Quantitative analysis of the reactivity of formate species seen by DRIFTS over a Au/Ce(La)O2 water-gas shift catalyst: First unambiguous evidence of the minority role of formates as reaction intermediates

Article abstract of DOI:10.1016/j.jcat.2007.02.013

The reactivity of the species formed at the surface of a Au/Ce(La)O₂ catalyst during the water-gas shift (WGS) reaction were investigated by operando diffuse reflectance Fourier transform spectroscopy (DRIFTS) at the chemical steady state during isotopic transient kinetic analyses (SSITKA). The exchanges of the reaction product CO₂ and of formate and carbonate surface species were followed during an isotopic exchange of the reactant CO using a DRIFTS cell as a single reactor. The DRIFTS cell was a modified commercial cell that yielded identical reaction rates to that measured over a quartz plug-flow reactor. The DRIFTS signal was used to quantify the relative concentrations of the surface species and CO₂. The analysis of the formate exchange curves between 428 and 493 K showed that at least two levels of reactivity were present. Slow formates displayed an exchange rate constant 10- to 20-fold slower than that of the reaction product CO₂. Fast formates were exchanged on a time scale similar to that of CO₂. Multiple nonreactive readsorption of CO₂ took place, accounting for the kinetics of the exchange of CO₂(g) and making it impossible to determine the number of active sites through the SSITKA technique. The concentration (in mol g^-1) of formates on the catalyst was determined through a calibration curve and allowed calculation of the specific rate of formate decomposition. The rate of CO₂ formation was more than an order of magnitude higher than the rate of decomposition of formates (slow + fast species), indicating that all of the formates detected by DRIFTS could not be the main reaction intermediates in the production of CO₂. This work stresses the importance of full quantitative analyses (measuring both rate constants and adsorbate concentrations) when investigating the role of adsorbates as potential reaction intermediates, and illustrates how even reactive species seen by DRIFTS may be unimportant in the overall reaction scheme.

Full text of DOI:10.1016/j.jcat.2007.02.013

Journal of Catalysis 247 (2007) 277–287
www.elsevier.com/locate/jcat
Quantitative analysis of the reactivity of formate species seen by DRIFTS
over a Au/Ce(La)O2 water–gas shift catalyst: First unambiguous evidence
of the minority role of formates as reaction intermediates
F.C. Meunier a,∗, D. Reid a, A. Goguet a, S. Shekhtman a, C. Hardacre a, R. Burch a, W. Deng b,
M. Flytzani-Stephanopoulos b
a
CenTACat, School of Chemistry and Chemical Engineering, Queen’s University Belfast, Belfast, BT9 5AG, Northern Ireland, UK
b
Department of Chemical and Biological Engineering, Tufts University, Medford, MA 02155, USA
Received 24 January 2007; revised 13 February 2007; accepted 13 February 2007
Available online 21 March 2007
Abstract
The reactivity of the species formed at the surface of a Au/Ce(La)O catalyst during the water–gas shift (WGS) reaction were investigated
2
by operando diffuse reflectance Fourier transform spectroscopy (DRIFTS) at the chemical steady state during isotopic transient kinetic analyses
(SSITKA). The exchanges of the reaction product CO and of formate and carbonate surface species were followed during an isotopic exchange
2
of the reactant CO using a DRIFTS cell as a single reactor. The DRIFTS cell was a modified commercial cell that yielded identical reaction
rates to that measured over a quartz plug-flow reactor. The DRIFTS signal was used to quantify the relative concentrations of the surface species
and CO . The analysis of the formate exchange curves between 428 and 493 K showed that at least two levels of reactivity were present. “Slow
2
formates” displayed an exchange rate constant 10- to 20-fold slower than that of the reaction product CO . “Fast formates” were exchanged on
2
a time scale similar to that of CO . Multiple nonreactive readsorption of CO took place, accounting for the kinetics of the exchange of CO (g)
2
2
2
−1
and making it impossible to determine the number of active sites through the SSITKA technique. The concentration (in mol g ) of formates on
the catalyst was determined through a calibration curve and allowed calculation of the specific rate of formate decomposition. The rate of CO
2
formation was more than an order of magnitude higher than the rate of decomposition of formates (slow + fast species), indicating that all of the
formates detected by DRIFTS could not be the main reaction intermediates in the production of CO . This work stresses the importance of full
2
quantitative analyses (measuring both rate constants and adsorbate concentrations) when investigating the role of adsorbates as potential reaction
intermediates, and illustrates how even reactive species seen by DRIFTS may be unimportant in the overall reaction scheme.
© 2007 Elsevier Inc. All rights reserved.
Keywords: Operando; In situ; SSITKA; DRIFT; FTIR; Spectroscopy; Water–gas shift; Formate; Carbonate; Surface species; Gold; Ceria; Spectator; Reaction
mechanism
1. Introduction
cently reviewed the field of WGS over Au-based catalysts and
proposed a unified WGS mechanism in which the redox route
would derive from other routes at the higher temperatures, at
which adsorbate surface coverage becomes low [21]. An impor-
tant feature of this model is the recognition that the dominant
mechanism will be a function of the choice of catalyst and the
experimental conditions.
Following earlier work by Shido and Iwasawa [3,22], Jacobs
et al. [23–27] proposed that the formates generated by the reac-
tion of CO with bridging OH groups associated with the Ce3+
defects were main reaction intermediates. The formates were
Ceria-supported noble metal-based materials are currently
receiving much interest as promising oxidation-resistant and
durable low-temperature water–gas shift (WGS; CO + H2O →
CO2 + H2) catalysts [1–12]. The role of formates as poten-
tial reaction intermediates has been much discussed, often as
an alternative route to a redox mechanism [13–20]. Burch re-
*
Corresponding author. Fax: +44 28 90 974890.
E-mail address: f.meunier@qub.ac.uk (F.C. Meunier).
0021-9517/$ – see front matter © 2007 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2007.02.013

278
F.C. Meunier et al. / Journal of Catalysis 247 (2007) 277–287
thought to decompose to CO2(g) via surface carbonate species.
