TAP study on carbon monoxide oxidation over supported gold catalysts Au/Ti(OH)*4 and Au/Fe(OH)*3: Moisture effect

Article abstract of DOI:10.1016/j.jcat.2007.03.004

The effect of moisture on the behavior of the active supported gold catalysts developed by us, Au/Ti(OH)^*₄ (I) and Au/Fe(OH)^*₃ (II), during CO activation and CO oxidation in the temperature range 298-473 K was investigated by the temporal analysis of products (TAP) technique. The oxidation reaction over the Au/Ti(OH)^*₄ catalyst was totally suppressed by the presence of water, independent of the water content of 200 and 1000 ppm, the carbon monoxide-to-oxygen ratio, and the temperature. In contrast, the presence of water vapors showed no effect on CO oxidation over the Au/Fe(OH)^*₃ catalyst. For both catalysts, the carbon monoxide adsorption feature was unalterable by the presence of water, namely CO molecules reversibly adsorbed on the catalyst surface, but the amount of CO adsorbed decreased by increasing water content. When a reaction mixture at a 1:1 ratio of CO to O₂ was pulsed on the catalyst in the absence of water, the CO response curve showed two peaks, whereas in the presence of water, the TAP response curve showed only one peak. At a 1:3 CO-to-O₂ ratio, the CO response curve showed only one peak, which shifted to a shorter time and narrowed in the absence of water. The shift was more pronounced on catalyst (II). The TAP experiments provided indirect evidence that CO and water adsorbed together (CO weakly and water strongly) and water blocked [irreversibly for Au/Ti(OH)^*₄ and reversibly for Au/Fe(OH)^*₃], the adsorption sites for oxygen.

Full text of DOI:10.1016/j.jcat.2007.03.004

Journal of Catalysis 248 (2007) 60–67
www.elsevier.com/locate/jcat
TAP study on carbon monoxide oxidation over supported gold catalysts
Au/Ti(OH)∗4 and Au/Fe(OH)*3: Moisture effect
Maria Olea ∗, Mizuki Tada, Yasuhiro Iwasawa
Department of Chemistry, Graduate School of Science, The University of Tokyo, Hongo, Bunkyo-ku, Tokyo 113-0033, Japan
Received 27 December 2006; revised 6 March 2007; accepted 6 March 2007
Abstract
The effect of moisture on the behavior of the active supported gold catalysts developed by us, Au/Ti(OH) (I) and Au/Fe(OH) (II), during
∗
∗
3
4
CO activation and CO oxidation in the temperature range 298–473 K was investigated by the temporal analysis of products (TAP) technique. The
∗
oxidation reaction over the Au/Ti(OH) catalyst was totally suppressed by the presence of water, independent of the water content of 200 and
4
1000 ppm, the carbon monoxide-to-oxygen ratio, and the temperature. In contrast, the presence of water vapors showed no effect on CO oxidation
∗
over the Au/Fe(OH) catalyst. For both catalysts, the carbon monoxide adsorption feature was unalterable by the presence of water, namely CO
3
molecules reversibly adsorbed on the catalyst surface, but the amount of CO adsorbed decreased by increasing water content. When a reaction
mixture at a 1:1 ratio of CO to O was pulsed on the catalyst in the absence of water, the CO response curve showed two peaks, whereas in
2
the presence of water, the TAP response curve showed only one peak. At a 1:3 CO-to-O ratio, the CO response curve showed only one peak,
2
which shifted to a shorter time and narrowed in the absence of water. The shift was more pronounced on catalyst (II). The TAP experiments
∗
provided indirect evidence that CO and water adsorbed together (CO weakly and water strongly) and water blocked [irreversibly for Au/Ti(OH)
4
∗
and reversibly for Au/Fe(OH) ], the adsorption sites for oxygen.
3
© 2007 Elsevier Inc. All rights reserved.
Keywords: Supported gold catalyst; Carbon monoxide oxidation; Reaction mechanism; Moisture effect; TAP study
1. Introduction
[2,3], the nature of the support [23,24], the methods of prepara-
tion and pretreatment [25], and the operating environment have
been proposed to be critical to the high activity. Despite previ-
ous efforts, however, there are still some uncertainty and debate
concerning the precise mechanism of the CO oxidation over
these catalysts and the effect of moisture on the catalytic per-
formance of gold.
To date, there have been several reports, sometimes contra-
dictory, about the effect of moisture on supported gold catalysts.
Daté and Haruta [26] reported that moisture enhanced the reac-
tion by more than 10 times up to 200 ppm H2O, and further
increases in moisture content suppressed the reaction over an
Au/TiO2 catalyst obtained by deposition–precipitation. At low
moisture content, the positive effect was attributed to a change
in the amount of water-derived species, which may activate O2
molecules or modify the electronic state of gold atoms exposed
at the surface. At high moisture content, the suppression of cat-
alytic activity was purportedly due to the adsorption of water
Supported gold nanoparticle catalysts have received intense
and broad interests since the report of Harada et al. [1] of their
high catalytic activity for low-temperature CO oxidation. In-
creasing efforts in both catalytic and industrial chemistry [2–
12] and cluster and theoretical science [13–15] have been de-
voted to clarifying the origins of the unique catalytic properties.
These properties have been related to changes in the electronic
properties, the presence of defect sites, and the existence of
strain for metallic gold nanoparticles [16–20]. Unique catalytic
properties also have been related to the presence of sites associ-
ated with the catalyst support, such as cationic gold species, and
sites at gold–support interfaces [21,22]. The size of Au particles
*
Corresponding author. Present address: Laboratorium voor Petrochemische
Techniek, Ghent University, Krijgslaan 281 S5, 9000 Ghent, Belgium. Fax:
+32 9 264 4999.
E-mail address: maria.olea@ugent.be (M. Olea).
