Ru, Rh, Pt, and Ir catalysts [5,6]. In con-
½OH ꢃ ¼
sffiffiffiffiffiffiffiffi
¼
1
3
b
trast, the C fractions in CO and CO
ent were 0.74 and 0.020, respectively, during CH
reactions on Pd/ZrO . Thus, CO activation steps that form CO
and O are not quasi-equilibrated during CH reactions on Pd
ꢀCO
cluster surfaces, even at modest CH
chemical conversions (ꢅ7%).
Water–gas shift and its reverse reaction also remained non-equili-
brated ( = 0.05–0.5 for CH and CH O reactions) at all
conditions, consistent with non-equilibrated CO
tion elementary steps. The value of was defined as:
2
molecules in the reactor efflu-
k
k5
ꢀ4 ½H
2
ꢃ
þ 1
12
12
13
4
ꢀ
CO
2
ꢀ
CO
K
7
*
2
2
ð20Þ
ð21Þ
*
4
2
where
4
sffiffiffiffiffiffiffiffi
kꢀ4 ½H
2
ꢃ
b ¼
þ 1
g
4
ꢀCO
2
4
ꢀH
2
k
5
K
7
2
and H
2
O activa-
*
g
Finally, the PSSH assumption for [O ] gives the equation:
k
b
4
½
P
COꢃ½PH2
O
ꢃ
ꢃ
1
k ½CO ꢃ þ ½H Oꢃ ꢀ
3
2
2
a
g
¼
ꢂ K
ð4Þ
d
c
½
P
CO2 ꢃ½P 2
H
WGS
ꢄ
½O ꢃ ¼
sffiffiffiffiffiffiffiffi
sffiffiffiffiffiffiffiffi½ꢄꢃ ¼ ½ꢄꢃ
ð22Þ
½
3
COꢃ
½H
2
ꢃ
k
b
ꢀ5 ½H ꢃ
2
where KWGS is the equilibrium constant for the water–gas shift
reaction.
k
ꢀ
þ kꢀ5
ꢀ
K
8
K
7
K
7

5
8
A. Yamaguchi, E. Iglesia / Journal of Catalysis 274 (2010) 52–63
4
Scheme 1. Sequence of elementary steps for CH reforming reactions on Pd-based catalysts.
in which
observed product inhibition effects. This equation describes rates
of CH reactions with both H O and CO as co-reactants, which
sffiffiffiffiffiffiffiffi
sffiffiffiffiffiffiffiffi
4
2
2
½
COꢃ
½H
2
ꢃ
k
b
ꢀ5 ½H
2
ꢃ
c
¼ kꢀ3
þ kꢀ5
ꢀ
ð23Þ
ð24Þ
lead, however, to different contributions from the various denom-
inator terms in Eq. (28). These denominator terms correspond to
the respective coverages of adsorbed species; their relative contri-
butions can only be discerned by a rigorous comparison of all ki-
netic data with the form of Eq. (28).
We consider next a detailed analysis of all rate data in the con-
text of Eq. (28) for three specific assumptions about the identity of
the most abundant surface intermediates (MASI):
K
8
K
7
K
7
k
4
d ¼ k
3
½CO
2
ꢃ þ ½H
2
Oꢃ ꢀ
a
b
*
The [C ] term can then be restated by combining Eqs. (19) and (22)
to give:
½COꢃ
a
þ k
6
Cꢄꢃ ¼
K
8
½ꢄꢃ
ð25Þ
½
d
k
ꢀ6
*
c
(i) CO and ꢄ as MASI:
The expressions for all adsorbed intermediates can be introduced
into the site balance (Eq. (18)) to solve for the concentration of
unoccupied sites:
k
1
½CH
4
ꢃ
r
f
¼
ꢀ
ꢁ
ð29Þ
À pffiffiffiffiffiffiffiffi
Á
2
½COꢃ
A
½H
2
ꢃ þ 1
þ 1
K
8
1
½ꢄꢃ ¼
ð26Þ
½
COꢃ
a
þ k
ꢀ6
6
K
8
d
c
½COꢃ
1
þ
þ
þ
K
8
d
c
k
*
(
ii) O and ꢄ as MASI:
4
The equation for the CH turnover rate can then be obtained by
substituting (ꢄ) into Eq. (16) to give:
k
1
½CH
4
ꢃ
pffiffiffiffiffiffiffiffi
a
2
r
f
¼
a
½ꢄꢃ ¼
ð27Þ
A
½H
2
ꢃ þ 1
!
2
½
COꢃ
r
f
¼
ð30Þ
ð28Þ
a
þ k
ꢀ6
6
2
4
0
@
½H Oꢃ
k
½CH
2
ꢃ
13
2
K
8
d
c
½COꢃ
pffi
2
ffiffiffiffiffi
p
1
ffiffiffiffiffiffi
4
D½CO
2
ꢃ þ E
ꢀ
1
þ
þ
þ
K
8
d
c
½H
2
ꢃ
A
½H
ꢃþ1
k
A5
1 þ
pffiffiffiffiffiffiffiffi
B½COꢃ þ C ½H
2
ꢃ
Substituting the defining equation for
gives a final equation for CH
a
, b,
c
, and d into Eq. (27)
4
reforming rates:
k
1
½CH
4
ꢃ
pffiffiffiffiffiffiffiffi
A
½H
2
ꢃ þ 1
r
f
¼ 2
0
1
3
0
1
2
!
pffiffiffiffiffiffiffiffi
½H
2
Oꢃ
2
k
1
½CH ꢃ
2
4
D½CO
2
ꢃ þ E pffiffiffiffiffiffiꢀ pffiffiffiffiffiffi
ꢂ
ꢃ
6
4
k
1
½CH
4
ꢃ
B
@
B½COꢃ þ C ½H
2
k
ꢃ
C
½H
ꢃ
A
½H
ꢃþ1
½COꢃ 7
@
A
1 þ
pffiffiffiffiffiffiffiffi
Aþ
pffiffiffiffiffiffiffiffi
þ
5
½
H
2
Oꢃ
1
½CH ꢃ
4
A
½H
2
ꢃ þ 1
D½CO ꢃ þ E pffiffiffiffiffiffiꢀ pffiffiffiffiffiffi
B½COꢃ þ C ½H
2
ꢃ
K
8
2
½
H
2
ꢃ
A
½H ꢃþ1
2
*
in which all remaining parameters (k
only of rate and equilibrium constants for elementary steps with-
out any residual concentration dependences, as shown in Table 5.