Kinetic isotope effect and isotopic tracer studies suggested that
the rate-limiting step of the forward formate decomposition was
the C–H bond rupture, aided by the noble metal [25]. Behm et
al. [28] carried out a quantitative analysis of the rate of formate
surface decomposition over a Au/CeO2 catalyst and proposed
that formates could account for about half of the CO2 pro-
duced. This calculation was based on the nonsteady-state mea-
surement of the rate constant of formate decomposition (i.e.,
during desorption under 2% H2O in N2) and an estimate of the
formate surface coverage determined via TPD under N2 of a
catalyst pretreated under WGS conditions. Meunier et al. [29]
recently proposed that formates can potentially be reaction in-
termediates over Pt/CeO2 above 473 K, whereas the formate
species seen by DRIFTS were merely reaction spectators be-
low this temperature. It is important to stress that in none of
the aforementioned studies, including the one carried out by
some of us [29], were the formate decomposition rate constant
and surface coverage determined simultaneously at the chemi-
cal steady state under WGS reaction conditions.
Meunier et al. showed that the reactivity of surface species
over Pt/CeO2 can dramatically depend on the experimental
procedure used [30]. Based on DRIFTS analyses combined
with the utilization of isotopic tracers, steady-state experi-
ments showed that formates were less reactive than carbonyl
and carbonate species under steady-state conditions, whereas
the reverse trend was observed during the desorption-type
nonsteady-state experiments carried out in an inert purge gas.
The operando DRIFTS–SSITKA method used in the present
study uses a single catalytic bed, which allows DRIFTS charac-
terization of the surface of the very same catalyst that is respon-
sible for the catalytic activity measured at the cell exit by gas
chromatography or mass spectrometry [8]. This methodology is
a powerful tool for an in-depth investigation of catalysts under
reaction conditions, similar to the method developed earlier for
transmission FTIR by Chuang et al. [31,32].
the surface during the calcination step at 673 K. The BET spe-
cific surface area was 161 m2 g−1. The purity of the gases used
(H2, Kr, CO, and Ar, supplied by BOC) was >99.95%. The
13CO was 99% pure (supplied by Cambridge Isotope Laborato-
ries Inc.).
The experimental setup consisted of an in situ high-tempera-
ture diffuse reflectance IR cell (from Spectra-Tech) fitted with
ZnSe windows. The DRIFTS cell was located in a Bruker
Equinox 55 spectrometer, operating at a resolution of 4 cm−1
.
The reactor crucible was modified to ensure plug-flow condi-
tions throughout the catalyst bed. The interface between the ce-
ramic reactor and the metallic base plate was sealed with some
PTFE tape to prevent any sample bypass. The original porous
bed to support the sample was replaced by an inert metallic
mesh. The reaction rates measured using the cell was equal to
that measured in a conventional tubular plug-flow reactor (vide
infra). No conversion of reactant was observed when the cru-
cible was heated up to the reaction temperature in the absence
of the catalyst. The cell was connected to the feed gas cylinders
through low-volume stainless-steel lines. The gas flows were
controlled by Aera mass flow controllers, which were regularly
calibrated. A 4-way valve was used to allow fast switching be-
tween two reaction feeds, when appropriate. High-purity water
was introduced using a single saturator, which delivered a con-
stant and accurate level of water. Note that the water delivery
was totally unaffected by the valve switching (as determined by
MS or GC analyses [29]).
The amount of catalyst used in the DRIFTS reactor was
78 2 mg (particle diameter <150 µm). Unless stated other-
wise, the material was brought up to reaction temperature under
Ar, and then 7 vol% of water was added to the Ar stream.
A reference DRIFTS single scan was recorded after stabiliza-
tion of the signal. The reaction mixture (typically 2% CO +
7% H2O in Ar) was subsequently introduced at a total flow rate
of 100 ml min−1. The reaction flow was going down the reactor
bed; therefore, the upper layer of the catalyst (which is probed
by the DRIFTS technique) was the front of the bed.
The present paper deals with reactivity in terms of isotopic
exchange of the surface species formed over an Au/Ce(La)O2
catalyst under forward WGS conditions (i.e., 2% CO + 7%
H2O). The main objective was to investigate the reactivity of the
surface formates species observed under reaction conditions.
A fully quantitative analysis was carried out to determine both
the rate constant of formate decomposition and the surface cov-
erage of formate under steady-state reaction conditions. This
methodology allowed us to compare the specific rate of CO2
formation to that of formate decomposition and to conclude that
the formates detected by DRIFTS are not important reaction in-
termediates in CO2 formation.
Steady-state conditions in terms of the concentration of the
surface species measured by DRIFTS were reached in less than
30 min. The assignment of the IR bands and the integration
method are described in Section 3. The IR data are reported
as log1/R, with R = I/I0, where R is the sample reflectance,
I0 is the intensity measured on the sample after exposure to
7% water, and I is the intensity measured under reaction con-
dition. The function log1/R (=“absorbance”) gives a better
linear representation of the band intensity against sample sur-
face coverage than that given by the Kubelka–Munk function
for strongly absorbing media such as Pt/CeO2 [33].
The determination of the level of exchange of the surface
species during the isotopic switch was carried out using a least
squares method based on the linear combination of the ini-
tial 12C-containing and fully exchanged 13C-containing spectra.
The 2650–3100 cm−1 region was used for the formates, 2400–
2250 cm−1 was used for CO2, and 750–1000 cm−1 was used
for the carbonates (Fig. 4). This method appeared to be accurate
and led to an improved signal-to-noise ratio compared with that
reported earlier [30]. A solver algorithm based on Microsoft
2. Experimental
The catalyst used in this study was a 0.6 at% Au + 7.3 at%
La/CeO2 (abbreviated 0.6AuCL) prepared by NaCN leaching
of a higher Au-loading parent material [7]. The bulk composi-
tion was determined by ICP. The concentration of La near the
surface was determined by XPS to be 19.1 at%, much higher
than the bulk content of 7.3 at% La, due to La enrichment of

F.C. Meunier et al. / Journal of Catalysis 247 (2007) 277–287
279
Excel was used for the parameter optimization. The 12C and
13C responses were symmetric, so any set of exchange curves
could be used to compare the relative rate of exchange.
observed. The apparent activation energy of CO2 formation was
40.0 kJ mol−1
.