0021-9517/$ – see front matter © 2007 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2007.03.004

M. Olea et al. / Journal of Catalysis 248 (2007) 60–67
61
on oxygen vacancies, which were then blocked for further ad-
sorption of O2.
monoxide oxidation mechanism on the catalysts previously re-
ported by Iwasawa’s group [35,37].
The main goal of the present paper is to reveal those de-
tails of the CO adsorption mechanism and features that provide
insight into the influence of water on CO oxidation activity
over Au/Ti(OH)∗4 (I) and Au/Fe(OH)∗3 (II) catalysts. For this
purpose, single-pulse TAP experiments were performed to in-
vestigate the diffusion of Ar and He and the adsorption of CO
(dry and moisturized) and O2. Pump-probe (or alternating) TAP
experiments were conducted to explore the behavior of the fresh
or oxygen precovered catalysts. The O2 pump molecule was
probed with both dry and moisturized CO at different time in-
tervals to provide information about the reactive surface species
and individual reaction steps. All TAP experiments in the pres-
ence of water were performed at the same temperatures as pre-
viously done in the absence of water: 298, 373, and 473 K. This
allowed a direct comparison of the data and a clear assessment
of the effect of water.
Augmentation of the oxidation activity in the presence of
water was reported for Au/Fe2O3 [27,28], Au/Al2O3 [28,29],
and Au/Mg(OH)2 [30] catalysts obtained by a deposition–
precipitation method. Bond et al. [31] proposed a possible
reaction mechanism for CO oxidation that involves hydroxyl
species. On the other hand, enhancement of activity by the ad-
sorbed water species was explained by the prevention of the
reduction of cationic gold species for those catalysts with ac-
tive sites consisting of cationic Au+ with a hydroxyl ligand and
neighboring metallic Au atoms [29,32].
For an impregnated Au/TiO2 catalyst, the presence of water
was found to suppress the catalytic activity [33]. The deactiva-
tion of the impregnated Au/TiO2 catalyst by water vapor may
be due to H2O adsorption on TiO2, because water is known to
adsorb both associatively and dissociatively (as OH−) on re-
duced Ti cations and to oxidize the surface [34], thus blocking
the coordinatively unsaturated sites at the TiO2 surface.
Iwasawa and coworkers developed iron and titanium oxide-
supported gold catalysts, Au/Fe(OH)∗3 [35,36] and Au/Ti(OH)∗4
[37], derived from Au–phosphine complex precursor for Au
particles and as-precipitated M(OH)∗x precursors for metal ox-
ide supports, that were very active for CO oxidation, and re-
ported that coexisting water vapor showed no effect and a nega-
tive effect, respectively. Based on FT-IR spectra of CO adsorbed
on Au/Fe(OH)∗ in the presence of various H2O pressures [35],
the insensitivit3y of this catalyst to water vapor was linked to
the presence of water, which greatly inhibited CO adsorption
on Fe3+ and dissociatively adsorbed on oxygen vacancies of
the Fe-oxide surface and reacted with CO to form CO2 by the
water-gas shift reaction. Thus, the oxygen vacancies were re-
covered to be available for O2 adsorption. To explain the differ-
ent behavior of the Au/Ti(OH)∗4 catalyst, additional experiments
were envisaged [35]. The key step seems to be CO adsorption.
A careful analysis of literature data leads to the conclu-
sion that the effect of water on the activity of supported gold
catalysts depends on the method of preparation, the nature of
the support, and the Au particle size, dispersion, and oxidation
state.
In the present study, the TAP technique was used to further
investigate the different effect of water on the CO oxidation
over the two catalysts, Au/Ti(OH)∗4 (I) and Au/Fe(OH)3∗ (II).
TAP is a technique sensitive to each step in a heterogeneous
catalytic process. We previously reported TAP studies [38,39]
on the diffusion and adsorption of carbon monoxide, oxygen,
and carbon dioxide adsorption/desorption, as well as catalytic
carbon monoxide oxidation over the titania-supported gold cat-
alyst (I), with all studies performed in the absence of mois-
ture. We found that carbon monoxide molecules reversibly
and weakly adsorbed on the catalyst surface; oxygen adsorbed
molecularly and strongly for high pulse intensities, with life-
times of 4500–6900 ms at 373–423 K; reaction occurred be-
tween adsorbed carbon monoxide and adsorbed molecular oxy-
gen; and lattice oxygen atoms were active in oxygen exchange
only with carbon dioxide. TAP results confirmed the carbon
2. Experimental
All experiments were conducted using a TAP reactor sys-
tem as described in detail elsewhere [38,39]. Briefly, the setup
consisted of a catalytic microreactor, a gas delivery system
for introduction of either high-speed gas pulses or a contin-
uous flow of gas, a high throughput, a high-vacuum system,
and a computer-controlled quadrupole mass spectrometer. The
microreactor was a stainless steel tube 2 mm in diameter and
25 mm long. Catalyst samples of 0.024 g were used in these ex-
periments, resulting in catalyst bed length of 10 mm. Through-
out the study, the 250–425 µm fraction was loaded and fixed in
the reactor by stainless steel mesh. The reactor could be heated
up to 773 K by cartilage heaters. The two high-speed pulse
valves could be operated at intervals of up to 50 pulses s−1. The
minimum pulse width was 200 µs. The pulse intensities were
varied between 1013 and 1018 molecules per pulse, allowing the
transport in the reactor to occur through Knudsen diffusion. Be-
fore each experiment, the catalyst was heated up to 473 K for
30 min in vacuum to desorb preadsorbed species. To increase
the signal-to-noise ratio in the response, 25 pulses per atomic
mass unit (AMU) value were monitored and averaged. Alter-
nating and simultaneous pulsing experiments were carried out
using 1:1 He/CO (dry or moisturized) mixture in one valve and
1:1 He/O2 mixture in the second valve. For dry experiments, the
two feeding mixtures were purified by being passed through a
silica trap at a dry ice-acetone temperature. The concentrations
of moisture in CO gas phase were 200 and 1000 ppm, and the
resulting mixtures were designated COm2 and COm10 to differ-
entiate them from the dry CO. These two particular moisture
values were chosen because they were found to be threshold
amounts for the enhancement and suppression, respectively,
of catalytic activity over an Au/TiO2 catalyst obtained by a
deposition–precipitation method [26].
