Au-catalyzed selective oxidation of CO: A steady-state isotopic transient kinetic study

Article abstract of DOI:10.1016/j.jcat.2005.06.012

Selective oxidation of CO in hydrogen, obtained from hydrocarbons, is an important reaction for producing hydrogen appropriate for use in fuel cells. This paper reports on the variations in activity and selectivity of alumina-supported Au catalysts for the selective oxidation of CO due to the presence/absence of chloride (Cl^-) species and to different Au particle sizes [dispersion, D (%)]. This study made use of elemental analysis, chemisorption, kinetic studies, and steady-state isotopic transient kinetic analysis (SSITKA) to explore these variations. A comparison of Au catalysts with different Au loadings to 5 wt% Pt/γ-Al₂O₃ is also made. Magnesium citrate was added during preparation of one of the Au catalysts [referred to as Au(Mg)], because it was found to displace Cl ^- species, resulting in smaller Au particle sizes. Rapid partial deactivation was observed for all of the catalysts during the first 30 min of time-on-stream irrespective of the presence or absence of Cl^- species. A reactivation after initial rapid deactivation was observed during reaction only for the Au(Mg) catalyst. This catalyst, the Au catalyst with the smallest particle size (D>90%) and the lowest Cl^- content, was more active than the catalysts with larger particle sizes (lower %D) at steady state. But all Au catalysts had similar initial activity. A decrease in the concentration of reactive intermediates (NI-CO2) along with a very large decrease in intrinsic site activity (k), both obtained from SSITKA, appear to have been the cause of the initial deactivation of the Au catalysts. For the Pt catalyst, the decreased concentration of reactive intermediates was the main cause of initial deactivation. Although the precise reasons for the increase in activity with time-on-stream after initial deactivation of the Au catalyst prepared with magnesium citrate are not clear, it can be hypothesized that increasing amounts of water formed during the reaction may have helped in decomposing deactivating Au carbonate species or in restructuring/ reconstructing the active sites. The smaller Au particle size and/or absence of Cl^- species may have contributed to this. Although there was a slight increase in the number of active CO₂ intermediates, most of the increase in rate for the Au(Mg) catalyst was due to an increase in average site activity. There is no evidence that Mg functions as a chemical promoter.

Full text of DOI:10.1016/j.jcat.2005.06.012

Journal of Catalysis 234 (2005) 151–160
www.elsevier.com/locate/jcat
Au-catalyzed selective oxidation of CO: a steady-state isotopic transient
kinetic study
a
b
a,∗
Sourabh S. Pansare , Amornmart Sirijaruphan , James G. Goodwin Jr.
a
Department of Chemical and Biomolecular Engineering, Clemson University, Clemson, SC 29634-0909, USA
b
Department of Chemical Engineering, King Mongkut’s University of Technology Thonburi, Bangkok, 10140, Thailand
Received 14 April 2005; revised 9 June 2005; accepted 13 June 2005
Available online 19 July 2005
Abstract
Selective oxidation of CO in hydrogen, obtained from hydrocarbons, is an important reaction for producing hydrogen appropriate for use
in fuel cells. This paper reports on the variations in activity and selectivity of alumina-supported Au catalysts for the selective oxidation of
−
CO due to the presence/absence of chloride (Cl ) species and to different Au particle sizes [dispersion, D (%)]. This study made use of
elemental analysis, chemisorption, kinetic studies, and steady-state isotopic transient kinetic analysis (SSITKA) to explore these variations.
A comparison of Au catalysts with different Au loadings to 5 wt% Pt/γ -Al O is also made. Magnesium citrate was added during preparation
2
3
−
of one of the Au catalysts [referred to as Au(Mg)], because it was found to displace Cl species, resulting in smaller Au particle sizes. Rapid
partial deactivation was observed for all of the catalysts during the first 30 min of time-on-stream irrespective of the presence or absence
−
of Cl species. A reactivation after initial rapid deactivation was observed during reaction only for the Au(Mg) catalyst. This catalyst, the
−
Au catalyst with the smallest particle size (D > 90%) and the lowest Cl content, was more active than the catalysts with larger particle
sizes (lower %D) at steady state. But all Au catalysts had similar initial activity. A decrease in the concentration of reactive intermediates
(
N
I-CO2
) along with a very large decrease in intrinsic site activity (k), both obtained from SSITKA, appear to have been the cause of the
initial deactivation of the Au catalysts. For the Pt catalyst, the decreased concentration of reactive intermediates was the main cause of initial
deactivation. Although the precise reasons for the increase in activity with time-on-stream after initial deactivation of the Au catalyst prepared
with magnesium citrate are not clear, it can be hypothesized that increasing amounts of water formed during the reaction may have helped
in decomposing deactivating Au carbonate species or in restructuring/reconstructing the active sites. The smaller Au particle size and/or
−
absence of Cl species may have contributed to this. Although there was a slight increase in the number of active CO intermediates, most
2
of the increase in rate for the Au(Mg) catalyst was due to an increase in average site activity. There is no evidence that Mg functions as a
chemical promoter.

2005 Elsevier Inc. All rights reserved.
Keywords: Selective oxidation of CO; CO oxidation; Au/γ -Al O , Pt/γ -Al O ; Magnesium effect on Au catalysts; Steady-state isotopic transient kinetic
2
3
2 3
analysis; Deactivation; Isotopic tracing
1
. Introduction
of hydrogen and oxygen as fuel and produce electricity, with
water the sole byproduct. The hydrogen is usually obtained
from hydrocarbons by partial oxidation or steam reforming,
with CO and CO2 as the byproducts. Unfortunately, CO is
a poison for fuel cell electrodes, and the acceptable limit of
CO in hydrogen for fuel cells is 10 ppm. Because the CO
concentration in the H2-rich stream obtained from hydrocar-
bons is around 1% (10,000 ppm), removing the CO from that
stream is extremely important for efficient operation of fuel
cells. Pressure swing adsorption, methanation, and selective
Fuel cells are gaining interest because of their potential
to replace internal combustion engines in automobiles and
their use in all sorts of electric power generation off the grid.