1
, A, B, C, D, and E) consist
(iii) C and ꢄ as MASI:
k
1
½CH
4
ꢃ
pffiffiffiffiffiffiffiffi
A
½H
2
ꢃ þ 1
r
f
¼ 2
0
13
2
!
pffiffiffiffiffiffiffiffi
In this final equation, k
tion in the elementary step (5) and K
1
is the rate constant for C–H bond activa-
is the equilibrium constant
6
4
k
1
½CH
4
ꢃ
B
B½COꢃ þ C ½H
2
2
k
ꢃ
C7
A5
8
1 þ
pffiffiffiffiffiffiffiffi
@
½
H
2
Oꢃ
2
1
½CH ꢃ
2
4
A
½H
2
ꢃ þ 1
D½CO ꢃ þ E pffiffiffiffiffiffiꢀ pffiffiffiffiffiffi
of CO adsorption and desorption steps (step (14)), and A, B, C, D,
and E are constants composed of the rate constants and equilib-
rium constants shown in Scheme 1 (Table 5).
½
H
ꢃ
A
½H ꢃþ1
ð31Þ
The underlying assumptions and the resulting functional form
of this equation are consistent with all isotopic data and with the
These three equations were compared with all rate data using
linear regression analyses with minimization of residuals (Appen-

A. Yamaguchi, E. Iglesia / Journal of Catalysis 274 (2010) 52–63
59
Table 5
16
Relations between kinetic parameters in Eq. (28) for
CH
4
ꢀCO
2
and CH
4
ꢀH
2
O reactions and rate and equi-
librium constants for elementary steps.
14
Kinetic parameter
A
1
1
2
0
8
6
4
2
0
sffiffiffiffiffiffi
k
k
ꢀ1
2
1
K
7
B
C
k
ꢀ3
K
8
sffiffiffiffiffiffi
1
k
ꢀ5
K
7
D
E
k
3
k
4
qffiffiffiffi
k
ꢀ4
1
k
5
K
7
*
dix II). For O and ꢄ as MASI, the values of B, or D and E in Eq. (30)
were determined to be negative, indicating that such an assump-
tion is inconsistent with the rate data. Indeed, Eq. (30) predicts that
rates would increase with increasing CO pressure, in contradiction
0
2
4
6
8
10
12
14
16
-1
Experimental r (moles (g-atom surface Pd-s) )
f
Fig. 7. Predicted experimental forward CH
d) Eq. (31) versus experimental forward CH
CH O reactions on 1.6% wt. Pd/ZrO (1023) at 823 K.
ꢀH
4
turnover rates using (j) Eq. (29) and
turnover rates for both CH and
*
with the data in Fig. 6. We conclude that O is not present at signif-
(
4
4
ꢀCO
2
icant coverages during steady-state CH
Thus, we retain Eqs. (29) and (31) and estimate their respective
parameters using all CH and CH O rate data at 823 K;
their respective parity plots are shown in Fig. 7, and the resulting
rate parameters are reported in Table 6. Eq. (31) provides a more
4
reforming on Pd clusters.
4
2
2
4
ꢀCO
2
4
ꢀH
2
Table 6
2
accurate description of the rate data than Eq. (29) (R = 0.986 vs.
Rate parameters of CH
823 K and correlation coefficients for the rate equations.
4
reactions on 1.6% wt. Pd/ZrO
2
(1023) for CH
4
reactions at
Eq. (33)c
*
0
.952), indicating that C and ꢄ are the MASI during CH
4
reforming
*
reactions and that the inhibition effects of CO reflect its effect on C
Equation
Eq. (29)a
Eq. (31)a
Eq. (32)b
*
*
coverages during catalysis. The choice of C (instead of CO ) as the
titrant for sites is consistent with the low CO coverages expected
from the heat of molecular adsorption of CO on Pd. Only three of
the six kinetic parameters in Eq. (31) are required to describe the
rate data (Supplementary material; Fig. S4). Forward CH
ꢀ1 ꢀ1
k
1
(kPa
s
)
2.3
0.92
–
–
–
–
4.3
–
1.7
0.74
1.8
0.1
5.6
5.3
–
1.9
1.2
–
–
–
2.1
1.1
–
–
–
–
–
–
*
A (kPaꢀ1/2
)
ꢀ
ꢀ
1
ꢀ1
B (kPa
C (kPa
D (kPa
s
)
1/2 ꢀ1
s
)
ꢀ
1 ꢀ1
s
)
4
turnover
ꢀ
1/2 ꢀ1
E (kPa
s
)
–
–
rates for CH
4
ꢀCO
2
reactions are given by:
K
8
(kPa)
B/D
B/E (kPa
–
–
0.30
–
0.975
k
1
½CH
4
ꢃ
ꢀ1/2
)
–
0.41
0.984
r
f
¼
! ;
2
2
B
D
R
(ꢀ)
0.952
0.986
pffiffiffiffiffiffiffiffi
k ½CH ꢃ
½COꢃ
1
4
ðA ½H
2
ꢃ þ 1Þ
pffiffiffiffiffiffiffiffi
þ 1
a
b
c
Kinetic parameters from CH
Kinetic parameters from CH
Kinetic parameters from CH
4
4
4
ꢀCO
2
2
and CH
rate data.
4
ꢀH O rate data.