For each experiment, the conversion was calculated accord-
ing to the following equation:
nin,i − nout,i
Xi =
,
(1)
nin,i

62
M. Olea et al. / Journal of Catalysis 248 (2007) 60–67
where n is the total number of mol and X is the conversion for
component i, CO or O2 (mol mol−1). The subscript “in” means
introduced into the reactor, and the subscript “out” means de-
tected at the reactor outlet. Using He as an internal standard
allows a precise determination of the number of molecules at
the reactor outlet. No influence on catalytic activity was found
when the catalysts were kept for one night under vacuum at
the highest temperature used during experiments. Because the
water–gas shift reaction could occur on these catalysts, hydro-
gen production was assessed by monitoring the changes in the
background spectra at AMU of 2 before and after the CO oxida-
tion reaction. Virtually no differences at AMU of 2 were found
before and after the oxidation reaction.
The catalysts were prepared as described in detail elsewhere
[35–37,40]. As-precipitated wet metal hydroxides, M(OH)∗x,
were obtained by hydrolysis of M(OiC3H7)x (99.999% purity)
with an aqueous solution containing 5% of Na2CO3 (99.9%
purity). The precipitates were then filtrated and washed with
deionized water to pH 7.0. The as-precipitated wet M(OH)x∗
was impregnated with an acetone solution of Au(PPh3)(NO3)
under vigorous stirring, followed by vacuum drying to remove
the solvent at room temperature. The sample thus obtained
was heated to 673 K at a ramp rate of 4 K min−1 and kept
at this temperature for 4 h in a flow of air (30 ml min−1).
Calcination of the samples led to decomposition of both pre-
cursors, Au(PPh3)(NO3) and M(OH)∗x, to metallic Au par-
ticles and TiO2 and Fe2O3, respectively. EXAFS measure-
ments in a transmission mode carried out at Au L3-edge be-
fore and after calcination indicated the formation of Au par-
ticles. The average Au particle size estimated by transmis-
sion electron microscopy was 2.9 nm for Au/Fe(OH)∗3 and
6.6 nm for Au/Ti(OH)∗4. The change in the Au species from
the phosphine complex to metallic particles was accompanied
by the development of the XRD peaks for crystalline TiO2
and Fe2O3 [40–42]. The catalysts thus obtained are denoted
as Au/M(OH)∗x where the term M(OH)∗x is used only to clar-
ify the origin of metal oxide, discriminating the preparation of
the hydroxide-based gold catalysts from that of conventional
Au/TiO2 and Au/Fe2O3 catalysts obtained by an incipient wet-
ness impregnation method. The Au loading on both supports
was controlled at 3.0 wt% [41]. The use of as-precipitated wet
M(OH)∗x as precursors was decisive in obtaining the highly ac-
tive catalyst [35–37,40–42].
intensity window for Knudsen diffusion only. Following the
procedure described previously [38], it was found that for pulse
intensities below 1 × 1018 molecules/pulse, the gas transport
through the catalyst bed fitted perfectly the Knudsen diffusion
flow for nonporous catalysts. The average values for He and Ar
effective Knudsen diffusivities were 0.443 and 0.140 cm2 s−1
,
respectively, considering an average porosity of bed of 0.23.
The effective Knudsen diffusivities for CO and O2 reactant
gases against He were 0.167 and 0.156 cm2 s−1, respectively.
The low content of moisture in the CO gas phase induced no
significant difference in the effective diffusivity of CO.
We performed experiments with the two inert gases, which
had significantly different molecular masses, to assess the intra-
particle Knudsen diffusion as if it would have been the case. But
this in fact was not the case, and, because our catalyst behaves
like a nonporous catalyst, we chose to report results relative to
He only.
3.2. CO and O2 adsorption
The interaction of CO and O2 with the Au/M(OH)∗x cata-
lysts was investigated through CO and O2 TAP single-pulse
experiments. Fig. 1 shows the CO, COm10, COm2, O2, and He
area normalized responses obtained when 25 pulses of CO,
CO, COm10, COm2, and O2 in He were passed on catalyst (I)
at 298 K. The normalized responses at 373 and 473 K have
similar shapes, with the only difference in the position of the
cross point, as explained below. Comparing the shape of the re-
sponses for the reactant gases with that for He, we concluded
that irrespective of the concentration of moisture and temper-
ature, CO reversibly adsorbed on the surface, because the CO
response curves crossed the He curve. On the other hand, O2
showed an irreversible adsorption at all temperatures, with the
O2 response curves under the He response curve [43]. The CO
intersection point shifted to a shorter time with increased mois-
ture content in CO at 298 K. The difference grew smaller with
increasing temperature, and there was almost no difference at
3. Results and discussion
The catalysts used in this study were very active for CO oxi-
dation in the absence of moisture. In contrast, in the presence of
moisture, the catalytic activity was totally suppressed for cata-
lyst (I) and remained unchanged for catalyst (II). To understand
the different behavior of the two catalysts in the presence of wa-
ter, single-pulse and alternating-pulse TAP experiments were
performed, and the TAP response curves were analyzed.
3.1. Gas transport
∗
Four single-pulse experiments with inert gases He and Ar
were performed at 298, 373, and 473 K to determine the pulse
Fig. 1. Area normalized responses at 298 K on 0.024 g Au/Ti(OH) catalyst for
3
He, CO, CO
, CO , and O .
m2 2
m10

M. Olea et al. / Journal of Catalysis 248 (2007) 60–67
63
Table 1
(cm−1), εb is porosity of the bed, and ka is adsorption rate con-
stant (cm3 of gas mol−1 s−1). The parameter as can be extracted
from the value of kaꢀ at given values of ka and SV . Therefore,
TAP single-pulse experiments evidenced the fact that the con-
centration of active sites available for CO adsorption is lower
on catalyst (I) than on catalyst (II) and also show that water and
part of CO should adsorb on the same type of adsorption sites
as the concentration of active sites available for CO adsorption
decreases by increasing the water content. Thus, competitive
adsorption of H2O and CO may occur on both catalysts.