Proton exchange membrane fuel cells (PEMFCs) make use
*
Corresponding author. Fax: +1-864-656-0784.
E-mail address: james.goodwin@ces.clemson.edu (J.G. Goodwin Jr.).
0021-9517/$ – see front matter  2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2005.06.012

152
S.S. Pansare et al. / Journal of Catalysis 234 (2005) 151–160
CO oxidation are some of the techniques available for this
purpose [1]. Of these, selective CO oxidation is the easiest
and most economical method of CO removal (by formation
of CO2).
Supported metal catalysts such as Pt, Rh, and Ru on ox-
ide supports have been shown to be useful for the selective
oxidation of CO in H2. Much research has been done on Pt
catalysts, including studies of deactivation behavior [2], ef-
fect of Fe promotion [3], kinetics [4], and effects of water
vapor and CO2 on activity and selectivity [5].
Au, when present in the bulk state, is inert and hence con-
sidered a poor catalyst. But highly dispersed nanosized Au
particles supported on metal oxides can be very active for the
CO oxidation reaction even at very low temperatures [6–11].
Au has a greater availability and lower cost than Pt. Because
of these reasons, researchers have started taking an interest
in Au catalysis for selective CO oxidation.
Kung [7]. The DP method was chosen as the catalyst prepa-
ration method because it yields hemispherical metal parti-
cles, which, according to Haruta [6], have larger turnover
frequencies. Au solutions of different concentrations were
used to produce catalysts with different metal dispersions.
Before the catalysts were prepared, the support (Al2O3,
◦
Vista B) was calcined at 500 C for 10 h to form γ -Al2O3.
The Au precursor, hydrogen tetrachloroaurate (HAuCl4·
3H2O, Alfa Aesar), was dissolved in water and stirred for
20 min. The aqueous solution (25 ml/g of support) was then
added to the support (γ -Al2O3) and stirred for 2 h. The stir-
ring was carried out at low temperature in an ice bath to
prevent coagulation of Au clusters in the solution [7]. The
pH of the slurry as measured was 4.5 and the temperature
◦
was 3.4 C after 1 h. During preparation of the lowest Au-
loaded catalyst, magnesium citrate (Alfa Aesar) was added
to the solution after 1 h of stirring. (Magnesium citrate is
believed to reduce the particle size by preventing the coag-
ulation of Au clusters during preparation [7].) The Mg/Au
ratio was 1.55. The pH of the slurry as measured was 5.97
after the addition of magnesium citrate. After stirring for 2 h,
the solutions were filtered and the catalysts washed three
times with water at room temperature and once with warm
Two important characteristics that catalysts for the se-
lective CO oxidation must have are high activity and high
selectivity toward CO2 formation. Au catalysts have been
reported to be highly active and selective for this reaction
at typical fuel cell operating temperatures [7]. Kahlich et
al. [8] studied the kinetics of this reaction at low temper-
ature on Au supported on Fe2O3, and Grunwaldt et al. [9]
conducted a comparative study of Au/TiO2 and Au/ZrO2 for
simple CO oxidation. Other studies have investigated the ef-
fect of particle size on activity and selectivity [7], the effects
of water concentration and the addition of MgO and MnOx
to the catalyst on catalyst activity [10], the nature of the ac-
tive sites [12], and the effects of preparation method and O2
and CO2 concentration in the feed stream [13]. Kung and
coworkers have studied in detail the effect of the presence
◦
water. The catalysts were dried at 100 C for 30 min and then
◦
calcined under flowing dry air (HC free) at 350 C for 4 h.
Throughout the paper, the two Au catalysts without magne-
sium citrate are referred to as “Au1” and “Au2” (with Au
loadings of around 3 and 2%, respectively), whereas the
Au catalyst with citrate is referred to as “Au(Mg)” (with an
Au loading of around 0.5%). The 5 wt% Pt/γ -Al2O3 catalyst
was prepared by incipient wetness impregnation method as
described in previous work [2]. This catalyst is referred to as
“Pt” throughout the paper.
−
of Cl species, which originate from the Au precursor so-
lution, on the activity of Au catalysts for CO oxidation and
selective CO oxidation [11,14].
2.2. Catalyst characterization
This paper reports on a study of γ -Al2O3-supported Au
catalysts (with varying Cl impurity concentrations and
−
−
Metal and Cl content (wt%) were determined by Gal-
average Au particle sizes) and their comparison with a
Pt/γ -Al2O3 catalyst using reaction kinetics and steady-state
isotopic transient kinetic analysis (SSITKA). Initially de-
veloped by Happel [15], Bennett [16], and Biloen [17],
SSITKA is a powerful technique for the in situ analysis of
surface reaction on heterogeneous catalysts. The major use
of SSITKA is for determining the average surface residence
time (τp), surface concentration (Np), and surface coverage
braith Laboratories (Knoxville, TN).
2.2.1. X-Ray diffraction
X-Ray diffraction (XRD) analysis was used to study the
Au and Pt catalysts using a Philips X’Pert X-ray diffrac-
tometer with monochromatized Cu-Kα radiation and a Ni
filter.
(θp) of the most active reaction intermediates. In this study
2.2.2. Pulse H2 titration
we used SSITKA to determine the variation of these parame-
ters for the Au and Pt catalysts with time on stream (TOS).