3.2. Nature of the IR spectra and assignment of IR bands
3. Results
The in situ DRIFTS spectrum observed over the 0.6AuCL
at 493 K under 7% H2O in Ar (using a mirror as background)
is shown in Fig. 2a. The sample surface was highly hydrox-
ylated, as evidenced by the large band observed in the 3700–
2500 cm−1 region, which corresponds to the stretching mode of
hydrogen-bonded hydroxyls. The band at 3658 cm−1 is associ-
ated with doubly coordinated hydroxyl groups (type II) [34].
The broad band observed in the region between 1800 and
1100 cm−1 indicated the presence of several OCO-containing
species, such as carbonates and carboxylates, typically ob-
served over basic solids, such as ceria [35–38]. The presence
of carbonate species was clearly evidenced by the 855-cm−1
band, assigned to the carbonate out-of-plane bending vibration.
The spectrum of the sample at steady-state conditions at 493
K under 2% CO + 7% H2O in Ar (using a mirror as back-
ground) exhibited some marked modifications (Fig. 2b). New
bands at 2946 and 2830 cm−1 were observed, which were as-
sociated with the combination band (C–H) + νs(OCO) and
ν(C–H) of a bidentate formate species, respectively [38]. The
additional bands in the 1800–1200 cm−1 region were due in
part to the stretching vibration modes (symmetric and asym-
metric) of the OCO group of formate and carbonate species.
The formates were likely formed by reaction of CO with the
ceria type II hydroxyls, which were no longer observable. Due
to the complexity of the band at 1800–1200 cm−1, the decon-
volution and interpretation of this region was not attempted.
The band observed at 2125 cm−1 is not related to any metal-
bonded carbonyl groups, because no shift of the band occurs
when switching to 13CO (see the next paragraph), but rather to
3.1. Sample activity and DRIFTS reactor evaluation
The WGS activity of the 0.6AuCL was measured between
398 and 473 K in our modified DRIFTS cell (in Belfast) and
compared with that evaluated in a traditional quartz plug-flow
reactor (in Medford) using the same feed, that is, 2% CO +
7% H2O in Ar (Fig. 1). Only the rates of CO2 formation for
which the CO conversion was <15% (i.e., essentially differ-
ential conditions) are reported in Fig. 1. A perfect agreement
between the rates measured over the two different reactors was
Fig. 1. Arrhenius plot relating to the rate of CO formation measured over the
2
0.6AuCL sample (!) in a quartz plug-flow reactor in Medford and (") in the
12
modified DRIFT cell in Belfast. The feed was 2% CO + 7% H O in Ar or
2
He.
Fig. 2. In situ DRIFT spectra obtained over the 0.6AuCL at steady-state conditions at 493 K under (a) 7% H O in Ar and (b) the WGS feed: 2% CO + 7% H O in
2
2
Ar. A mirror signal was used as background.

280
F.C. Meunier et al. / Journal of Catalysis 247 (2007) 277–287
12
13
Fig. 3. In situ DRIFT spectra obtained over the 0.6AuCL at steady-state conditions at 493 K under WGS feed (a) 2% CO + 7% H O in Ar and (b) 2% CO +
2
7% H O in Ar. The measured CO conversion was 26%. A spectrum of the same sample under 7% H O /Ar at the same temperature was used as background.
2
2
a normally forbidden electronic transition of partially reduced
ceria [39]. This actually indicates that ceria was partly reduced
under our reaction conditions.
tures. The 12CO2 and 13CO2 relative concentrations were mea-
sured using the gas-phase DRIFTS bands at 2400–2250 cm−1
(Fig. 3b). Formate and carbonate exchange curves were deter-
mined using the 3100–2700 and 1000–750 regions, respectively
(Fig. 4). The data shown in Fig. 4 revealed that the exchange of
formates and carbonates over the sample was not fully com-
pleted after 30 min under labeled feed, even at the highest
temperature investigated (493 K), at which it reached 95%. The
level of exchange of formates after 30 min was 73% at 458 K
and 48% at 428 K.
Although the exchange of formates and carbonates was only
partial after 30 min, the exchange of CO2 was always essen-
tially completed within 20 min at the three temperatures inves-
tigated (see the 13CO2 signal in Figs. 5 and 6 for the cases
at 428 and 493 K; the data at 458 K are not shown for the
sake of brevity).The curves associated with the CO2 exchange
were approximated by a single exponential function (Fig. 7),
and the corresponding rate constant Arrhenius plot (Fig. 7, in-
sert) yielded an apparent activation energy of CO2 exchange of
23.7 kJ mol−1. It is worth stressing that the apparent activation
energy for the formation of CO2 was equal to 40.0 kJ mol−1
(see Section 4.1), significantly different from that of the ex-
change of CO2.
The spectrum obtained under WGS conditions (i.e., 493 K
under 2% CO + 7% H2O in Ar) is also shown using the spec-
trum of the sample recorded in the presence of water only as
background (Fig. 3a). The minor wavenumber differences ob-
served between Figs. 2b and 3a are due simply to the baseline
corrections. The formate (2945 and 2832 cm−1) and carbonate
(853 cm−1) bands were clearly visible. The spectrum obtained
30 min after switching to the corresponding 13C-labeled feed
(i.e., 2% 13CO + 7% H2O in Ar) is reported in Fig. 3b. Both
the formate and carbonate bands shifted to lower wavenumbers,
for example, from 2832 to 2816 cm−1 for the formate and from
853 to 834 cm−1 for the carbonates. The isotopic exchange
of these surface species was almost completed after 30 min
at this temperature (vide infra). The band at 2132 cm−1 was
unchanged when switching to the 13C-feed; a red shift of ca.
50 cm−1 would be expected when replacing a 12CO(ads) with
13CO(ads). Such CO(ads) red shifts are easily observed on Pt-
based catalysts [8], which are known to bond CO even more
strongly than Au-based materials do. This observation stresses
that the 2132 cm−1 band observed here was not due to any car-
bonyl species, but simply to an electronic transition occurring
over reduced ceria, as discussed by Lavalley et al. [39].