To determine how strongly the water was adsorbed on the
two catalysts, the TAP single-pulse experiments on CO adsorp-
tion in the absence of water were repeated after pulsing He over
the catalysts for 30 min. Whereas for catalyst (II), the former
position of CO intersection time with He curve was recovered,
for catalyst (I), the new intersection time of dry CO with the
He curve was shifted to shorter times (about 50 ms at 298 K,
35 ms at 373 K, and about 20 ms at 473 K). On other words,
on catalyst (II), the adsorption of water seems to be weak and
reversible, whereas for catalyst (I), stronger and irreversible ad-
sorption of water seems to occur.
Mean apparent adsorption rate constants for CO, CO , and CO
373, and 473 K, respectively, and standard deviations
, at 298,
m2
m10
Adsorption rate constant Temperature (K)
−1
(s ) and (σ)/species
298
0.014 (0.0003) 0.021 (0.0004)
0.010 (0.0005) 0.018 (0.00055) 0.025 (0.0008)
0.007 (0.0006) 0.013 (0.0007) 0.024 (0.0009)
373
473
CO
0.026 (0.0006)
CO
m2
CO
m10
473 K. The shift of the intersection point to shorter times can be
explained as follows. For a reversible adsorption, the time po-
sition at which the reversible adsorption and inert gas response
curves intersect depends on the adsorption and desorption rate
constants. The adsorption rate depends partly on the concentra-
tion of adsorption sites at the surface. If some of the active sites
are occupied with water, then the concentration of active sites
available for CO adsorption decrease and thus the apparent ad-
sorption rate constant decreases, and in this case the influence
of desorption on the intersection point becomes less significant,
leading to a shorter time of the intersection point.
Analysis of the zeroth, first, and second moments of the CO
TAP curves led to simple relations who were used to calculate
adsorption and desorption rate constants [38]. Table 1 presents
the apparent adsorption rate constants for the adsorption of CO,
COm2, and COm10, at 298, 373, and 473 K, respectively. The
apparent rate constants increase with temperature for all three
species considered, and at a constant temperature they decrease
slightly with increasing water content in the moisture. At 473 K,
only slight differences in the CO apparent adsorption rate con-
stants are seen for dry CO and the two moistures.
But are these small differences between values of apparent
adsorption rate constants really significant, or are they due to
experimental uncertainties? To explore this question, the ex-
periments were repeated three times for each temperature and
CO mixture. In all cases, the standard deviation was less than
the difference between the values of apparent adsorption rate
constants. As a general tendency, the standard deviation of all
samples in a pulse response was lower at lower temperature.
The level of noise was higher with higher amounts of water in
the CO mixture.
For the catalyst (II), the adsorption of CO in the absence and
presence of water and O2 adsorption show the same features as
for the catalyst (I). The only difference was in the time posi-
tion at which the CO curve intersects the He curve, which was
shifted to longer time. At 298 K, the time position of the inter-
section of the two curves was shifted 300 ms; in this case, the
apparent adsorption rate constant was 1.5 times higher than on
catalyst (I). Similarly, we can conclude that the shift of the in-
tersection point to longer time is due to the existence of a higher
concentration of active sites for CO adsorption on catalyst (II).
The apparent adsorption rate constant, kaꢀ , is defined by:
Concerning the oxygen responses at a pulse intensity of
about 1 × 1016 molecules/pulse, an irreversible adsorption oc-
curs on both catalysts. Fig. 1 shows the O2 adsorption features
for catalyst (I) at 298 K. The normalized responses at 373 and
473 K follow the same trend as that at 298 K. The amount of
adsorbed oxygen increases with increasing temperature up to
373 K. Following the procedure described in [38], the oxygen
was found to adsorb molecularly, and the conversion was esti-
mated to be 13% at 298 K and 26% at 373 K, whereas at 473 K,
the oxygen conversion decreased at 11%. We roughly calculate
apparent activation energy to be 11 kJ mol−1 from room tem-
perature to 373 K. For catalyst (II), more oxygen was adsorbed:
14.5% at 298 K, 29% at 373 K, and about 12.5% at 473 K,
and the apparent activation energy was around 7.5 kJ mol−1
,
which implies a weaker interaction of adsorbed oxygen with
Fe2+ oxygen vacancy sites. In addition, O2-TPD experiments
suggested that of all the adsorbed oxygen species on catalyst (I)
and (II) are active for CO oxidation at room temperature [35].
To explain the maximum conversion at 373 K, the following
mechanism of the oxygen adsorption was proposed:
+e−
O2 + Y ⇔ O2Y −→ O2-Y.
At low pulse intensities, only reversible physical adsorption
occurs. In this case, the oxygen-adsorbed species could be con-
sidered a mobile precursor for the oxygen chemisorption. If
a molecule adsorbs in the precursor state, it may (i) become
chemisorbed, (ii) be inelastically scattered back into the gas
phase, or (iii) hop to a neighboring site, in which case pathways
(i) and (ii) are again open. For high pulse intensities, as those
used within this study, the oxygen chemisorption becomes com-
petitive with the oxygen desorption. Previous experimental re-
sults have indicated that significant changes in the shape of the
TAP response curves occur when differences in conversions are
at least 9%. Because for both catalysts, the differences in con-
version from 298 to 373 K were >9%, we concluded that these
asSV (1 − εb)ka
kaꢀ =
,
(2)
εb
where as is surface concentration of active sites (mol cm−2 of
catalyst), SV is surface area of catalyst per volume of catalyst

64
M. Olea et al. / Journal of Catalysis 248 (2007) 60–67
data are significant and thus the activation energies for the two
catalysts can be differentiated.
sorb on 0.5% of the Ti sites at the surface for the COm10 mixture
and 0.1% of the Ti sites for the COm2 mixture.