Pulse H2 titration was used to determine dispersion, D
(%), and metal particle size in the Au catalysts. The analysis
was carried out at 1 atm using an Altamira AMI-1 system.
Initially, the catalyst in a U-quartz tube reactor was pre-
treated with 30 sccm of 5% H2 in Ar (National Welders)
2
. Experimental
◦
at 400 C for 1 h [18]. The temperature was increased from
◦
◦
2
.1. Catalyst preparation
25 to 400 C with a ramp rate of 10 C/min. After pretreat-
ment, the catalyst was flushed with Ar (National Specialty
Gases, UHP) for 30 min to desorb the hydrogen; during
Three Au catalysts were prepared by the deposition–
precipitation (DP) method as described by Bethke and
◦
this step, the temperature was reduced from 400 to 300 C

S.S. Pansare et al. / Journal of Catalysis 234 (2005) 151–160
153
Fig. 1. Schematic diagram of the SSITKA—CO selective oxidation system.
◦
at 10 C/min. In the next step, 2% O2 in He (Holox) was
quartz wool and placed at center of the reactor with a ther-
mocouple at the bottom of this catalyst bed. The reaction was
carried out at 1.8 atm and 90 C. The reactants were fed to
◦
passed over the catalyst for 3 h at 300 C to adsorb O2 on
◦
the metal surface atoms. The catalyst was then flushed with
Ar for 1 h to remove any weakly adsorbed oxygen. The
strongly adsorbed oxygen was then titrated with pure hydro-
gen (National Specialty Gases, UHP) pulses. The H2 pulses
were passed over the catalyst, and unreacted hydrogen was
detected using a thermal conductivity detector (TCD). A wa-
ter trap was placed between the reactor and TCD to remove
any water formed during the reaction. The H2 pulses were
continued until the areas under the pulses coming out of
the reactor were constant. The amount of O2 adsorbed on
alumina-supported Au catalysts increases with temperature,
the reactor in a ratio of H2/CO/O2/He = 45/1/1/53, with a
total flow rate of 100 sccm. All Au catalysts were pretreated
◦
in a flow of hydrogen (45 sccm) at 150 C for 1 h. The tem-
◦
perature was increased from 30 to 150 C with a ramp rate
◦
◦
of 1 C/min. The Pt catalyst was pretreated at 550 C for 1 h
in a flow of hydrogen (45 sccm).
The effluent from the reactor during reaction was ana-
lyzed by gas chromatography (GC) (Varian CP-3380). In the
GC, the products were separated by a 6-ft-long 80/100 mesh
◦
carbosphere column (Alltech). At 35 C, H2, CO, and O2
◦
◦
reaching a maximum at about 340 C followed by desorption
of O2 at higher temperatures [18]. Adsorption experiments
carried out at 200 C showed very slow adsorption of O2.
Hence 300 C was chosen as the temperature for O2 adsorp-
were first separated, and then the GC was ramped to 150 C
at 20 C/min for separating CO2. The CO conversion and
◦
◦
selectivity were determined periodically using a definition
given by Manasilp and Gulari [5].
◦
tion and pulse titration in this study.
Reaction on the Au2 catalyst for different reactant flow
rates showed that external mass transfer did not affect re-
action rate measurements. The Weisz–Prater parameter was
calculated for all of the catalysts and indicated no internal
mass transfer effects at the reaction conditions used.
The dispersion of the Pt catalyst was determined by static
H2 chemisorption, the details of which were given previ-
ously [2].
2
.3. Reaction studies
2
.4. SSITKA
The kinetic studies were carried out in a stainless-steel
plug flow microreactor (0.3-inch id). A schematic of the re-
action system used is shown in Fig. 1. About 20–25 mg of
the catalyst diluted with α-Al2O3 was sandwiched between
The SSITKA studies were carried out by making a switch
from 12CO to CO (Isotec, 99 atom% C, 12 atom% O).
13
13
18
12
A trace of argon was present in the CO stream to measure

154
S.S. Pansare et al. / Journal of Catalysis 234 (2005) 151–160
Fig. 3. Variation of τ
with inverse flow rate (1/F ) for the Au1 catalyst.
I-CO
2
τI-CO2 (and consequently of NI-CO ) were essentially free
2
from readsorption effects on the Au catalysts at these con-
ditions (Fig. 3). The same was true for the Pt/γ -Al2O3 cata-
lyst [21].
3
. Results and discussion
The actual metal loadings as determined by Galbraith
Laboratories for the Au and Pt catalysts are given in Table 1.
Fig. 2. Typical SSITKA transients: (a) actual transients, (b) normalized tran-
sients.
3
.1. X-Ray diffraction
the gas-phase holdup in the reaction system. A quadrupole
mass spectrometer (Pfeiffer Vacuum, Prisma) was used to
monitor the decrease in the old label and the increase in the
new label in the CO2 along with CO coming out of the reac-
tor, as shown in Fig. 2a. The mass spectrometer was coupled
with a high-speed data acquisition system interfaced to a
personal computer using Balzers Quadstar 422 v 6.0 soft-
ware. The SSITKA parameters, including average surface
residence time (τI-CO ), surface concentration (NI-CO ), and
No peaks for metallic Au were seen in powder XRD. This
was due in large part to the small particle sizes of the Au
present in the catalysts. Also, the detection of Au peaks was
more difficult, because of overlap with the large γ -Al2O3
peaks present. The XRD pattern of the Pt catalyst after cal-
cination also showed peaks corresponding to γ -Al2O3. No
peaks for Pt metal were seen, for the same reason they were
not seen for Au.
2
2
surface coverage (θI-CO ) of reactive intermediates, were cal-
3.2. Au dispersion
2
culated using the method described by Shannon and Good-
win [19]. The average residence time was determined by
integrating the normalized isotopic transient (Fig. 2b) of the
product relative to that of the inert tracer (Ar).