The exchange of the carbonate and formate species appeared
to proceed at a similar rate (see the 12C-formate and 12C-
carbonate signals in Figs. 5 and 6). It is obvious from Figs. 5
and 6 that a fraction of the formates and carbonates exchanged
on a similar time scale to that of the CO2, whereas the re-
mainder exchanged much more slowly. Accordingly, the curves
associated with the exchange of the formate and carbonate sig-
nals could not be fitted to a simple exponential plot, as would
be expected for a first-order exchange of uniform species. As an
3.3. Formate, carbonate and CO2 signal exchange over
0.6AuCL
The kinetics of the exchange of the formate and carbonate
species and that of the reaction product CO2 were followed
using the DRIFTS spectra recorded at three different tempera-

F.C. Meunier et al. / Journal of Catalysis 247 (2007) 277–287
281
12
Fig. 4. Details of the formate and carbonate regions of the in situ DRIFT spectra obtained over of the 0.6AuCL under 2% CO + 7% H O in Ar (bottom spectrum)
2
13
and 30 min under 2% CO + 7% H O in Ar at various temperatures (three top spectra). A spectrum of each same sample under 7% H O /Ar at the same
2
2
temperature was used as background.
12
12
Fig. 6. Relative evolution of the intensity of C-containing carbonate band ("),
Fig. 5. Relative evolution of the intensity of C-containing carbonate band ("),
12
13
12
13
C-containing formate band (×) and CO (!) signal with time on stream at
C-containing formate band (×) and CO (!) signals with time on stream
2
2
13
12
13
12
493 K under 2% CO + 7% H O, following steady-state under 2% CO +
at 428 K under 2% CO + 7% H O, following steady-state under 2% CO
2
2
7% H O.
2
+ 7% H O.
2
mates (i.e., 29.9 kJ mol−1). The pre-exponential factors As and
Af represent the fraction of the slow and fast species (i.e.,
As + Af = 1), respectively. It can be noted that the proportion
of fast formate increased at the higher temperatures. Essentially
similar observations were made in the case of the carbonates,
but both slow and fast carbonates exhibited similar activation
energies of decomposition. Again, we see that the apparent
activation energy of 40.0 kJ mol−1 (see Section 4.1) for the for-
mation of CO2 (determined from the reaction rate values) is
significantly different from that of the decomposition of any of
the surface species reported above.
example, the formate exchange curves are reported in Fig. 8 us-
ing a semilogarithmic plot, which shows that the corresponding
lines clearly are not straight.
The sum of two exponential terms was sufficient to obtain a
satisfactory fit of the exchange signals for formates and carbon-
ates (not shown). The coefficients of the exponential terms (i.e.,
pre-exponential factor A and rate constant k) for the modeling
of the formate and carbonate signals are reported in Table 1.
Assuming the relevance of a two-species model, the parame-
ters were assigned the letters “s” for slow and “f” for fast with
respect to the slower- and faster-exchanging species related to
the lower and higher rate constant, respectively. The Arrhenius
plots for the slow and fast formates and carbonates rate con-
stants are shown in Fig. 9. The fast formates were exchanged
between 10 (at 493 K) and 20 times faster (at 428 K) than the
slower counterpart. The slower-exchanging formate species ex-
hibited a higher apparent activation energy of decomposition
(i.e., 60.8 kJ mol−1) compared with the faster-exchanging for-
3.4. Quantification of the formate concentration over 0.6AuCL
The quantification of the specific concentration of formate
species at the surface of the 0.6AuCL was determined via a cal-
ibration curve, which was obtained from standard samples pre-
pared by incipient wetness impregnation of the ceria–lanthana

282
F.C. Meunier et al. / Journal of Catalysis 247 (2007) 277–287
13
Fig. 7. Relative evolution of the CO signal with time on stream at three
2
13
different temperatures 428 K (+), 458 K ("), 493 K (!) under 2% CO + 7%
Fig. 9. Arrhenius plots of the rate constant related to the exchange of the fast
formates (ꢀ), the slow formates (Q), the fast carbonates (!) and the slow car-
bonates ("). The corresponding apparent activation energies are also reported.
12
H O, following steady-state in 2% CO + 7% H O. Insert: Arrhenius plot
2
2
and corresponding apparent activation energy of the rate constant related to the
exchange of CO (g).
2
formate decomposition to that of CO2 formation per mass of
catalyst. The specific rate of CO2 formation was simply deter-
mined from the accurate measurement of the CO2 concentration
in the DRIFTS cell exhaust gas by gas-chromatography, the
sample weight in the DRIFTS crucible and the flow rate of re-
actants used. The rate of formate decomposition was calculated
as the sum for the fast and slow formates of the corresponding
product of the rate constant for exchange and the concentration
(Table 1), that is, rate of formate decomposition = ks [slow
formates] + kf [fast formates], where [slow formates] = As
[formates] and [fast formates] = Af [formates].
Note that the decomposition rate of the slow formate species
was almost negligible compared with that of the fast formates.
The semilogarithmic plot reporting the rates in Fig. 11 shows
that the rate of formate decomposition was more than an order
of magnitude smaller (ca. 60-fold) than the rate of CO2 forma-
tion.
Fig. 8. Semilogarithmic plots of the relative evolution of the intensity of
12
C-containing formate band with time on stream at different temperatures un-
13
12
der 12% CO + 7% H O, following steady-state under 2% CO + 7% H O.