Now the problem is to determine the nature of adsorption
sites on which the competitive adsorption of CO and water oc-
curs. On the Au/Ti(OH)∗4 catalyst, three CO adsorption sites
were characterized by FT-IR spectra at νCO = 2119, 2136, and
2188 cm−1. The band at 2188 cm−1 was assigned to linear
CO on Ti4+, the band at 2119 cm−1 was assigned to linear
CO on Au metallic particles, and the band at 2136 cm−1 was
tentatively assigned to CO on Au particles strongly interact-
ing with the support surface. For the Au/Fe(OH)∗3 catalyst,
only two bands were observed, at 2170 and 2137 cm−1. The
Taking into account all of these findings, we can assume
competitive adsorption of a part of CO and water on the Ti3+
-
oxygen vacancy sites situated at the gold interface, where oxy-
gen has the highest probability to adsorb as well. At low tem-
perature, water, as a polar molecule, has a higher adsorption
probability. Thus, these sites will be blocked for CO adsorp-
tion, and the total amount of adsorbed CO will decrease with
increasing water content.
For catalyst (II), an increased amount of adsorbed oxygen
(3.5 µmol g−1) was found [37]. This indicates a slightly in-
creased number of Fe2+oxygen vacancies at the surface than
band at 2170 cm−1 can be assigned to linear CO on Fe3+
,
of Ti3+ oxygen vacancies, that is, 2.1 × 1018 sitesg−1
.
whereas the higher-frequency νCO peak at 2137 cm−1 may be
attributed to probably more finely dispersed Au particles com-
pared with Au particles in the Au/Ti(OH)∗4 catalyst [35,37].
The smaller size of Au particles in Au/Fe(OH)∗3 compared with
Au/Ti(OH)∗4 was confirmed by the Au coordination number de-
rived from EXAFS analysis (7.4 1 and 9.8 1, respectively)
[36,40]. The former studies demonstrated that the CO mole-
cules reversibly adsorbed on Au particles and irreversibly on
the metal oxide support surfaces, and that only those adsorbed
on Au particles strongly interacting with the support surface
contributed to the catalytic CO oxidation [35,37]. Therefore,
on both catalysts, a part of CO is expected to adsorb on oxygen
vacancies of the support. The finding of a reversible adsorp-
tion feature of CO on the TAP experiment most likely indicates
that CO adsorbs preferentially on Au nanoparticles, with only
a small amount of CO adsorbing on M sites. The more Au
nanoparticles on the catalyst surface, the greater the CO ad-
sorption.
3.3. CO–O2 reaction
A series of experiments for CO–O2 reactions were per-
formed in which dry and moisturized carbon monoxide and
oxygen were pulsed over the Au/Ti(OH)∗4 (I) and Au/Fe(OH)∗3
(II) catalysts, either simultaneously or at various time intervals
between the reactant pulses. Fig. 2 shows the normalized re-
sponse of CO when the two reactants CO and O2 (at a 1:1 ratio)
were pulsed simultaneously at 298 K, along with the He re-
sponse, over catalyst (II). The normalized response of CO from
a CO pulse at 298 K is also shown for comparison, providing
a better understanding of the behavior of the catalyst during re-
action. Even at this low temperature, the catalysts shows high
activity for CO oxidation. The CO conversion was 45% and in-
creased to 52% at 473 K for (I) and was between 50 and 79%
for (II). As expected, the TAP conversions are lower than those
obtained under steady-state conditions in a fixed-bed flow re-
actor, in which larger amounts of catalysts were used, that is,
0.200 g instead of 0.024 g.
Moreover, it is highly likely that at low temperature, the
dissociative adsorption of water occurs on coordinatively unsat-
urated M sites that are available for CO adsorption. Increasing
the temperature will likely cause desorption of water from the
surface, making the M sites available for CO adsorption once
again. At 298 K, the number of CO molecules pulsed over the
catalyst in a single-pulse experiment was around 2.5 × 1017.
For the COm10 mixture, the number of water molecules pulsed
together with CO was 2.5 × 1014; for the COm2 mixture, it was
5 × 1013. Because the catalyst surface has about 1020 Ti4+ and
about the same number of Fe3+ sites per gram, the number of
Mx+ sites for a 0.024-g catalyst loading is about 2.4 × 1018.
If all of these sites would have been available for the adsorp-
tion of water, then the amount of adsorbed water would have
been as little as 0.01% of the exposed Mx+ ions for the COm10
mixture and 0.002% for the COm2 mixture. These values are
too low to explain the difference in CO adsorption as a function
of the moisture concentration. More likely, water adsorbs only
on coordinatively unsaturated M(x−1)+ sites adjacent to oxygen
vacancies, which are as little as 1% of the exposed Mx+ ions.
On the other hand, ESR spectra show that O2 also adsorbed
on those sites to form Ti4+–O−2 species, which may be located
more at the gold interface [37]. The amount of oxygen adsorbed
on the catalyst (I) was 3.1 µmol g−1 [37], meaning that the num-
ber of the Ti3+-oxygen vacancy sites is 1.9 × 1018 sitesg−1, or
4.56 × 1016 sites in the TAP reactor. In this case, water can ad-
∗
Fig. 2. Area normalized responses at 298 K on 0.024 g Au/Fe(OH) catalyst
3
for He, CO from CO single-pulses and CO from simultaneous CO + O pulse.
2

M. Olea et al. / Journal of Catalysis 248 (2007) 60–67
65
∗
Fig. 3. Area normalized responses at 473 K on 0.024 g Au/Ti(OH) catalyst for
4
He, CO from CO single-pulses, CO from simultaneous CO + O pulse, CO
2
m10
+ O pulse.