A potential problem for SSITKA measurements of CO
adsorption and reaction can be readsorption of product CO2.
If this were to occur, it would cause an increase in the av-
Dispersions and particle sizes obtained from pulse H2
titration for the three Au catalysts are reported in Table 1,
Table 1
Physical properties of Au and Pt catalysts
a
c
Catalyst
Metal
wt%)
Concentration Dispersion Avg. metal
b
d
(
of Aus
(%)
particle size
(nm)
erage surface reaction residence time (τI-CO ), resulting in
2
(µmol/gcat)
overestimation of NI-CO and underestimation of k. Meth-
2
Au/γ -Al O
2.89
1.54
13.2
14.1
15.6
9
18
93
45
13.9
2
3
3
ods for removing readsorption effects to determine the true
surface reaction residence time, taking into account both in-
terparticle and intraparticle readsorption, have been given
by Ali and Goodwin [20]. Experiments performed under
the conditions of this study for different flow rates to check
whether readsorption affected the average surface residence
time of product CO2 on Au1 indicated that the values of
[
Au1]
Au/γ -Al O
6.8
2
[
Au2]
Au-Mg/γ -Al O 0.33
1.3
2
3
[Au(Mg)]
Pt/γ -Al O [Pt]
5
115.4
2.4
2
3
a
b
c
d
Maximum measured error:
± 5%;
± 10%;
± 3%;
± 10%.

S.S. Pansare et al. / Journal of Catalysis 234 (2005) 151–160
155
Table 2
Chloride (Cl ) content of the Au and Pt catalysts
−
−
a
Catalyst Chloride (Cl ) content (%)
b
c
Before pretreatment,
as prepared
After pretreatment
After reaction
Au1
Au2
Au(Mg) 0.03
Pt 0.48
0.29
0.22
0.33
0.22
0.03
–
–
–
0.06
–
–
Mg/Au2 0.67
–
a
Maximum measured error, ± 5%.
b
◦
Pretreatment in the presence of H (45 sccm) at 150 C for 1 h. Tem-
2
◦
◦
perature ramp rate, 1 C/min from 30 to 150 C.
c
After 230 min TOS of the reaction.
as are these parameters for Pt/γ -Al2O3 (determined by sta-
tic H2 chemisorption). Fukushima et al. [22] studied O2
chemisorption on supported Au and found dissociative ad-
Fig. 4. Time-on-stream behavior of CO oxidation rate for Au1, Au2,
Au(Mg) and Pt catalysts (maximum error measured: ± 5% for Au1, Au2
catalysts and ± 10% for Pt and Au(Mg) catalysts).
◦
sorption of O2. The stoichiometry of Aus:O was 2 at 200 C
◦
and 1 at 300 C [18]. Because the temperature chosen for
◦
titration, as discussed earlier, was 300 C, a stoichiometry of
Pt catalysts. For Au1 and Pt catalysts, the rate dropped by
about 90% during the first 30 min TOS before a constant
steady-state value was established. For Au2, the decrease
was about 80%. Such an initial fast partial deactivation for Pt
has been noted and studied in detail earlier [2]. The Au(Mg)
catalyst also exhibited the same rapid initial decrease in rate,
but then, surprisingly, it showed a significant increase in rate
to its steady-state value—the highest rate/gcat of any of the
catalysts. The activity of the catalyst was almost 10 times
that of the Pt catalyst at steady state. Thus an apparent reacti-
vation occurred for this catalyst after the initial deactivation.
A similar result on Au catalysts after injection of pulses of
water during selective CO oxidation was reported by Grisel
and Nieuwenhuys [10]. An important observation from the
present study is that the initial partial deactivation behavior
1
was used for calculations here. The reactions taking place
on the catalyst surface can be summarized as follows:
Aus + O → Aus–O
and
(1)
Aus–O + H2 → Aus + H2O.
From the results given in Table 1, the Au(Mg) catalyst had
the largest dispersion and, consequently, the smallest particle
size.
(2)
+
During Au catalyst preparation, HAuCl4 forms H and
−
AuCl4 ions after dissolving in water. The chloride ligands
in AuCl4 are then displaced by hydroxyl groups and water
molecules, resulting in a tendency to coagulate by condensa-
tion into clusters of Au complexes and hence forming larger
particles. Another possible reason for formation of the larger
−
−
of the Au catalysts was independent of the amount of Cl
species present or the Au particle size.
−
Au clusters is the presence of residual Cl species from the
Fig. 5 shows the selectivity variations for the Au and Pt
catalysts. The trend for selectivity was similar to the rate of
the reaction. All of the catalysts showed initial rapid declines
in CO2 selectivity from initial values of 80–100%. At steady
state, Au1 and Au2 catalysts showed selectivities of around
18 and 22%, respectively. The Pt catalyst exhibited a steady-
state CO2 selectivity of around 35%. Although these three
catalysts showed similar behavior, Au(Mg) displayed quite
different and reproducible behavior, similar to the rate of re-
action. During the initial period (i.e., ꢀ30 min of TOS), the
CO2 selectivity decreased to a minimum (around 10%) be-
fore increasing to a high steady-state value (around 65%).
A question can be raised about the effect on CO2 selec-
tivity of the presence of MgO in the Au(Mg) catalyst, be-
cause CO2 can react with MgO. The concentration of MgO
present on the catalyst surface was around 210 µmol/gcat.