2
2
4. Discussion
(CL) support with a solution of sodium formate. The standard
samples did not contain any Au to prevent the decomposi-
tion of the adsorbed formates with time. The spectrum of the
standard loaded with 0.125 wt% formate on CL is shown in
Fig. 10a, along with an in situ spectrum obtained under WGS
conditions over the 0.6AuCL at 458 K. The calibration plot
was drawn by reporting the area of the 2830 cm−1 peak as a
function of the concentration of the standard (Fig. 10b). The
values of formate concentration at the three temperatures inves-
tigated are reported in Table 1, in 10−6 mol per gram of cat-
alyst. These values corresponded to a surface density of about
4.1. Ceria oxidation state and nature of the surface species
The combination of the in situ DRIFTS and the SSITKA
technique facilitated the assignment of the bands observed un-
der WGS conditions over the 0.6% Au/Ce(La)O2 catalyst. In
particular, the band observed here at 2132 cm−1 (located at
2125 cm−1 before baseline correction) appears to be solely due
to the normally forbidden electronic transition of reduced ce-
ria [39] and not to any CO(ads), because no band shift was
observed on CO(g) isotopic exchange (Fig. 3). Note that a band
at ca. 2130–2140 cm−1 was observed on the same sample from
120 to 373 K and assigned to CO adsorbed on Au+n clusters in
work reported by some of us elsewhere [10]. However, the ex-
perimental conditions used here are markedly different, that is,
significantly higher temperatures (428–493 K) and the presence
of 7% H2O. The bonding of CO on Au is typically weak, which
explains the fact that no carbonyl bands could be detected in the
1.2 × 10−7 molm−2
.
3.5. Comparison of the formate decomposition rate and the
rate of CO2 formation
The relevance of the formate seen by DRIFTS in the forma-
tion of CO2 must be evaluated by comparing the molar rate of

F.C. Meunier et al. / Journal of Catalysis 247 (2007) 277–287
283
(a)
(b)
Fig. 10. (a) Upper spectrum: DRIFT spectrum observed over a Au-free Ce(La)O material at 373 K in Ar impregnated with 0.125 wt% of formate deposited from
2
13
sodium formate. Lower spectrum: in situ DRIFT spectrum observed over the 0.6AuCL under 2% CO + 7% H O in Ar. T = 458 K. (b) Calibration plot relating
2
the formate CH stretching band area of the Ce(La)O sample impregnated with various loading of sodium formates. The dotted line is associated with the in situ
2
spectrum shown in (a).
Table 1
Parameters representing the two-term exponential model of the formate and carbonate exchange curves
T = 428 K
T = 458 K
T = 493 K
App. activation
energy (kJ mol
−1
)
a
CO conversion
Formate concentration in 10 mol g
7.98%
21.5
14.8%
24.1
25.2%
19.2
40.0
−6
−1
CO exchange
2
k in min
−1
−1
−1
−1
2.5 × 10
4.0 × 10
6.0 × 10
23.7
Formate exchange
A
f
0.382
0.536
0.690
−1
−1
−1
−1
k
f
in min
1.36 × 10
2.20 × 10
4.11 × 10
29.9
60.8
A
s
0.617
0.454
0.314
−1
−3
−2
−2
k in min
s
5.85 × 10
1.73 × 10
5.54 × 10
Carbonate exchange
A
s
0.591
0.504
0.296
−1
−3
−1
−3
−1
−2
−1
k in min
s
6.81 × 10
7.62 × 10
1.65 × 10
24.0
20.4
A
f
0.413
0.499
0.695
−1
k
f
in min
1.81 × 10
2.83 × 10
3.84 × 10
−k xt
−k xt
. The CO exchange plot was fitted with a single exponential function.
2
f
s
Signal form = A e
+ A e
s
f
a
Determined from data measured in differential conditions, i.e. conversion <15% (Fig. 1).
range of higher temperatures used here. The corresponding low
surface coverage of Au by CO suggests that CO adsorption may
be the rate-determining step of the WGS reaction. As a matter
of fact, a CO reaction order of 1 is often observed for Au-based
WGS catalysts [40]. This is in contrast with Pt/CeO2 catalysts,
for which a strong carbonyl band is observed even around 500 K
[30] and the CO reaction order is typically 0 [5,15].
The bands associated with formate and carbonate species
were broad and ill-defined (Figs. 2–4), as would be expected
over a high-surface area nanocrystalline sample. Nonetheless,
our DRIFTS–SSITKA experiments were able to identify a
nonuniform reactivity of the surface formates and carbonates,
because neither the formate (Fig. 8) nor the carbonate (not
shown) exchange curve fit a simple exponential decay. Note
that first-order exchanges, which are associated with single ex-
ponential laws, have been observed for formates under similar
reaction conditions over Pt/CeO2 [29] and Pt/ZrO2 [41]. The
fact that a sum of two exponential terms could represent the
exchange of formates and carbonates suggests that two main
types of reactivity were present. This observation is crucial,
because formate and carbonate species in the WGS reaction
are usually considered to exhibit the same reactivity. Davis et
al. [27] recently reported two different types of formates over
Pt/Ce(Zr)O2 materials relating to Ce-rich or Zr-rich regions, but
it was not clear whether these species displayed the same reac-
tivity.
As mentioned by one of the referees, an apparent twofold
increase in formate reactivity also could have arisen from a
variation of the molar absorption coefficient (absorptivity) with
surface coverage. But we believe that this was not the case
here, for the following reasons. First, the proportion of slow
formate species varied quite significantly (between 60% and

284
F.C. Meunier et al. / Journal of Catalysis 247 (2007) 277–287
Table 2
Density of CO precursors as measured by SSITKA over 0.6AuCL
2
2
−1
)
(161 m g
T = 428 K
T = 458 K
T = 493 K
N
= number of CO
3.67
4.00
5.49
CO
2
−1
2
−4
precursors (10 mol g
Surface O/N
)
9.9
154
17
9.1
168
18
6.6
231
29
CO
2
N
2 /Au
2 /formate
CO
CO
N
Au was assumed to be 100% dispersed, leading to a surface O/Au ratio of ca.
1500. CeO has a cubic cell with four oxygen atoms on each face, the surface
2
−19
2
area of which is ca. 2.92 × 10
m . The theoretical number of moles of
2
−6
oxygen atoms, or oxide ions, per m of catalyst surface is 22.7 × 10
.
Fig. 11. Rate of CO production and rate of formate decomposition over the
2
12
0.6AuCL at three different temperatures under 2% CO + 7% H O.
2
30%) over the temperature range investigated, whereas the total
formate concentration remained essentially constant (Table 1).