∗
from CO
single-pulses, and CO
from simultaneous CO
m10
m10
m10
2
Fig. 4. Dimensionless responses at 373 K on 0.024 g Au/Ti(OH) catalyst (a
4
∗
and b) and on 0.024 g Au/Fe(OH) catalyst (c and d) for CO : a and c in the
2
3
CO pulse, b and d in the CO + O pulse.
2
2
On both catalysts, the shape of the CO response was differ-
ent from that obtained when CO alone was pulsed. The reactant
curve shifts to shorter time and became slightly narrower. More-
over, after very careful analysis, we concluded that CO as a
reactant gives a response curve with two peaks. The two peaks
can be hardly distinguished at this temperature, but by increas-
ing the temperature to 473 K, the peaks are more pronounced
as shown in Fig. 3. Both peaks show their maxima at shorter
times than that of CO alone, which suffers only adsorption on
the catalyst surface. Taking into account the findings related to
the adsorption of CO, we can assign the first peak from CO re-
sponse to the CO which attempted to adsorb on the same sites as
oxygen and from which it was forced to desorb earlier because
O2 adsorption is stronger than CO adsorption. More CO seems
to be adsorbed on other types of adsorption sites, as demon-
strated by the presence of the second peak in the CO response.
Fig. 3 also shows the responses of COm10 over catalyst (I)
when COm10 was pulsed simultaneously with O2 at a 1:1 ratio
or alone at 473 K. The only difference between the two curves is
the shift to shorter times for COm10 pulsed simultaneously. The
same behavior was found at the other two temperatures studied,
as well for the COm2 mixture. Moreover, no reaction product
was detected when water was present in the reactant mixture.
The suppression of CO2 production by water vapor was also
observed in the catalytic oxidation in a closed circulating sys-
tem. Therefore, the shifts in the COm10 and COm2 responses
are due only to the forced earlier CO desorption in the presence
of O2.
The CO2 adsorption/reaction features over the two catalysts
were well characterized by comparing the formation of CO2
from the single-pulse of the 1:1 mixture of CO and O2 with
the TAP response of CO2 pulse (see Fig. 4). The dimensionless
CO2 response curve for the CO2 pulse at 373 K was almost sim-
ilar over the two catalysts. Slightly slower CO2 desorption can
be seen over catalyst (I). In contrast, at the same temperature,
the CO2 response curve for the CO + O2 pulse over catalyst (II)
is sharper than that over catalyst (I). This means that the CO2
formation rate is higher on the Au/Fe(OH)∗3 catalyst.
Next, alternating pulse experiments were performed at
373 K over catalyst (I). O2 was pulsed first from the pump
valve and then interrogated with the CO pulse from the probe
valve at time intervals varying from 50 to 10 000 ms. The re-
action product, CO2, was monitored as a function of the time
interval between the two pulses. When dry CO was used as
the probe molecule, CO2 formation increased rapidly with in-
creasing time interval, reaching a maximum at an interval of
1000 ms. CO2 yield decreased with further increases in the time
interval and almost vanished after a 6000-ms interval. These
findings demonstrate the participation of the adsorbed oxygen
and rule out the possibility that lattice oxygen atoms of the cata-
lyst contribute to the oxidation reaction. When COm10 or COm2
was the probe mixture, no reaction product was observed irre-
spective of time interval.
The experimental procedure was repeated for a mixture of
CO and O2 at a 1:3 ratio. When CO and O2 were pulsed simul-
taneously, the CO response showed only a single peak on both
catalysts, which shifted to shorter time and narrowed compared
with the CO response for adsorption only. For example, over
catalyst (I), the time shift decreased with increasing tempera-
ture, from 500 ms at 298 K to 300 ms at 473 K. CO conversion
was 50% at 298 K and 56% at 473 K for catalyst (I), with higher
For catalyst (II), the CO adsorption features when CO dry
and O2 were pulsed simultaneously were similar to those ob-
served over catalyst (I), but the maxima appeared earlier. An-
other difference in the behavior of catalyst (II) was the finding
that the presence of water, irrespective of its content, did not af-
fect catalyst activity. Virtually the same conversion of CO was
obtained in the absence and the presence of water.

66
M. Olea et al. / Journal of Catalysis 248 (2007) 60–67
conversions obtained on catalyst (II): 53% at 298 K and 82% at
473 K. This finding was expected. Increasing the amount of
oxygen pulsed over the catalyst caused an increase in the num-
ber of active M sites occupied with O2 at the expense of CO.
The greater the quantity of oxygen adsorbed on the M(x−1)+
oxygen vacancy sites at the support–gold interface, the higher
the reaction probability; in other words, the shape of the CO
response curve is increasingly determined by the reaction rate.
No reaction product was detected in the presence of mois-
ture, irrespective of water content and temperature over cata-
lyst (I). Comparing COm10 and COm2 responses resulting from
the simultaneous pulse with oxygen or from their single pulses
over catalyst (I) demonstrated a shift to the shorter time in the
presence of oxygen. But the time shift was constant within the
temperature range, supporting the assumption that in the pres-
ence of oxygen, CO is forced to desorb earlier.
at the gold–oxide support interface increases. Oxygen vacan-
cies on TiO2 and Fe2O3 are electrically compensated for by
the formation of Ti3+ and Fe2+, on which O2 adsorbs to form
Ti4+–O2− species and Fe3+–O−2 species. The former species is
more stable than the latter species. Indeed, the temperature peak
in TPD of oxygen adsorbed on the Au/Fe(OH)∗3 catalyst, 400 K,
was 50 K lower than that for the Au/Ti(OH)∗4 catalyst [35]. The
stronger adsorption of O2 on the Ti(OH)∗4 support may lead to
the lower reactivity with CO.