Considering the maximum rate of CO2 formation of around
18 µmol/(gcat s), it would have taken around 11.5 s for the
MgO to be totally saturated by CO2. The initial rapid de-
precursor solution, which causes agglomeration of Au par-
ticles during heat treatment [14]. The greater the amount of
−
Cl species present, the larger the particle size. The amounts
−
of Cl present in the various catalysts prepared in this study
are given in Table 2. The Mg citrate added during prepara-
tion serves a dual purpose. First, it acts as a ligand to Au3
+
ions and competes with hydroxyl groups for attaching to
3+
Au [7]. This results in displacement of the hydroxyl lig-
ands, which ultimately breaks up the Au clusters. Second, it
−
displaces the residual Cl species, resulting in smaller Au
particle sizes. Thus magnesium citrate added during prepa-
ration plays an important role in reducing the Au particle
size.
3
.3. Selective CO oxidation reaction
Fig. 4 compares the rates of selective CO oxidation with
TOS for the various Au and Pt catalysts studied. A rapid ini-
tial partial deactivation was observed for the Au1, Au2, and

156
S.S. Pansare et al. / Journal of Catalysis 234 (2005) 151–160
Fig. 5. Selectivity (%) as a function of time-on-stream for Au1, Au2,
Au(Mg) and Pt catalysts.
Fig. 6. Turnover frequency comparison of Au and Pt catalysts as a function
of time-on-stream.
activation with loss of CO2 selectivity in the current study
occurred over a period of 1800 s. Hence the possibility of a
significant decrease in selectivity resulting from the forma-
tion of Mg carbonates can be ruled out, even if the process
was not totally efficient.
It is obvious that two phenomena were occurring in the
case of the Au(Mg) catalyst: one leading to partial deac-
tivation and the other, a type of induction phenomenon,
causing activation of new sites and/or reactivation of de-
activated sites. Together, these two phenomena yielded the
results shown in Figs. 4 and 5.
Fig. 6 compares the TOFs (determined based on pulse
H2 titration and static H2 chemisorption for the Au and Pt
catalysts, respectively) with the TOS of these catalysts. This
figure shows that the Pt catalyst had a much lower TOF than
all of the Au catalysts for all TOS. This suggests that the Pt
catalyst, while having the highest metal surface area/g, had a
much lower site activity or fraction of Pt surface atoms active
for reaction. It is especially important to note that all of the
Au catalysts had approximately similar initial TOF values
as well as similar initial rates (at 5 min TOS). These results
further confirm the similar concentrations of Aus obtained
from pulse H2 titration for all the Au catalysts, as shown in
Table 1.
Fig. 7. Intrinsic SSITKA site activity variation with time-on-stream for Au1,
Au2, Au(Mg) and Pt catalysts (maximum error measured, ± 10%).
Au(Mg) catalysts; Fig. 8b shows the variation with TOS for
the Pt catalyst. NI-CO is the concentration of intermediates
2
on the catalyst surface leading to CO2. There was a rapid de-
crease in the concentration of these intermediates during the
first 30 min TOS for all catalysts. The decreases for the Au
catalysts (around 50%) were significantly less than the de-
crease for the Pt catalyst (around 90%). Hence, considering
The activities and the deactivation behaviors of these
catalysts can be better explained by considering the con-
tributions to the overall activity from the intrinsic site ac-
the changes in k and NI-CO , we can conclude that for the Pt
2
−1
tivity (k, s ) and the concentration of surface intermedi-
catalyst, deactivation was due mainly to a decreased concen-
tration of reactive surface intermediates. In contrast, for the
Au1 and Au2 catalysts, there were decreases in both NI-CO2
and k, but the k decrease (around 70%) was the primary
ates (NI-CO ), both of which can be obtained by SSITKA.
2
Overall activity is the product of these two parameters
{
Rate = kNI-CO = (1/τI-CO )NI-CO } [23]. Fig. 7 shows the
2
2
2
variations in k with TOS for the Au and Pt catalysts. For
Au1 and Au2 catalysts, the reduction in k was significant,
whereas for the Pt catalyst, there was very little decrease
in k. The Au(Mg) catalyst exhibited a variation in k similar
to that observed for the rate.
cause of deactivation. The NI-CO decrease was probably due
2
mainly to deposition of carbon from CO on the active sites,
as has been shown previously for Pt [2]. This deposition
decreases adsorption of the reactants and hence increases
formation of reactive intermediates. We sent a sample of the
Au2 catalyst after it had reached steady state during reaction
to Galbraith Laboratories for carbon analysis; the amount
Fig. 8a shows the variations in the concentration of re-
active intermediates (NI-CO ) with TOS for Au1, Au2, and
2

S.S. Pansare et al. / Journal of Catalysis 234 (2005) 151–160
157
Fig. 9. Variations in surface coverage of reactive intermediates (θ
) for
I-CO
2
Au1, Au2, Au(Mg) and Pt catalysts (maximum error measured, ± 10%).
at 5 min TOS; the value for the Au2 catalyst was around
5
3
0%. The θ values of all of the catalysts dropped during first
0 min TOS, explaining the decreases in TOF and reaction
rate. The Au(Mg) catalyst showed an increase in (θI-CO2 )
after the initial drop, exhibiting a value of around 60% at
steady state. An important fact to remember here is that k
also increased with TOS after initial deactivation.
Five samples of the Au(Mg) catalyst (two at 30 min TOS
and other three after 130, 230 and 330 min TOS) were sent to
Galbraith for carbon analysis. The results showed a decrease
in the amount of carbon between the minimum activity and
steady state (2.18± 0.1% at 30 min vs 1.65% at steady state).
Although the decrease was not substantial, this could be one
of the reasons for the increased activity of the catalyst after
Fig. 8. Concentrations of reactive intermediates as
time-on-stream for (a) Au1, Au2 and Au(Mg) catalysts, (b) Pt catalyst
maximum error measured, ± 10%).
a function of
3
0 min TOS.