A change of absorptivity with coverage cannot explain these
observations. In other words, a variation of absorptivity would
have to be considered only if the proportion of slow formates
was the same at all temperatures, which could then be possibly
explained by a coverage-dependent absorptivity. Second, the
experiments were carried at constant coverage, with only the
isotopic nature of the adsorbates varied. Third, in the present
case, we can assume that the distance between formates was
sufficiently large so as to limit any lateral interactions. This is
supported by the fact that the number of surface formates is
<1% of the number of surface oxide ions (see Table 2; mul-
tiply the values in the last and antepenultimate lines). Fourth,
the decay of the formate signal over similar catalysts during
isotopic exchange [29,41] or during an Ar purge (data to be
published) showed that a perfect single first-order decay was
obtained, demonstrating that the absorptivity of formates was
not coverage-dependent in these cases. Finally, the calibration
plot given in Fig. 10b clearly shows that the DRIFTS signal of
formates was proportional to formate surface concentration. In
view of all of these arguments, it is unlikely that the absorptivity
of formates was coverage-dependent in the present case.
Fig. 12. Schematic representation of the species populating the surface of the
0.6AuCL catalyst during the water–gas shift reaction. It is suggested that the
surface species close to the Au centers are the most reactive, while those fur-
ther away are essentially spectators. The origin of the difference of reactivity is
unclear.
constant can be explained by a variable extent of adsorbate
spillover with temperature [42]. Second, the promoting effect of
the Au could spread to an increasing distance with higher tem-
peratures. This could be due to an enhanced “normal-support
activation” electronic effect [43], to an increased extent of the
“active” reduced ceria area around the Au particles as proposed
for Pt/CeO2 [25,44], or to an increased capture zone of slowly
diffusing surface species, as proposed on Au/CeO2 [28]. Third,
and somewhat similar to the aforementioned point, the mobil-
ity of surface Au atoms should increase with temperature, ex-
plaining the larger fraction of reactive intermediates. A scheme
summarizing these various possibilities is given in Fig. 12.
It is interesting to note that the 0.6AuCL catalyst used here
has the characteristic CO temperature reduction profile reported
before for these Au–O–Ce surfaces (devoid of gold particles
before reaction) [7], with two resolved surface oxygen reduc-
tion peaks, noted as Os1 and Os2. The Os1 species was re-
duced below 473 K and thus would be expected to be partly
4.2. Origin of the nonuniform reactivity of formates and
carbonates
The twofold or greater increase in reactivity of surface inter-
mediates could have various origins, which we did not explore
in the present study. First, the sites of different reactivities could
be related to Ce or La cations, or even to surface patches of
the segregated oxides. The fact that the ratio of fast to slow
species varied widely over the temperature range used (see the
values of Af and As in Table 1) while the fraction of La close to
the surface (ca. 20 at%) was expected to remain approximately

F.C. Meunier et al. / Journal of Catalysis 247 (2007) 277–287
285
Fig. 13. Suggested reaction scheme taking place over the 0.6AuCL catalysts during the water–gas shift reaction. The intermediate X could possibly be a species
formed by reaction between CO adsorbed on Au and an oxide ion, as in a redox mechanism, or an adduct species, possibly a formate not resolved on the DRIFT
spectrum.
reduced under our reaction conditions (2% CO + 7% H2O),
in agreement with the fact that ceria was partly reduced dur-
ing our investigation. Following an initial report by Löffler et
al. [9], ceria surface reduction and the formation of Ce(III) car-
bonates has a widely proposed cause of catalyst deactivation
[11,29,45,46] during the WGS reaction over ceria-based cata-
lysts. Thus, it is possible that the most stable adsorbate (i.e.,
so-called “slow” species) would be that adsorbed on Os1 sites.
This stable formate would then become more reactive as the
temperature increases and water activation and O-diffusion fa-
cilitate the reoxidation of Os1 sites. More work would be re-
quired to probe the various possibilities proposed in this section
and determine the true origin of the nonuniform reactivity ob-
served.
overestimation. The number of CO2 precursors increased with
temperature and was always two orders of magnitude higher
than the number of surface Au atoms (assuming 100% disper-
sion). Such a large number of precursors and the well-known
facts that ceria is a strongly basic oxide and CO2 is an acidic
molecule raise the concern of extensive CO2 readsorption dur-
ing the course of the reaction.
Typical CO2 desorption rate constant values over similar ba-
sic materials measured during 12CO2–13CO2 exchanges (with
no other reactive gas present) have been reported [56,57] and
were of a similar order of magnitude to that of the exchange
rate constant k determined here, that is, 0.1–1 min−1 (Table 1).
The apparent activation energy for CO2 formation (determined
from the reaction rate values) was equal to 40.0 kJ mol−1
,
It must be noted that the level of reduction of ceria under
our less-reducing conditions must be limited, although it is ob-
servable. Mogensen et al. [47] have reported equilibrium plots
describing the extent of ceria reduction as a function of the
partial pressure of O2. Based on these plots, we estimate that
bulk CeO2 should be reduced to CeO1.9 under our conditions,
for which the corresponding theoretical O2 pressure was ca.
10−42 atm. WGS catalyst deactivation due to Ce(III) carbon-
ate formation is established in the literature [11,20]. As long
as CeO2 is completely oxidized, it does not adsorb CO2 [48].
The presence of Ce3+ in catalysts after the WGS reaction was
identified by XRD [11] and by XPS [49] even though the Ce3+
concentration may be overestimated by XPS [50–52]. The oxy-
gen defect concentration increases with addition of trivalent
La. Near the surface, the La (19.1 at%) content was more than
twofold that in the bulk of ceria; the same is true for the corre-
sponding concentration of oxygen vacancies. In the presence of
Au, the surface reduction of ceria is much easier as evidenced
by TPR data [18]. All surface oxygen of ceria was reduced
below 473 K. Reactivation of ceria requires heating in air at
T > 350◦C to decompose the carbonate [11].
significantly greater than that of the exchange of CO2 (i.e.,
23.7 kJ mol−1). The apparent activation energy of CO2 ex-
change (23.7 kJ mol−1) was actually similar to that of carbon-
ate exchange (slow carbonate, 24.0 kJ mol−1; fast carbonates,
20.4 kJ mol−1). These observations strongly support the hy-
pothesis that a significant fraction of CO2 readsorbed on the
catalyst as carbonates, and that the CO2(g) exchange was dom-
inated by this “unreactive” adsorption/desorption phenomenon
and masked the process of conversion of CO to CO2. A scheme
summarizing the different events occurring during the reaction
is given in Fig. 13.