2
ðA ½H
2
ꢃ þ 1Þ ½CO ꢃ
2
ꢀCO
sffiffiffiffiffi
ꢀH
2
O rate data.
k
ꢀ1
1
B
D
k
ꢀ3
A ¼
;
¼ k
3 8
K
ð32Þ
k
2
K
7
Forward CH
4
turnover rates for CH
4
ꢀH
2
O reactions are given by:
CH
4
ꢀCO
2
or CH
4
ꢀH
2
O rate data were used in the regression (Ta-
nor H O is in-
ble 7). These similar values indicate that neither CO
volved in determining the magnitude of these two terms,
consistent with Scheme 1, in which their respective values depend
only on CH
small differences in measured CO
(Fig. 4) reflect their respective roles in providing O species at dif-
ferent rates; these O species, even at their low prevalent surface
coverages, act as the reactive scavengers that determine the con-
centration of the C species formed via dissociation of both CH
2
2
k
1
½CH ꢃ
4
r
f
¼
! ;
pffiffiffiffiffiffiffiffi
2
B
E
pffiffiffiffiffiffiffiffi
k ½CH ꢃ
½COꢃ ½H ꢃ
1
4
2
ðA ½H
2
ꢃ þ 1Þ
pffiffiffiffiffiffiffiffi
þ 1
ꢄ
*
4
3
dissociation and CH –H recombination rates. The
ðA ½H
2
ꢃ þ 1Þ
½H
2
Oꢃ
2
2
and H O reforming rates
sffiffiffiffiffiffi
sffiffiffiffiffiffi
*
k
k
ꢀ1
1
B
E
k
ꢀ3
k
ꢀ4
1
A ¼
;
¼ k
ð33Þ
*
2
K
7
4
k
5
K
8
K
7
*
The parameters for the best fit were estimated using the data of
CH reactions for Eq. (32) and CH ꢀH O reactions for Eq.
ꢀCO
33) at 823 K separately. The parity plots are shown in Fig. 8 and
4
*
and CO. The C species from CH
crease in coverage with increasing concentration of O formed in
4
and CO dissociation, which de-
4
2
4
2
*
(
*
the rate parameters in Table 6. Three parameters are used to fit
the respective data to each of Eqs. (29), (32), and (33), but the good-
ness of fit was significantly better for Eqs. (32) and (33) (R = 0.975
CO
2
and H
2
O activation, represent the relevant C in the denomina-
and H O influence the
tor of Eqs. (32) and (33); as a result, CO
magnitude of this term and the severity of CO inhibition effects
to different extents.
2
2
2
2
*
and 0.984) than for Eq. (29) (R = 0.952), indicating that C and ꢄ are
the most plausible MASI during CH reforming reactions.
These rate equations (Eq. (32) for CH and Eq. (33) for
ꢀCO
ꢀH O) were used to estimate the relevant kinetic parameters
from rate data at each temperature; the values of these parameters
are shown in Table 7. Rate constants for C–H bond activation (k
and for the H inhibition parameter (A) were similar whether
Fig. 9 shows Arrhenius plots for C–H bond activation rate con-
4
stants with CO
on 1.6% wt. Pd/ZrO
CO and H O are very similar (84 and 81 kJ mol , respectively; Ta-
2
and H
2 1
O co-reactants (k in Eqs. (32) and (33))
4
2
CH
4
2
2
(1023). Measured activation energies with
ꢀ
1
2
2
1
)
ble 7), as expected from the identical kinetic origins of the two rate
constants. These activation barriers are somewhat smaller than
2

6
0
A. Yamaguchi, E. Iglesia / Journal of Catalysis 274 (2010) 52–63
and H
2
inhibition, deactivation consequences of temperature,
A
1
1
1
1
6
4
2
0
8
6
4
2
0
and/or transport and thermodynamic artifacts. Our intrinsic barri-
ers for C–H bond activation on Pd clusters are larger, however, than
those estimated from density functional theory calculations on
ꢀ1
ꢀ1
Pd(1 1 1) (63.7 kJ mol ), Pd-step (36.7 kJ mol ), and Pd-kink
ꢀ1
(
39.6 kJ mol ) [18]. It seems plausible that such differences may
*
reflect the strong binding of C on corners and edges, which leads
to their low activation barrier for C–H activation [21] and would
render such highly active structures unable to turnover because
*
of their permanent titration by C . It is also conceivable that theo-
retical treatments underestimate C–H activation barriers because
they neglect specific effects of clusters with significant atomic
mobility and possibly containing trace amounts of dissolved car-
bon under typical conditions of reforming catalysis. We remark,
however, that the experimental barriers reported here are those
of CH reactions, devoid of transport or thermodynamic corrup-
4
tions, on Pd clusters as they exist at conditions relevant to CH
reforming catalysis.
4
0
2
4
6
8
10
12
14
16
-1
Experimental r (moles (g-atom surface Pd-s) )
f
4
3.8. Effects of Pd cluster size on CH reforming turnover rates
B
1
1
1
1
6
4
2
0
8
6
4
2
0
Turnover rates were measured with CH
reactants at 923 K also on 1.6% wt. Pd/ZrO
Pd/ZrO
ꢀLa
centrations. These samples gave smaller Pd dispersions (3.5% and
5.3%, respectively) than the 1.6% wt. Pd/ZrO (1023) (8.9%) sample
for which detailed kinetic data were presented in earlier sections.
4
ꢀCO
2
and CH
(1123) and 1.6% wt.
4
ꢀH
2
O
2
2
2 3
O (1023) as a function of reactant and product con-
2
20
10
9
8
7
6
5
0
2
4
6
8
10
12
14
16
4
3
-1
Experimental r (moles (g-atom surface Pd-s) )
f
Fig. 8. Predicted experimental forward CH
CH reaction and (B) Eq. (33) for CH
ꢀCO
forward CH turnover rates for (A) CH and (B) CH
ꢀCO
Pd/ZrO (1023) at 823 K.
4
turnover rates using (A) Eq. (32) for
ꢀH O reaction versus experimental
ꢀH O reactions on 1.6% wt.