To explain the different behavior of the two catalysts in the
absence and presence of water, the support effect also should
be considered. In our opinion, the difference in the electronega-
tivity of their cations from the supports is responsible for the
stronger dissociative adsorption of water on the Au/Ti(OH)4∗
catalyst, which leads to lower reactivity to adsorbed CO to form
CO2 by the water–gas shift reaction. The water–gas shift re-
action is more facile on the Au/Fe(OH)∗3 catalyst, and the Fe
sites become free for oxygen adsorption by the water–gas shift
reaction, allowing further reaction to occur. Andreeva’s stud-
ies on low-temperature water–gas shift over gold catalysts [45,
46] prepared by a modified deposition–precipitation technique
(i.e., gold hydroxide deposited on freshly precipitated metal hy-
droxide) showed a large variation in activity depending on the
nature of support. Thus, the temperature for 50% conversion
was 423 K for the Au/FeO3 catalyst with 3.5-nm-diameter Au
particles, but 573 K for the Au/TiO2 catalyst with Au particle
diameters <3 nm. Moreover, a recent DRIFTS study combined
with mass spectrometry showed that the water–gas shift reac-
tion occurred even at room temperature over an Au/Fe2O3 cat-
alyst obtained by co-precipitation, with an Au loading of 4.48%
and Au particle sizes of 3–5 nm (mean size, 2.8 nm) [47]. The
adsorptive mechanism comprising CO adsorption on Au parti-
cles, formation of carbonates and bicarbonates at the Au–Fe2O3
interface, and their decomposition to CO2 appears to prevail
at low temperature, whereas at around 473 K, the regenerative
mechanism is dominant. Under these conditions, the catalyst
support Fe2O3 is easily reducible in the presence of Au. The
study also revealed that the Au particles are essential for H2
production and the reoxidation step involving H2O as the re-
actant. The greatest effect is associated with the presence of a
higher concentration of low-coordinated sites on the surface of
smaller particles on the Au/Fe(OH)∗3 catalyst, which leads to
a higher stabilization of reaction intermediates, such as CO or
oxygen, on the surface of gold.
3.4. Au/Ti(OH)∗4 versus Au/Fe(OH)3∗
TAP experiments revealed that the CO oxidation reaction on
the Au/Ti(OH)∗4 catalyst was totally suppressed by the presence
of moisture (200 or 1000 ppm) and that water had no effect on
the activity of the Au/Fe(OH)3∗ catalyst. Previous experiments
also showed the poisoning effect of water on the catalyst (I)
and the insensitivity of catalyst (II) to the presence of mois-
ture [35–37]. TAP also showed that whereas on catalyst (II),
the adsorption of water seems to be weak and reversible, on
catalyst (I), strong and irreversible adsorption of water seems
to occur.
Moreover, in the absence of water, TAP confirmed that the
Au/Fe(OH)∗3 catalyst is more active than the Au/Ti(OH)∗4 cat-
alyst. Other studies have shown that the temperature needed
to start CO oxidation is 50 K lower for the Au/Fe(OH)∗3 cata-
lyst than for the Au/Ti(OH)∗4 catalyst, whereas the temperatures
for 50 and 100% conversions are lower by 100 and 160 K, re-
spectively [40]. The question is why these two catalysts behave
differently, given that they were prepared and pretreated in the
same way and had the same gold content (3 wt%). Experimental
results to date have confirmed the same mechanism for CO ox-
idation on the two catalysts [35,37]. This mechanism involves
reversible CO adsorption on the Au nanoparticle surface; O2
molecular adsorption, probably as O−2 bonded to Ti or Fe ions
as indicated by ESR [36] on oxygen vacancies on the oxide
support near to the Au nanoparticles; and subsequent reaction
between the adsorbed CO and O−2 to form a CO2 molecule and
an active oxygen atom, which rapidly reacts with CO to form
another CO2 [35,37].
4. Conclusion
The efficiency of supported gold catalysts depends on gold
dispersion and the gold–support interaction on one hand, and
on the nature, crystalline type, and pore structure of metal oxide
supports, on the other hand. The catalytic activity may change
by several orders of magnitude due to the effect of the Au–Au
coordination number, making this effect dominant and crucial
for catalysis by Au [44]. More CO that may adsorb on the Au
nanoparticles is seen on the Au/Fe(OH)∗3 catalyst than on the
Au/Ti(OH)∗4 catalyst. On the other hand, as the size of the Au
nanoparticles decreases, the number of the O2 adsorption sites
The TAP study revealed that moisture in the reactant gas had
different effects on the activity of Au/Ti(OH)∗4 and Au/Fe(OH)3∗
catalysts with respect to CO oxidation. Whereas the activity of
the Au/Ti(OH)∗4 catalyst was totally suppressed irrespective of
the water content between 200 and 1000 ppm, no effect was
observed for the Au/Fe(OH)∗3 catalyst. When the CO:O2 ratio
was 1:1 in the absence of water, CO oxidation was character-
ized by a two-peak feature in the CO response curve, with only
one peak seen in the presence of water. At a CO:O2 ratio of
1:3, the CO response curve exhibited one peak with a peak

M. Olea et al. / Journal of Catalysis 248 (2007) 60–67
67
shift to shorter time. These TAP results indicate that CO and
water absorbed competitively and that water blocked the ad-
sorption sites for O2 irreversibly on Au/Ti(OH)∗4 and reversibly
on Au/Fe(OH)∗3. The CO adsorption was weak and reversible,
whereas O2 adsorption was strong and irreversible, but weaker
on Au/Fe(OH)∗3 than on Au/Ti(OH)∗4.
[21] R.J. Davis, Science 301 (2001) 926–927.
[22] Q. Fu, H. Saltsburg, M. Flytzani-Stephanopoulos, Science 301 (2003)
935–938.
[23] N.M. Gupta, A.K. Tripathi, J. Catal. 187 (1999) 343–347.
[24] N.M. Gupta, A.K. Tripathi, Gold Bull. 34 (2001) 120–128.
[25] S.T. Daniells, A.R. Overweg, M. Makkee, J.A. Moulijn, J. Catal. 230
(2005) 52–65.
[26] M. Daté, M. Haruta, J. Catal. 201 (2001) 221–224.
[27] M. Haruta, T. Takase, T. Kobayashi, S. Tsubota, in: S. Yoshida, N.