(
The selective CO oxidation reaction may be a structure-
sensitive reaction on Au catalysts with strong metal–support
interactions [6]. According to Haruta [6], the particle–
support perimeter interface plays a very important role in
the reaction. This interface is composed of Au oxide or hy-
droxide and is stabilized by interaction with metal oxide
supports, making it more active. The reaction is purported
to proceed at the perimeter interface with CO adsorbing
on the edge and corner sites of the Au nanoparticles while
oxygen in molecular form adsorbs at the support–metal in-
terface. The smaller the particle size, the more perimeter
per gm of Au is available for reaction, and hence the higher
the activity of the catalyst. The order of average metal par-
ticle diameters of the Au catalysts used in this study was
Au1 > Au2 > Au(Mg), and the activity of these catalysts at
steady state was the reverse: Au(Mg) > Au2 > Au1, which
is consistent with the foregoing discussion.
of carbon found by the combustion/colorimetric titration
method was around 0.5 wt%.
For the Au(Mg) catalyst, both NI-CO and k attained min-
2
imum values at 30 min TOS and then increased to their
steady-state values. The decrease in NI-CO2 was smaller
(
around 50%) than the decrease in k (around 83%), sug-
gesting that the decline in intrinsic site activity, not the de-
crease in the concentration of reactive intermediates, was
the major cause of deactivation. There was an increase in
the concentration of reactive intermediates during reactiva-
tion, as shown in Fig. 8a, but this alone is not sufficient to
explain the large increase in rate. The increase in intrinsic
site activity (k) after the initial rapid decrease was more sig-
nificant (Fig. 7), suggesting the possible creation of sites
of higher activity and/or reactivation of deactivated higher
activity sites during this activation/reactivation period. This
result was completely replicable.
There is increasing support in the literature for the fact
that moisture plays a very important role in enhancing the
activity of supported Au catalysts for CO oxidation [24–27]
and selective CO oxidation [10,28]. In the present study,
Another important parameter obtained from the SSITKA
data was the surface coverage (θI-CO ) of the most reactive
2
intermediates. Fig. 9 shows the variations in (θI-CO ) for all
the water formed from the oxidation of H during the re-
2
2
of the catalysts. The Au1, Au(Mg), and Pt catalysts showed
surface coverages of active intermediates of around 70%
action may have played an important role in increasing the
activity of the Au(Mg) catalyst with TOS. Fig. 10 shows

158
S.S. Pansare et al. / Journal of Catalysis 234 (2005) 151–160
towards the left, producing the original active bicarbonate in-
termediate species in place of a deactivated site. Thus active
sites may have been “regenerated” as the reaction proceeded,
resulting in increased activity of the catalyst. Hence regener-
ation could be another reason, along with reconstruction of
sites, for the increase in activity. Further research is needed
to address this issue in detail.
A question can be raised as to why the water produced
during the reaction reactivated only the Au(Mg) catalyst and
not the Au1 and Au2 catalysts. Daté et al. [25] recently
studied the effect of moisture on the activity of supported
Au catalysts for the CO oxidation reaction. According to
these authors, the moisture activates the oxygen and aids de-
composition of carbonate (as discussed earlier), resulting in
enhanced catalytic activity. Another observation from their
study is that for the Au/γ -Al2O3 catalyst, the enhancement
in the activity is significant only above 200 ppm of wa-
ter. The mean diameter of the supported Au nanoparticles
in that study was 3.9 nm, and the reaction temperature was
around 273 K. In the present study, for Au1, the concentra-
tion of water initially increased but then rapidly decreased to
a steady-state value of around 250 ppm (which is very close
to 200 ppm). Au2 had a higher water concentration than Au1
Fig. 10. Rate of water formation and concentration of water (ppm) as a
function of TOS for Au and Pt catalysts.
the rate of water formation during the selective CO oxi-
dation reaction on the Au and Pt catalysts as a function
of TOS. It can be speculated that the water formed on the
Au(Mg) catalyst is able to reactivate deactivated Au sites
and/or create additional sites through reconstruction of the
Au particles. This reconstruction may increase the perimeter
interface and/or change its structure, resulting in increased
activity. Grisel and Nieuwenhuys [10] have confirmed the
fact that for some Au catalysts, adding water has a positive
effect on the activity of the catalyst. They injected a 50-µl
pulse of water during the selective CO oxidation reaction and
found an increase in conversion on Au/MnOx/MgO/Al2O3,
Au/MgO/Al2O3, and Au/Al2O3 catalysts. In the present
case, the water formed during the reaction (which is signif-
icant after the initial deactivation period), not added water,
would have been the cause of the increased activity for the
Au(Mg) catalyst.
(
also > 200 ppm). It would seem highly likely that Au par-
−
ticle size and/or the presence of Cl species may play an
important role in determining the effect of water on cata-
lyst activity. The Au particles in Au1 and Au2 were 3.5 and
1
.7 times larger, respectively, than the particles in the study
of Daté et al. [25] and 10.7 and 5.2 times larger on aver-
age, respectively, than those of Au(Mg). Another noteworthy
parameter that may have an impact is the reaction tempera-
ture, which differs between the present study (363 K) and
the study of Daté et al. (273 K) [25].
It is important to consider the relationship of the SSITKA
parameters to the concentrations of the intermediates for the
Kung et al. [11] recently addressed the issue of the de-
activation mechanism of Au and the role of water in regen-
eration of the active sites. According to these authors, the
deactivation may result from formation of inactive carbonate
intermediates through dehydration of bicarbonate intermedi-
ates. The reaction occurring at the surface can be written as
selective CO oxidation reaction. NI-CO is related only to the
2
carbon intermediates leading to formation of CO2. The in-
trinsic site activity (k) value is related to the rate constant
in a pseudo-first-order power law rate expression and poten-
tially includes the partial pressure of O2. If new sites were
created that increased the amount of O2 able to adsorb, then
it would be likely that: (i) k would increase; (ii) more carbon
would be removed from the catalyst, possibly reactivating
sites; (iii) the number of CO reacting sites would increase;
and (iv) the rate of CO2 formation would increase.