2
4
2
4
2
4
4
2
4
2
2
1
apparent activation energies reported previously from CO
2
reform-
1.00
1.05
1.10
1.15
1.20
1.25
ing rate data without specific mechanistic analysis on 1% wt. Pd/
03 K / T
1
ꢀ1
ꢀ1
Al
2 3 2
O (87.1 kJ mol [25] and 93.1 kJ mol [7]), 1% wt. Pd/TiO
ꢀ1
ꢀ1
(
91.5 kJ mol ) [7], 1% wt. Pd/CeO
2
/Al
O
2 3
(100 kJ mol ) [10], 1%
Fig. 9. Arrhenius plots for (d) CH
4
ꢀCO
2
and (N) CH
, for C–H bond activation on 1.6% wt. Pd/ZrO
4
ꢀH
2
O reactions of the rate
(1023).
ꢀ1
ꢀ1
constants, k
1
2
wt. Pd/MgO (114 kJ mol ) [7], and 1% wt. Pd/SiO
2
(141 kJ mol
)
[
7]. These previous values may contain unintended effects of CO
Table 7
Kinetic parameters of CH
4
reactions on 1.6% wt. Pd/ZrO
2
(1023) for CH
4
ꢀCO
2
and CH
4
ꢀH
2
O reactions.
a
H
2
Ob
Co-reactant
CO
2
Temperature (K)
823
1.9
1.2
0.30
ꢀ
873
4.0
1.0
0.29
ꢀ
923
8.0
0.51
0.17
ꢀ
973
12.2
0.19
0.14
ꢀ
823
2.1
1.1
ꢀ
873
4.3
1.0
ꢀ
923
7.9
973
12.9
0.18
ꢀ
ꢀ
1 ꢀ1
k
1
(kPa
s )
)
ꢀ
1/2
A (kPa
B/D
0.52
ꢀ
B/E (kPaꢀ1/2)
0.41
0.35
0.22
0.19
Activation energy (kJ mol 1)
ꢀ
84 ± 4
4 ± 2
81 ± 3
3 ± 1
Pre-exponential factor (105
s
ꢀ1
kPa )
ꢀ1
a
Parameters defined in Eq. (32).
Parameters defined in Eq. (33).
b

A. Yamaguchi, E. Iglesia / Journal of Catalysis 274 (2010) 52–63
61
Table 8
Kinetic parameters of CH
10
A
4
reactions on Pd/ZrO
ꢀCO
2
(1123), Pd/ZrO
2
ꢀLa
2 3
O (1023), and Pd/
ZrO
2
(1023) (1.6% wt. Pd) at 923 K for CH
4
2
and CH O reactions.
4
ꢀH
2
Pd
Catalyst
Pd/ZrO
CO a
2
(1123)
Ob
Pd/ZrO
2
ꢀLa
2
O
3
(1023)
Ob
Pd/ZrO
2
(1023)
Ob
a
a
Co-reactant
2
H
2
CO
2
H
2
CO
2
H
2
ꢀ
1 ꢀ1
k
1
(kPa
s
)
7.8
0.49
0.16
ꢀ
8.0
0.54
ꢀ
8.1
0.52
0.17
ꢀ
7.8
0.50
ꢀ
8.0
0.51
0.17
ꢀ
7.9
0.52
ꢀ
ꢀ
1/2
A (kPa
B/D
)
Pt
B/E (kPaꢀ1/2)
0.25
0.23
0.22
1
Ir
a
Parameters defined in Eq. (32).
Parameters defined in Eq. (33).
b
Rh
These data and their kinetic analysis were used to probe the conse-
quences of Pd cluster size on the rate constant for C–H bond
activation.
Ru
Ni
Rate data on these three catalysts were accurately described by
0.1
10
Eq. (32) for CH
and the rate constants (k
4
ꢀCO
2
reactions and Eq. (33) for CH
4
ꢀH
2
O reactions,
0
20
40
60
80
1
) for C–H bond activation and inhibition
Metal dispersion (%)
parameters are shown in Table 8 for all three samples. The C–H
bond activation rate constants did not vary with Pd dispersion
B
(
3.5–8.9%) within experimental accuracy (Table 8), in spite of a
Pd
monotonic increase in the fraction of exposed surface Pd atoms
at corner and edge location with increasing Pd dispersion [16].
Such insensitivity to structure, which contrasts the small increase
in C–H bond activation rates with increasing dispersion for other
metals (Fig. 10) [6], appears to reflect the predominant contribu-
tions of low-index planes on large Pd clusters to measured rates
and the difficulty in stabilizing small Pd metal clusters against sin-
tering at reforming reaction temperatures. Such low-index surfaces
prevail on large clusters and exhibit much higher reactivity than on
the other metals, apparently because of a concomitant stronger
Pt
1
Ir
Rh
*
binding of C on Pd compared with that on other metals. Such high-
*
er C binding energies, evident from CO inhibition effects that were
Ru
not evident on other Group VIII metals [5,6], make contributions
from edge and corner sites even less likely than on the other met-
Ni
*
0.1
als, as a result of their irreversible titration by C species more
0
20
40
60
80
strongly bound on corners and edges than on low-index planes.
Metal dispersion (%)
Supports (ZrO
stant for C–H bond activation or the H
2
and ZrO
2
ꢀLa
2
O
3
) did not influence the rate con-
inhibition parameter (Ta-
2
Fig. 10. Rate constants for C–H bond activation in CH
4 2 2
for (A) CO and (B) H O
reforming of CH at 873 K on different metal clusters as a function of metal
dispersion. The data for Ni, Ru, Rh, Ir, and Pt have been previously reported [5,6].
ble 8), consistent with the proposal that all elementary steps in
Scheme 1 proceed on Pd cluster surfaces. Detailed kinetic data
4
2 3
are not available on 1.6% wt. Pd/Al O (1023) because of difficulties
imposed by its less stable reforming rates (Fig. 1). A value for the
C–H bond activation rate constant can be estimated, however, from
initial reforming turnover rates (extrapolated to zero time-on-
4. Conclusions
stream) by correcting these rates for the (small) effects H
2
and
) and B/D (0.30) (Ta-
on Pd/Al (1023) was
at 823 K, which is quite similar to that measured on
A reaction mechanism and a detailed sequence of elementary
ꢀ1/2
CO pressures using the values of A (1.2 kPa
ble 7, Fig. S5). The value obtained for k
steps required for CO
catalysts were investigated by kinetic and isotopic tracer studies.