Although fast CO2 readsorption/desorption occurred, in-
volving a greater proportion of the surface at higher temper-
atures, it is still worth pointing out that a significant fraction
of carbonates did not exchange over durations much greater
than the time needed to complete CO2 exchange, stressing the
complex nature of the sample reactivity. A full analysis of the
SSITKA data [53] gathered over such samples that exhibit a
large concentration of adsorption sites and sites with varying
reactivity is too complex to yield any unambiguous conclusions
and thus was not attempted. Note that a more detailed SSITKA
analysis was possible in the case of a low-basicity Pt/ZrO2 ma-
terial, for which the amount of CO2 precursors was found to
be even less than the number of surface Pt atoms [41]. In the
present case, the number of CO2 precursors was more than an
order of magnitude higher than that of surface formates (Ta-
ble 2). Because CO2 readsorption occurred, such a high ratio
4.3. CO2 precursors and readsorption effects
The number of reaction product precursors of CO2 mea-
sured by the SSITKA method [53–55] over the 0.6AuCL are
given in Table 2. It should be stressed that these numbers are
by no means to be taken as the number of active sites, because
product readsorption can occur, leading to active site number

286
F.C. Meunier et al. / Journal of Catalysis 247 (2007) 277–287
cannot serve as evidence to rule in or rule out formates as sur-
face intermediates.
the U.S. Department of Energy, Basic Energy Sciences, Hydro-
gen Fuel Initiative Grant #DE-FG02-05ER15730.
References
4.4. Role of the formates detected by DRIFTS in CO2
production
[1] T. Tabakova, F. Boccuzzi, M. Manzoli, J.W. Sobczak, V. Idakiev, D. An-
dreeva, Appl. Catal. B 49 (2004) 73.
[2] T. Tabakova, F. Boccuzzi, M. Manzoli, J.W. Sobczak, V. Idakiev, D. An-
dreeva, Appl. Catal. A 298 (2006) 127.
[3] T. Shido, Y. Iwasawa, J. Catal. 136 (1992) 493.
[4] G. Jacobs, L. Williams, U. Graham, D. Sparks, B.H. Davis, J. Phys. Chem.
B 107 (2003) 10398.
[5] G. Jacobs, L. Williams, U. Graham, D. Sparks, G. Thomas, B.H. Davis,
Appl. Catal. A 252 (2003) 107.
[6] G. Jacobs, E. Chenu, P. M Patterson, L. Williams, D. Sparks, G. Thomas,
B.H. Davis, Appl. Catal. A 258 (2004) 203.
[7] Q. Fu, H. Saltsburg, M. Flytzani-Stephanopoulos, Science 301 (2003) 935.
[8] A. Goguet, D. Tibiletti, F.C. Meunier, J.P. Breen, R. Burch, J. Phys. Chem.
B 108 (2004) 20240.
[9] J.M. Zalc, V. Sokolvskii, D.G. Löffler, J. Catal. 206 (2002) 169.
[10] M. Manzoli, F. Boccuzzi, A. Chiorino, F. Vindigni, W. Deng, M. Flytzani-
Stephanopoulos, J. Catal. 245 (2007) 306.
[11] W. Deng, M. Flytzani-Stephanopoulos, Angew. Chem. Int. Ed. 45 (2006)
2285.
[12] C.M.Y. Yeung, F. Meunier, R. Burch, D. Thompsett, S.C. Tsang, J. Phys.
Chem. B 110 (2006) 8540.
[13] Y. Amenomiya, Appl. Spectrosc. 32 (1978) 484.
[14] Y. Li, Q. Fu, M. Flytzani-Stephanopoulos, Appl. Catal. B 27 (2000) 179.
[15] T. Bunluesin, R.J. Gorte, G.W. Graham, Appl. Catal. B 15 (1998) 107.
[16] S. Hilaire, X. Wang, T. Luo, R.J. Gorte, J. Wagner, Appl. Catal. A 215
(2001) 271.
[17] X. Wang, R.J. Gorte, Catal. Lett. 73 (2001) 15.
[18] Q. Fu, A. Weber, M. Flytzani-Stephanopoulos, Catal. Lett. 77 (2001) 87.
[19] T. Luo, R.J. Gorte, Catal. Lett. 85 (2003) 139.
[20] X. Wang, R.J. Gorte, J.P. Wagner, J. Catal. 212 (2002) 225.
[21] R. Burch, Phys. Chem. Chem. Phys. 8 (2006) 5483.
[22] T. Shido, Y. Iwasawa, J. Catal. 141 (1993) 71.
[23] G. Jacobs, S. Ricote, U.M. Graham, P.M. Patterson, B.H. Davis, Catal.
Today 106 (2005) 259.
[24] G. Jacobs, B. Davis, Appl. Catal. A 285 (2005) 43.
[25] G. Jacobs, U.M. Graham, E. Chenu, P.M. Patterson, A. Dozier, B.H.
Davis, J. Catal. 229 (2005) 499.
[26] G. Jacobs, B.H. Davis, Appl. Catal. A 284 (2005) 31;
G. Jacobs, B.H. Davis, Appl. Catal. A 284 (2005) 499.
[27] S. Ricote, G. Jacobs, M. Milling, Y. Ji, P.M. Patterson, B.H. Davis, Appl.
Catal. A 303 (2006) 35.