To confirm the fact that H2O plays an important role in
the recovery of catalyst activity, several experiments were
carried out in the absence of H2 (non-selective CO oxida-
tion), with the other conditions kept the same as for selec-
tive CO oxidation. Fig. 11 shows the TOFs as a function of
TOS for the nonselective and selective CO oxidation on the
Au(Mg) catalyst. It was observed that in the absence of H2,
there was also an initial partial deactivation of the catalyst
during the first 30 min TOS; this occurred for the same rea-
son as for the other Au catalysts, that is, deposition of carbon
on active sites. But after that, the activity of the catalyst was
Au–(CO3H) + Al–OH ⇐⇒ Au–(CO3)–Al + H2Oads . (3)
The concentration of the reactive intermediates obtained
from SSITKA confirms that a decrease in active intermedi-
ates (presumably bicarbonates, in accordance with the fore-
going equation) occurred after 30 min TOS for the Au1
and Au2 catalysts. Both of these catalysts were not very ac-
tive, and hence water formed during the reaction was much
less (around 250 ppm for Au1 and 600 ppm for Au2, at
steady state). During the first 5 min TOS, the Au(Mg) cat-
alyst was very active, and the selectivity toward CO2 was
1
00% (Fig. 5); hence all of the O2 was used in producing
CO2. But as the reaction proceeded, the selectivity toward
CO2 dropped, and more water molecules were formed, as
shown in Fig. 10 (0 ppm at 5 min TOS vs around 900 ppm at
4
05 min TOS), which were adsorbed in part on the support
surface. As seen in Eq. (3), equilibrium would have shifted

S.S. Pansare et al. / Journal of Catalysis 234 (2005) 151–160
159
Fig. 11. Turnover frequency for non-selective and selective CO oxidation
on the Au(Mg) catalyst.
Fig. 12. Effect of the addition of magnesium citrate to the Au2 catalyst
Mg/Au2) on the rate of CO oxidation with TOS. For comparison, the rate
(
for the Au2 catalyst without Mg is also shown.
almost constant (around 0.5 s 1). No significant increase in
the TOF for non-selective CO oxidation on Au(Mg) was
observed, and the catalyst behaved like other Au catalysts
−
this is true, a Au catalyst was prepared by adding an aque-
ous solution of magnesium citrate to the Au2 catalyst using
the incipient wetness impregnation method. The catalyst was
dried for 1 h in static air and then calcined at 350 C for 4 h
under flowing dry air (HC free). This catalyst is referred to
as “Mg/Au2.”
The selective oxidation reaction was carried out on
Mg/Au2 in the same way as for the other catalysts. Fig. 12
shows the variations in the rate of reaction with TOS for the
Mg/Au2 and Au2 catalysts. The Mg/Au2 catalyst displayed
similar deactivation characteristics as the original Au2 cat-
alyst. No increase in the activity of Mg/Au2 was observed
after the initial partial deactivation, suggesting that Mg pro-
motion alone may not cause the increased catalyst activity.
Because the activity of the Mg/Au2 catalyst was similar
to that of Au2, it would appear that Mg does not have a
chemical-promoting effect on Au-catalyzed reaction, assum-
ing that there is no need for the Au and Mg to be contacted
in a specific way that is possible only during preparation.
(
Au1 and Au2 catalysts for selective CO oxidation from this
work).
Many studies of CO oxidation [12,24,25,27] have re-
◦
ported a regain of original activity after the introduction of
water vapor in the CO oxidation feed. These results also
support the fact that water may play an important role in re-
covering the catalytic activity during the selective oxidation
of CO in H2.
The steady-state value of TOF for nonselective CO ox-
idation was higher than that for selective CO oxidation.
There is evidence in the literature that the effect of H2
on Au-catalyzed CO oxidation activity can be either neg-
ative [10] or positive [29], depending on the reaction con-
ditions, method of catalyst preparation, and other factors.
◦
Clearly, in the present study, at 90 C, the CO oxidation
activity of a Au catalyst without significant amounts of chlo-
ride was greater than that for the selective oxidation of CO
in the presence of H2. In the absence of H2, only CO and O2
adsorb on the active sites. But with the addition of H2, there
is a competitive adsorption between H2 and CO and a com-
petition for reaction with O2. This may reduce the number of
active sites available for CO adsorption as compared to that
under non-selective CO oxidation.
4. Conclusions
Alumina-supported Au catalysts catalyze the selective
◦
oxidation of CO at 90 C with activities (TOF, site activity)
greater than that of Pt/γ -Al2O3. The important conclusions
3
.4. Role of magnesium in the activity of the catalyst
of this study can be summarized as follows:
The kinetic experiments showed that the Au(Mg) catalyst
– All of the Au and Pt catalysts studied showed high ini-
tial activities and selectivities, which decreased rapidly
during the first 30 min TOS. The activities of the Au
was more active than the Au1 and Au2 catalysts at steady
state because of reactivation/modification of the catalyst. Be-
cause the presence of Mg is the only difference between
Au1, Au2, and Au(Mg) other than Au particle size, the ques-
tion can be asked as to whether Mg is important only during
initial preparation as a structural promoter. Could it be pos-
sible that Mg acts as a chemical promoter for Au, much like
Fe for Pt, in this reaction [3]? To determine whether or not
−
catalysts with lower dispersions and higher Cl content
(Au1 and Au2) and the Pt catalyst remained constant af-
ter the initial significant partial deactivation.