The initial C–H bond activation step is reversible during CH
2 2 4
or H O reforming of CH on supported Pd
1
2 3
O
ꢀ1 ꢀ1
2
.5 kPa
s
4
ꢀ
1 ꢀ1
Pd/ZrO
2
(1023) (1.9 kPa
s
) at the same temperature.
reforming reaction on Pd clusters even when overall reforming
reactions are far from equilibrium, a conclusion confirmed by rates
The k
.3 kPa
1
values for C–H bond activation steps at 873 K (4.0 and
ꢀ
1
ꢀ1
4
s
, with CO
2
2
and H O, respectively) on 1.6% wt. Pd/
of formation of CH
that are similar to those for chemical conversion. On other Group
VIII metals, C–H bond activation steps are irreversible and H does
not inhibit CH reforming reactions. CO and H O activation steps
are not quasi-equilibrated during CH and CH ꢀH O reac-
ꢀCO
tions on Pd clusters, consistent with the slow isotopic exchange be-
3
D isotopomers from CH
4
ꢀCO
2
ꢀD
2
mixtures
2
ZrO (1023) are four times larger than the highest values measured
on any other Group VIII catalysts [5,6] (Fig. 10) and more than 10
times larger than turnover rates extrapolated to the same disper-
sion of the Pd samples on the other metals. The high activity of
Pd surfaces for C–H bond activation leads to the rapid formation
2
4
2
2
4
2
4
2
*
*
13
of C , which scavenges active O species from co-reactants and pre-
vents the equilibration of their activation steps on Pd surfaces and
leads to reversibility for both C–H activation and C–O formation
steps. The reactivity of Pd surfaces in C–H activation and their ten-
tween CO and CO
HDO/D O ratio during reactions of CH
gas shift reactions also remained non-equilibrated during
CH and CH ꢀH O reactions, as expected from the irrevers-
ꢀCO
ible nature of CO and H O activation steps. CO dissociation steps
(to C and O ) are reversible but not quasi-equilibrated during
2
in CH
4
ꢀCO
2
ꢀ
CO mixtures and with the low
2
4
ꢀD O mixtures. Water–
2
4
2
4
2
*
dency to be inhibited by C during reforming appear to reflect their
2
2
*
*
*
strong affinity for C , which can lead to significant deactivations
1
3
and to bifunctional effects of support when reforming experiments
CH
4
reforming reactions, as shown from the C content in CO
1
2
12 16
12 16
13 18
are conducted at low H
2
O/CH
4
or CO
2
/CH
4
reactant ratios.
and CO
2
during CH
4
ꢀ
C
O
2
ꢀ
C
Oꢀ
C O reactions. These

6
2
A. Yamaguchi, E. Iglesia / Journal of Catalysis 274 (2010) 52–63
ꢄ
ꢄ
ꢄ
ꢄ
kinetic and isotopic results on Pd clusters are consistent with a se-
quence of elementary steps similar to that proposed on other
Group VIII metals. The resulting rate equation accurately describes
k
2
½CH ꢃ½ꢄꢃ þ k
6
½CO ꢃ½ꢄꢃ ¼ kꢀ6½C ꢃ½O ꢃ
ðA8Þ
ðA9Þ
3
½
COꢃ
2
2
ꢄ
ꢄ
a
½ꢄꢃ þ k
6
½ꢄꢃ ¼ kꢀ6½C ꢃ½O ꢃ
K
8
*
measured rates and indicates that chemisorbed carbon (C ) and
The PSSH is also applied to [OH*].
unoccupied Pd atoms (ꢄ) are the most abundant surface intermedi-
ates. First-order rate constants, corresponding to C–H bond activa-
tion steps, are ꢅ10-fold larger than on all other Group VIII metals.
This high reactivity of Pd surfaces causes this step to become
2
ꢄ
ꢄ
ꢄ
ꢄ
ꢄ
k
4
½H
2
Oꢃ½ꢄꢃ þ kꢀ5½H ꢃ½O ꢃ ¼ kꢀ4½OH ꢃ½H ꢃ þ k
5
½OH ꢃ½ꢄꢃ
ðA10Þ
ðA11Þ
sffiffiffiffiffiffiffiffi
!
sffiffiffiffiffiffiffiffi
½H ꢃ
2
2
½H ꢃ
½
OHꢄꢃ
k
ꢀ4
þ k
5
¼ k
4
½H
2
Oꢃ½ꢄꢃ þ kꢀ5
½Oꢄꢃ
reversible, leading to inhibition effects by H
2
mediated by its
microscopic reverse and by C species introduced as a result of
the reversible nature of both CO and CH dissociation steps. The
reactive nature of Pd surface atoms appears to reflect their strong
affinity for both C and H atoms formed in CH activation steps,
K
7
K
7
*
qffiffiffiffiffiffi
qffiffiffiffiffiffi
k
k
4
5
k
k
ꢀ5
½H
2
ꢃ½Oꢄꢃ ¼ ½H
k
k
4
k
ꢀ5
½H
2
ꢃ½Oꢄꢃ
½H
2
Oꢃ½ꢄꢃ þ
2
Oꢃ½ꢄꢃ þ k
4
K
K
ꢄ
5
7
5
5
7
½
OH ꢃ ¼
qffiffiffiffiffiffi
ðA12Þ
k
k
ꢀ4
½H
2
ꢃ
b
þ 1
5
K
7
4
which lead to more stable transition states for the initial C–H bond
activation. The catalytic consequences of Pd cluster size are quite
weak over the limited range of Pd dispersion accessible at typical
in which
sffiffiffiffiffiffiffiffi
k
ꢀ4 ½H
2
ꢃ
conditions of CH
4
reforming. Supports did not influence C–H bond
b ¼
þ 1
ðA13Þ
ðA14Þ
k
5
K
7
activation or reforming turnover rates, but those supports with low
but detectable reactivity in the activation of co-reactants were able
to reverse the infrequent blockage of clusters surfaces by carbon
overlayers, thus inhibiting deactivation.