Takezawa, T. Ono (Eds.), Catalytic Science and Technology, vol. 1, Ko-
dansha, Tokyo, 1991, pp. 331–335.
Acknowledgments
This work was supported by Core Research for Evolutional
Science and Technology (CREST) of the Japan Science and
Technology Corporation (JST). The authors thank Dr. H. Liu
and Dr. A.I. Kozlov for preparing the catalysts.
[28] E.D. Park, J.S. Lee, J. Catal. 186 (1999) 1–11.
[29] C.K. Costelo, M.C. Kung, H.-S. Oh, Y. Wang, H.H. Kung, Appl. Catal.
A 232 (2002) 159–168.
[30] D.A.H. Cunningham, W. Vogel, M. Haruta, Catal. Lett. 63 (1999) 43–47.
[31] G.C. Bond, D.T. Thompson, Gold Bull. 33 (2000) 41–51.
[32] H.-S. Oh, C.K. Costelo, C. Cheung, H.H. Kung, M.C. Kung, Stud. Surf.
Sci. Catal. 139 (2001) 375–381.
References
[33] M.A. Bollinger, M.A. Vanice, Appl. Catal. B 8 (1996) 417–443.
[34] G.B. Raupp, J.A. Dumisec, J. Phys. Chem. 89 (1985) 5240–5246.
[35] H. Liu, A.I. Kozlov, A.P. Kozlova, T. Shido, Y. Iwasawa, Phys. Chem.
Chem. Phys. 1 (1999) 2851–2860.
[36] A.P. Kozlova, A.I. Kozlov, S. Sugiyama, Y. Matsui, K. Asakura, Y. Iwa-
sawa, J. Catal. 181 (1999) 37- 48 .
[37] H. Liu, A.I. Kozlov, A.P. Kozlova, T. Shido, K. Asakura, Y. Iwasawa,
J. Catal. 185 (1999) 252–264.
[38] M. Olea, M. Kunitake, T. Shido, K. Asakura, Y. Iwasawa, Bull. Chem.
Soc. Jpn. 74 (2001) 255–265.
[1] M. Haruta, T. Kobayashi, H. Sano, N. Yamada, Chem. Lett. 16 (1987)
405–408.
[2] G.C. Bond, D.T. Thompson, Catal. Rev. Sci. Eng. 41 (1999) 319–388.
[3] G.C. Bond, Gold Bull. 34 (2001) 117–140.
[4] M. Haruta, M. Daté, Appl. Catal. A 222 (2001) 427–437.
[5] M. Haruta, CATTECH 6 (2002) 102–115.
[6] C.W. Corti, R.J. Holliday, D.T. Thompson, Gold Bull. 35 (2002) 111–136.
[7] D.T. Thompson, Appl. Catal. A 243 (2003) 201–205.
[8] H.H. Kung, M.C. Kung, J. Catal. 216 (2003) 425–432.
[9] M. Haruta, J. New Mater. Electrochem. Syst. 7 (2004) 163–172.
[10] M. Haruta, Gold Bull. 37 (2004) 27–36.
[39] M. Olea, M. Kunitake, T. Shido, Y. Iwasawa, Phys. Chem. Chem. Phys. 3
(2001) 627–631.
[11] C.W. Corti, R.J. Holliday, D.T. Thompson, Appl. Catal. A 291 (2005) 253–
261.
[40] Y. Yuan, A.P. Kozlova, K. Asakura, H. Wan, K. Tsai, Y. Iwasawa,
J. Catal. 170 (1997) 191–199.
[12] J.D. Henao, T. Caputo, J.H. Yang, M.C. Kung, H.H. Kung, J. Phys. Chem.
B 110 (2006) 8689–8700.
[41] Y. Yuan, K. Asakura, A.P. Kozlova, H. Wan, K. Tsai, Y. Iwasawa, Catal.
Today 44 (1998) 333-342 .
[13] F. Cosandey, T.E. Madey, Surf. Rev. Lett. 8 (2001) 73–93.
[14] P. Pyykko, Angew. Chem. Int. Ed. 43 (2004) 4412–4456.
[15] A. Cho, Science 299 (2003) 1684–1685.
[16] M. Valden, X. Lai, Dz.W. Goodman, Science 281 (1998) 1647–1650.
[17] M. Mavrikakis, P. Stoltze, J.K. Norskov, Catal. Lett. 64 (2000) 101–106.
[18] H.G. Boyen, G. Kastle, F. Weigl, B. Koslowski, C. Dietrich, P. Ziemann,
J.P. Spatz, S. Riethmuller, C. Hartmann, M. Moller, G. Schmid, M.G. Gar-
nier, P. Oelhafen, Science 297 (2002) 1533–1536.
[42] A.I. Kozlov, A.P. Kozlova, K. Asakura, Y. Matsui, T. Kogure, T. Shido,
Y. Iwasawa, J. Catal. 196 (2000) 56–65.
[43] J.T. Gleaves, G.S. Yablonskii, P. Phanawadee, Y. Shuurman, Appl. Catal.
A 160 (1997) 55–88.
[44] N. Lopez, T.V.W. Jansen, B.S. Clausen, Y. Xu, M. Mavrikakis, T. Bligaard,
J.K. Nørskov, J. Catal. 223 (2004) 232–235.
[45] Ed. Andreeva, V. Idakiev, T. Tabakova, A. Andreev, J. Catal. 158 (1996)
354–355.
[19] S.K. Shaikhutdinov, R. Meyer, M. Naschitzki, M. Bäumer, H.-J. Freund,
Catal. Lett. 86 (2003) 211–219.
[20] Y. Xu, M. Mavrikakis, J. Phys. Chem. B 107 (2003) 9298–9307.
[46] D. Andreeva, Gold Bull. 35 (2002) 82–88.
[47] B.A.A. Silberova, G. Mul, M. Makkee, J.A. Moulijn, J. Catal. 243 (2006)
171–182.