−
– Presence or absence of Cl species had little effect on
the initial high activities and initial partial deactivation
of the Au catalysts.

160
S.S. Pansare et al. / Journal of Catalysis 234 (2005) 151–160
–
SSITKA data showed that the decline in the concentra-
References
tion of reactive intermediates (NI-CO ) and a drop in the
intrinsic site activity (k) were two major reasons for the
loss of activity for the Au1 and Au2 catalysts. For the Pt
2
[
[
[
1] O. Korotkikh, R. Farrauto, Catal. Today 62 (2000) 249.
2] A. Sirijaruphan, J.G. Goodwin Jr., R.W. Rice, J. Catal. 221 (2004) 288.
3] A. Sirijaruphan, J.G. Goodwin Jr., R.W. Rice, J. Catal. 224 (2004) 304.
catalyst, only a drop in NI-CO was the major cause of
2
[4] M.J. Kahlich, H.A. Gasteiger, R.J. Behm, J. Catal. 171 (1997) 93.
5] A. Manasilp, E. Gulari, Appl. Catal. B: Environ. 37 (2002) 17.
deactivation.
[
–
The Au(Mg) catalyst, which had much smaller Au parti-
cles and the least Cl content, showed reactivation with
[6] M. Haruta, Catal. Technol. 6 (2002) 102.
[7] G.K. Bethke, H.H. Kung, Appl. Catal. A: Gen. 194–195 (2000) 43.
−
[
[
8] M.J. Kahlich, H.A. Gasteiger, R.J. Behm, J. Catal. 182 (1999) 430.
9] J.-D. Grunwaldt, M. Maciejewski, O.S. Becker, P. Fabrizioli, A.
Baiker, J. Catal. 186 (1999) 458.
TOS after the initial rapid partial deactivation. It can be
speculated that this modification of the Au(Mg) catalyst
was due to activation/reactivation and regeneration of
active sites because of the water formed during the re-
action (which was significant at steady state compared
with that for the Au1 and Au2 catalysts), leading to in-
creased activity.
[10] R.J.H. Grisel, B.E. Nieuwenhuys, J. Catal. 199 (2001) 48.
[11] H.H. Kung, M.C. Kung, C.K. Costello, J. Catal. 216 (2003) 425.
12] C.K. Costello, M.C. Kung, H.-S. Oh, Y. Wang, H.H. Kung, Appl.
Catal. A: Gen. 232 (2002) 159.
13] A. Luengnaruemitchai, S. Osuwan, E. Gulari, Int. J. Hydrogen En-
erg. 29 (2004) 429.
[
[
–
Finally, it can be suggested that Mg, which was present
in the Au(Mg) catalyst, may not act as a chemical pro-
moter. Its main role may be only to help reduce Au
[14] H.-S. Oh, J.H. Yang, C.K. Costello, Y.M. Wang, S.R. Bare, H.H. Kung,
M.C. Kung, J. Catal. 210 (2002) 375.
[15] J. Happel, Chem. Eng. Sci. 33 (1978) 1567.
[
16] C.O. Bennett, Catalysis Under Transient Conditions, ACS Symposium
Series, American Chemical Society, Washington, DC, 1982, pp. 1–32.
17] P. Biloen, J. Mol. Catal. 21 (1983) 17.
−
particle size and Cl content during preparation.
[
The results also provide further insight into the impact of
[18] H. Berndt, I. Pitsch, K. Evert, K. Struve, M.-M. Pohl, J. Radnik,
A. Martin, Appl. Catal. A: Gen. 244 (2003) 169.
−
Au particle size and/or Cl content (difficult to separate be-
cause they are intertwined) on catalyst behavior. It is highly
likely that the presence of small Au particles and/or low
[
[
[
19] S.L. Shannon, J.G. Goodwin Jr., Chem. Rev. 95 (1995) 677.
20] S.H. Ali, J.G. Goodwin Jr., J. Catal. 171 (1997) 339.
21] A. Sirijaruphan, J.G. Goodwin Jr., R.W. Rice, J. Catal. 227 (2004) 547.
−
Cl content in a Au catalyst, produced in the presence of
[22] T. Fakushima, S. Galvagno, G. Parravano, J. Catal. 57 (1979) 177.
[23] J.G. Goodwin Jr., S.Y. Kim, W.D. Rhodes, Turnover Frequencies in
Metal Catalysis: Meanings, Functionalities and Relationships, Cataly-
sis, The Royal Society of Chemistry, 2003, pp. 1–29.
Mg citrate during preparation, will cause that catalyst to be
significantly affected by the formation of water, resulting in
reactivation of the catalyst as the rate of water formation in-
creases with TOS.
[
[
24] M. Daté, M. Haruta, J. Catal. 201 (2001) 221.
25] M. Daté, M. Okumura, S. Tsubota, M. Haruta, Angew. Chem. Int.
Ed. 43 (2004) 2129.
[26] S.T. Daniells, A.R. Overweg, M. Makkee, J.A. Moulijn, J. Catal. 230
(
2005) 52.
Acknowledgment
[27] J.T. Calla, R.J. Davis, J. Phys. Chem. B 109 (2005) 2307.
[28] M.M. Schubert, A. Venugopal, M.J. Kahlich, V. Plzak, R.J. Behm,
J. Catal. 222 (2004) 32.
This work was funded by US Department of Energy grant
DE-FG-26-01NT41277.
[29] D.A. Bulushev, L. Kiwi-Minsker, I. Yuranov, E.I. Suvorova, P.A. Buf-
fat, A. Renken, J. Catal. 210 (2002) 149.