*
The PSSH equation for [O ] gives the equation.
2
ꢄ
ꢄ
ꢄ
ꢄ
k
3
½CO
2
ꢃ½ꢄꢃ þ k
5
½OH ꢃ½ꢄꢃ þ k
6
½CO ꢃ½ꢄꢃ ¼ kꢀ3½CO ꢃ½O ꢃ
ꢄ
ꢄ
ꢄ
ꢄ
þ kꢀ5½H ꢃ½O ꢃ þ kꢀ6½C ꢃ½O ꢃ
sffiffiffiffiffiffiffiffi
Acknowledgments
k
4
k
b
ꢀ5 ½H
2
ꢃ
½COꢃ½ꢄꢃ
2
2
ꢄ
½Oꢄꢃ
k
3
½CO
2
ꢃ½ꢄꢃ þ ½H
2
Oꢃ½ꢄꢃ þ
½O ꢃ½ꢄꢃ ¼ kꢀ3
b
K
7
K
8
This study was supported by BP as part of the Methane Conver-
sion Cooperative Research Program at the University of California
at Berkeley. We acknowledge Dr. Theo Fleisch (BP), Dr. Junmei
Wei and Ms. Cathy Chin (UC-Berkeley) for helpful technical discus-
sions. Transmission electron micrographs were kindly provided by
Professor Miguel Avalos Borja of the Centro de Ciencias de la Mate-
ria Condensada of the Universidad Nacional Autonoma de Mexico.
sffiffiffiffiffiffiffiffi
½
H
2
ꢃ
ꢄ
2
þ kꢀ5
½ꢄꢃ½O ꢃ þ
a
½ꢄꢃ
ðA15Þ
K
7
sffiffiffiffiffiffiffiffi
sffiffiffiffiffiffiffiffi!
½COꢃ
½H
2
ꢃ
k
ꢀ5 ½H
2
ꢃ
Oꢄꢃ
½
k
ꢀ3
þ kꢀ5
ꢀ
K
8
K
7
b
K
7
ꢂ
ꢃ
k
4
¼
½ꢄꢃ k
3
½CO
2
ꢃ þ ½H
2
Oꢃ ꢀ
a
ðA16Þ
ðA17Þ
b
Appendix I
k
b
4
k
3
½CO
2
ꢃ þ ½H
2
Oꢃ ꢀ
a
d
c
ꢄ
½
O ꢃ ¼
qffiffiffiffiffiffi
qffiffiffiffiffiffi½ꢄꢃ ¼ ½ꢄꢃ
A sequence of elementary steps for CH
4
reforming reactions on
½COꢃ
½H
ꢃ
k
½H
ꢃ
2
ꢀ5
2
K
7
k
ꢀ3
þ kꢀ5
ꢀ b
K
8
K
7
Pd-based catalysts is shown in Scheme 1. The steps (13) and (14)
give following formulas.
in which
½COꢃ
sffiffiffiffiffiffiffiffi
sffiffiffiffiffiffiffiffi
sffiffiffiffiffiffiffiffi
½
H
2
ꢃ
ꢄ
½
½
H ꢃ ¼
½ꢄꢃ
ðA1Þ
ðA2Þ
½H ꢃ
k
b
½H ꢃ
2
ꢀ5
2
K
7
c
¼ kꢀ3
þ kꢀ5
ꢀ
ðA18Þ
ðA19Þ
K
8
K
7
K
7
½
COꢃ½ꢄꢃ
ꢄ
CO ꢃ ¼
k4
b
K
8
d ¼ k ½CO ꢃ þ ½H Oꢃ ꢀ
a
3
2
2
ꢄ
The pseudo-steady-state hypothesis (PSSH) is applied to [CH ].
3
2
ꢄ
ꢄ
ꢄ
Eq. (A9) can be rewritten by the following equation using Eq. (A17).
k
1
½CH
4
ꢃ½ꢄꢃ ¼ kꢀ1½CH ꢃ½H ꢃ þ k
2
½CH ꢃ½ꢄꢃ
ðA3Þ
ðA4Þ
3
3
k
1
½CH
4
ꢃ
2
½COꢃ
2
d
c
ꢄ
ꢄ
½
CH ꢃ ¼
qffiffiffiffiffiffi
½ꢄꢃ
2
a
½ꢄꢃ þ k
6
½ꢄꢃ ¼ kꢀ6 ½C ꢃ½ꢄꢃ
ðA20Þ
ðA21Þ
3
½
H
2
ꢃ
K
8
k
ꢀ1
þ k
K
7
½COꢃ
6
K
a
þ k
Cꢄꢃ ¼
8
½ꢄꢃ
½
Forward CH
ing formula.
4
turnover rates in step (5) are expressed by the follow-
d
c
k
ꢀ6
2
ꢄ
ꢄ
ꢄ
Using Eqs. (A2), (A7), (A17), and (A21), we can obtain the following
formula.
r
f
¼ k
1
½CH
1
4
ꢃ½ꢄꢃ ꢀ kꢀ1½CH ꢃ½H ꢃ ¼ k
2
½CH ꢃ½ꢄꢃ
3
3
k
½CH ꢃ
4
2
2
¼
qffiffiffiffiffiffiffiffiffiffiffiffiffiffi½ꢄꢃ ¼
a
½ꢄꢃ
ðA5Þ
ðA6Þ
1
k
k
ꢀ1
½H ꢃ
2
½
ꢄꢃ ¼
ðA22Þ
þ 1
½COꢃ
2
K
7
a
þk
k
6
K
½COꢃ
8
d
c
1
þ
d
c
þ
þ
K
8
ꢀ6
in which
k
1
½CH
4
ꢃ
4
Forward CH turnover rates are expressed by the following formula
using Eq. (A5).
a
¼
qffiffiffiffiffiffi
k
k
ꢀ1
½H
2
ꢃ
þ 1
2
K
7
a
2
The concentration of free sites (*) can be derived from the site
balance equation:
rf ¼
a
½ꢄꢃ ¼
ðA23Þ
ꢂ
ꢃ
2
½
COꢃ
a
þk
k
6
K8
d
c
½COꢃ
1
þ
d
c
þ
þ
K
8
ꢀ6
ꢄ
ꢄ
ꢄ
½
ꢄꢃ þ ½C ꢃ þ ½O ꢃ þ ½CO ꢃ ¼ L
by evaluating the steady-state coverages of the various adsorbed
ðA7Þ
COꢃ
ꢄ
We conclude that
a
ꢆ k ½
6
a
) is
), as shown by the small
8
K
8
*
½COꢃ
much faster than CO dissociation (k
6
K
intermediates. The PSSH is applied to [C ].