Whereas CO2 readsorption hindered the analysis of the CO2-
exchange signals, the observed exchange rate of formates and
the corresponding steady-state surface coverage allowed us to
calculate the specific rate of formate decomposition, which
turned out to be an order of magnitude lower than the rate
of CO2 formation (Fig. 11). These data unambiguously prove
that the formats detected by DRIFTS are not main reaction
intermediates in the formation of CO2 during the WGS over
this 0.6AuCL catalyst under our reaction conditions. These for-
mates (slow + fast) should be labeled “minor reaction interme-
diates,” not “spectators,” because they still led to some reaction
product, although at an almost negligible rate compared with
that of the main, yet unknown route. As indicated by Behm
et al. [28], these formates could diffuse and react at the Au–
support interface, being a kind of “buffer” species storing CO
at the surface of the sample. Nonetheless, it should be stressed
that this does not prove by any means that the main reactive in-
termediate leading to the formation of CO2 is a formate. The
main WGS pathway remains unclear (Fig. 13), for example,
redox route or interface adduct (not excluding formates) of a
concentration too low to be observable by in situ DRIFTS.
5. Conclusion
The following conclusions can be drawn from the results of
our in situ DRIFTS–SSITKA study of the WGS reaction over a
0.6% Au/Ce(La)O2 catalyst between 428 and 493 K:
1. The importance of fully quantifying not only the rate of ex-
change, but also the absolute amounts of surface species
has been established, showing that otherwise very mislead-
ing conclusions about possible reaction intermediates can
be drawn.
2. The specific rate of CO2 formation was ca. 60 times
higher than the rate of formate decomposition (fast + slow
species) decomposition, showing that the formates detected
by DRIFTS are not important in the main reaction pathway.
3. Nonuniform reactivity of the surface formates and carbon-
ates was observed, which could be fitted to a bimodal dis-
tribution relating to slow- and fast-exchanging species.
4. A significant readsorption of the reaction product CO2 oc-
curred, which disguised the exchange curves and the num-
ber of precursors as determined by the SSITKA method.
[28] R. Leppelt, B. Schumacher, V. Plzak, M. Kinne, R.J. Behm, J. Catal. 244
(2006) 137.
[29] F.C. Meunier, D. Tibiletti, A. Goguet, S. Shekhtman, C. Hardacre,
R. Burch, Catal. Today (2007), doi:10.1016/j.cattod.2006.10.003, in press.
[30] D. Tibiletti, A. Goguet, F.C. Meunier, J.P. Breen, R. Burch, Chem. Com-
mun. (2004) 1636.
[31] M.W. Balakos, S.S.C. Chuang, G. Srivinas, J. Catal. 140 (1993) 281.
[32] R.W. Stevens, S.S.C. Chuang, J. Phys. Chem. B 108 (2004) 696.
[33] J.M. Olinger, P.R. Griffiths, Anal. Chem. 60 (1988) 2427.
[34] C. Binet, M. Daturi, J.C. Lavalley, Catal. Today 50 (1999) 207.
[35] M. Daturi, C. Binet, J.C. Lavalley, G. Blanchard, Surf. Interface Anal. 30
(2000) 273.
[36] C. Li, Y. Sakata, T. Arai, K. Domen, K.-I. Maruya, T. Onishi, J. Chem.
Soc. Faraday Trans. 85 (1989) 1451.
Acknowledgments
[37] P. Bazin, O. Saur, J.C. Lavalley, M. Daturi, G. Blanchard, Phys. Chem.
Chem. Phys. 7 (2005) 187.
[38] G. Busca, J. Lamotte, J.C. Lavalley, V. Lorenzelli, J. Am. Chem. Soc. 109
(1987) 5197.
[39] C. Binet, A. Badri, J.C. Lavalley, J. Phys. Chem. 98 (1994) 6392.
[40] G. Jacobs, P.M. Patterson, L. Williams, E. Chenu, D. Sparks, G. Thomas,
B.H. Davis, Appl. Catal. A 285 (2005) 43.
This work was partly supported by the EPSRC, under the
CARMAC project and the European Social Funds (D.R.). W.D.
and M.F.-S. acknowledge the support of this work by the U.S.
National Science Foundation, NIRT Grant #0304515; and by

F.C. Meunier et al. / Journal of Catalysis 247 (2007) 277–287
287
[41] D. Tibiletti, F.C. Meunier, A. Goguet, D. Reid, R. Burch, M. Boaro,
M. Vicario, A. Trovarelli, J. Catal. 244 (2006) 183.
[42] M. Daturi, O. Marie, personal communication, Caen, 2006.
[43] J.C. Frost, Nature 334 (1988) 577.
[49] Q. Fu, Ph.D. thesis, Department of Chemical and Biological Engineering,
Tufts University, 2004.
[50] E. Paparazzo, G.M. Ingo, N. Zacchetti, J. Vac. Sci. Technol. A 9 (3) (1991)
1416.
[44] A. Goguet, F.C. Meunier, J.P. Breen, R. Burch, M.I. Petch, A. Faur Ghen-
ciu, J. Catal. 226 (2004) 382.
[51] M.V. Rama Rao, T. Shripathi, J. Electron. Spectrosc. Relat. Phenom. 87
(1997) 121.
[45] D. Tibiletti, A. Goguet, D. Reid, F.C. Meunier, R. Burch, Catal. Today 113
(2006) 94.
[52] F. Zhang, P. Wang, J. Koberstein, S. Khalid, S.-W. Chan, Surf. Sci. 563
(2004) 74.
[46] Q. Fu, W. Deng, H. Saltsburg, M. Flytzani-Stephanopoulos, Appl. Catal.
B 56 (2005) 57.
[47] M. Mogensen, N.M. Sammes, G.A. Tompsett, Solid State Ionics 129
(2000) 63.
[48] S. Sharma, S. Hilaire, J.M. Vohs, R.J. Gorte, H.-W. Jen, J. Catal. 190
(2000) 199.
[53] S.L. Shannon, J.G. Goodwin, Chem. Rev. 95 (1995) 677.
[54] C.N. Costa, A.M. Efstathiou, J. Phys. Chem. B 108 (2004) 2620.
[55] Y. Schuurman, C. Mirodatos, Appl. Catal. A 151 (1997) 305.
[56] M. Xu, E. Iglesia, J. Phys. Chem. B 102 (1998) 961.
[57] F.C. Meunier, D. Tibiletti, A. Goguet, D. Reid, R. Burch, Appl. Catal.
A 289 (2005) 104.