A. Yamaguchi, E. Iglesia / Journal of Catalysis 274 (2010) 52–63
63
extent of isotopic scrambling in 12C16Oꢀ13C18O mixtures during
The goodness of fit was assessed from the respective correlation
coefficients (R).
CH
4
ꢀCO
2
reactions.
P
n
i¼1
ðr
f
ðexperimentalÞ ꢀ r
f
ðexperimentalÞ Þðr
f
ðpredictedÞ ꢀ r
f
ðpredictedÞ Þ
i
av
i
av
R ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðA33Þ
P
P
n
i¼1
2
n
2
ðr
f
ðexperimentalÞ ꢀ r
f
ðexperimentalÞ Þ
ðr
f
ðpredictedÞ ꢀ r
f
ðpredictedÞ Þ
i
av
i¼1
i
av
a
In these equations, r
age values of r
f
(experimental)av and r
f
(predicted)av are aver-
r
f
¼ ꢂ
ꢃ
ðA24Þ
2
½COꢃ
f
f
(experimental) and r (predicted), respectively.
a d
d
c
1
þ k
þ
þ
K
8
ꢀ6
c
Appendix C. Supplementary material
in which
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.jcat.2010.06.001.
k
1
½CH
4
ꢃ
k
1
½CH
4
ꢃ
a
¼
qffiffiffiffiffiffi
¼
pffiffiffiffiffiffiffiffi
ðA25Þ
ðA26Þ
k
k
ꢀ1
½H
2
ꢃ
A
½H ꢃ þ 1
þ 1
2
2
K
7
sffiffiffiffiffiffiffiffi
k
k
ꢀ4 ½H
2
ꢃ
b ¼
þ 1
References
5
K
7
0
1
sffiffiffiffiffiffiffiffi
sffiffiffiffiffiffiffiffi
sffiffiffiffiffiffiffiffi
[
1] S. Wang, G.Q.M. Lu, G.J. Millar, Energy Fuels 10 (1996) 896–904;
T. Inui, Appl. Oranometal. Chem. 15 (2001) 87–94.
½
COꢃ
½H
2
ꢃ
k
b
ꢀ5 ½H
2
ꢃ
½COꢃ
½H
2
ꢃB
1
C
A
c
¼ kꢀ3
þ kꢀ5
ꢀ
¼ kꢀ3
þ kꢀ5
@1 ꢀ qffiffiffiffiffiffi
K
8
K
7
K
7
K
8
K
7
k
k
ꢀ4
½H
2
ꢃ
[2] M.C.J. Bradford, M.A. Vannice, Catal. Rev. Sci. Eng. 41 (1999) 1–42.
[
þ 1
5
K
7
3] K.P. De Jong, J.W. Geus, Catal. Rev. Sci. Eng. 42 (2000) 481–510;
L. Dussault, J.C. Dupin, C. Guimon, M. Monthioux, N. Latorre, T. Ubieto, E.
Romeo, C. Royo, A. Monzon, J. Catal. 251 (2007) 223–232;
I. Kvande, D. Chen, Z. Yu, M. Ronning, A. Holmen, J. Catal. 256 (2008) 204–
214.
ðA27Þ
ðA28Þ
k
4
k
4
k
1
½CH
4
ꢃ
d ¼ k
3
½CO
2
ꢃ þ ½H
2
Oꢃ ꢀ
a
¼ k
3
½CO
2
ꢃ þ qffiffiffiffiffiffi
½H
2
Oꢃ ꢀ pffiffiffiffiffiffiffiffi
b
k
k
ꢀ4
½H
2
K
7
ꢃ
A
½H ꢃ þ 1
þ 1
2
5
[
4] S.J. Blanksby, G.B. Ellison, Acc. Chem. Res. 36 (2003) 255–263.
qffiffiffiffiffiffi
k
ꢀ4
½H
2
7
ꢃ
[5] J. Wei, E. Iglesia, Angew. Chem. Int. Ed. 43 (2004) 3685–3688;
J. Wei, E. Iglesia, Phys. Chem. Chem. Phys. 6 (2004) 3754–3759;
J. Wei, E. Iglesia, J. Catal. 225 (2004) 116–127;
When k
ꢆ 1,
5
K
sffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffi
½
COꢃ
½H
2
ꢃ
J. Wei, E. Iglesia, J. Catal. 224 (2004) 370–383.
c
ꢇ kꢀ3
þ kꢀ5
¼ B½COꢃ þ C ½H
2
ꢃ
ðA29Þ
[
[
6] J. Wei, E. Iglesia, J. Phys. Chem. B 108 (2004) 7253–7262;
J. Wei, E. Iglesia, J. Phys. Chem. B 108 (2004) 4094–4103.
7] A. Erdohelyi, J. Cserenyi, E. Papp, F. Solymosi, Appl. Catal. A 108 (1994) 205–
219.
8] K. Nagaoka, K. Aika, Bull. Chem. Soc. Jpn. 74 (2001) 1841–1846.
9] R. Craciun, W. Daniell, H. Knozinger, Appl. Catal. A 230 (2002) 153–168.
K
8
K
7
k
4
k
1
½CH
4
ꢃ
d ꢇ k
3
½CO
2
ꢃ þ qffiffiffiffiffiffi½H
2
Oꢃ ꢀ pffiffiffiffiffiffiffiffi
[
[
k
k
ꢀ4
½H
2
K
7
ꢃ
A
½H ꢃ þ 1
2
5
[
[
10] P.G. Schulz, M.G. Gonzalez, C.E. Quincoces, C.E. Gigola, Ind. Eng. Chem. Res. 44
(2005) 9020–9029.
11] C.E. Gigola, M.S. Moreno, I. Costilla, M.D. Sanchez, Appl. Surf. Sci. 254 (2007)
½
H
2
Oꢃ
k
1
½CH
4
ꢃ
¼
D½CO
2
ꢃ þ E pffiffiffiffiffiffiffiffiꢀ pffiffiffiffiffiffiffiffi
ðA30Þ
½
H
2
ꢃ
A
½H
2
ꢃ þ 1
325–329;
L.S.F. Feio, C.E. Hori, S. Damyanova, F.B. Noronha, W.H. Cassinelli, C.M.P.
Marques, J.M.C. Bueno, Appl. Catal. A 316 (2007) 107–116;
W.H. Cassinelli, L.S.F. Feio, J.C.S. Araujo, C.E. Hori, F.B. Noronha, C.M.P. Marques,
J.M.C. Bueno, Catal. Lett. 120 (2008) 86–94.
4
Forward CH turnover rates are expressed by the following formula
using the parameters (k , K , A, B, C, D, and E), which do not contain
1 8
any residual dependences on concentrations of reactants and
products.
[
12] T. Osaki, T. Horiuchi, K. Suzuki, T. Mori, Catal. Lett. 44 (1997) 19–21;
Z. Zhang, X.E. Verykios, Catal. Lett. 38 (1996) 175–179;
H.-Y. Wang, C.-T. Au, Catal. Lett. 38 (1996) 77–79.
k
1
½CH
2
4
ꢃ
[13] J.E. Benson, H.S. Hwang, M. Boudart, J. Catal. 30 (1973) 146–153.
[14] G.L. Price, E. Iglesia, Ind. Eng. Chem. Res. 28 (1989) 839–844.
[15] R.-J. Liu, P.A. Crozier, C.M. Smith, D.A. Hucul, J. Blackson, G. Salaita, Appl. Catal.
A 282 (2005) 111–121;
pffiffiffiffiffiffi
A
½H
ꢃþ1
r
f
¼ 2
0
@
1
0
@
1
A
3
5
D½CO2ꢃþEp½H2ffiffiOffiffiꢃffi
k1½CH4ꢃ
2
ꢂ
ꢃ
pffiffiffiffiffiffi
ꢀ
pffiffiffiffiffi
ꢀ
ꢁ
½H2ꢃ
A
½H2ꢃþ1
k
1
½CH
2
4
ꢃ
B½COꢃþC ½H
2
ꢃ
½COꢃ
4
pffiffiffiffiffiffi
A
pffiffiffiffiffiffi
1
þ
½H Oꢃ
k
½CH
ꢃ
þ
þ
A
½H
ꢃþ1
D½CO2ꢃþE
p
2
ffiffiffiffiffi
ꢀ
p
1
ffiffiffiffi
4
ffi
B½COꢃþC ½H
2
ꢃ
K
8
M.S. Moreno, F. Wang, M. Malac, T. Kasama, C.E. Gigola, I. Costilla, M.D.
Sanchez, J. Appl. Phys. 105 (2009) 083531.
½
H2ꢃ
A
½H2ꢃþ1
[
[
16] R. Van Hardeveld, F. Hartog, Surf. Sci. 15 (1969) 189–230.
17] K. Klier, J.S. Hess, R.G. Herman, J. Chem. Phys. 107 (1997) 4033–4043;
G. Jones, J.G. Jakobsen, S.S. Shim, J. Kleis, M.P. Andersson, J. Rossmeisl, F. Abild-
Pedersen, T. Bligaard, S. Helveg, B. Hinnemann, J.R. Rostrup-Nielsen, I.
Chorkendorff, J. Sehested, J.K. Norskov, J. Catal. 259 (2008) 147–160.
18] Z.-P. Liu, P. Hu, J. Am. Chem. Soc. 125 (2003) 1958–1967.
19] M. Boudart, G. Djega-Mariadasson, Kinetics of Heterogeneous Catalytic
Reactions, Princeton University Press, Princeton, NJ, 1984.
ðA31Þ
In this final equation, k
1
is the rate constant for C–H bond activation
is the equilibrium constant of CO
in the elementary step (5) and K
8
[
[
adsorption and desorption steps (step (14)), and A, B, C, D, and E are
constants composed of the rate constants and equilibrium constants
shown in Scheme 1.
[
[
20] D.R. Stull, E.F. Westrum Jr., G.C. Sinke, The Chemical Thermodynamics of
Organic Compounds, Krieger, Malabar, FL, 1987.
21] M. Neurock, M. Janik, S. Wasileski, R.A. van Santen, J. Am. Chem. Soc.,
submitted for publication (private communication).
Appendix II
[
[
22] H. Conrad, G. Ertl, E.E. Latta, Surf. Sci. 41 (1974) 435–446.
23] H. Conrad, G. Ertl, J. Koch, E.E. Latta, Surf. Sci. 43 (1974) 462–480;
P.W. Davies, R.M. Lambert, Surf. Sci. 111 (1981) L671–L674;
R.D. Ramsier, K.-W. Lee, J.J.T. Yates, Surf. Sci. 322 (1995) 243–255.
24] E.H. Voogt, L. Coulier, O.L.J. Gijzeman, J.W. Geus, J. Catal. 169 (1997) 359–
364.
Rate parameters were determined using linear regression anal-
yses with minimization of residuals (S):
[
n
X
2
S ¼
ðr
f
ðexperimentalÞ ꢀ r
f
ðpredictedÞiÞ
ðA32Þ
[25] F. Solymosi, G. Kutsan, A. Erdohelyi, Catal. Lett. 11 (1991) 149–156.
i
i¼1