Propene epoxidation over TiO2-supported Au-Cu alloy catalysts prepared from thiol-capped nanoparticles

Article abstract of DOI:10.1016/j.jcat.2008.06.010

Propene oxidation to propene oxide (PO) was performed with N₂O in the temperature range of 473-673 K using TiO₂-supported Au and Au-Cu alloy nanoparticles synthesized from pre-formed thiol-capped nanoparticles of controlled composition and size. Catalysts were activated by calcination at different temperatures in the range 573-873 K and characterized by HRTEM, XPS, and TPR. Among a series of catalysts with different Au/Cu ratio and metal loading, the Au₁Cu₃/TiO₂ with 1.2 wt% exhibited the best catalytic performance for epoxidation, both in terms of propene conversion rate and selectivity towards PO (0.25 mol PO g_M^-1 h^-1 at 573 K). The highest TOF was obtained over a catalyst calcined at 673 K. At this calcination temperature, HRTEM revealed a large perimeter interface between the nanoparticles and the support, which was accompanied by an intense TPR hydrogen uptake at low temperature. At increasing calcination temperature, the surface of Au-Cu alloy nanoparticles was progressively decorated with oxidized Cu species, which were detrimental for epoxidation and favored allylic oxidation products. Isolation effects and control of the extent of Cu oxidation by Au in the alloy nanoparticles as well as the perimeter interface between Au-Cu alloy nanoparticles and TiO₂ are imagined to play pivotal roles in the epoxidation of propene.

Full text of DOI:10.1016/j.jcat.2008.06.010

Journal of Catalysis 258 (2008) 187–198
Contents lists available at ScienceDirect
Journal of Catalysis
www.elsevier.com/locate/jcat
Propene epoxidation over TiO -supported Au–Cu alloy catalysts prepared from
2
thiol-capped nanoparticles
a,∗
a
a
b
b
Jordi Llorca , Montserrat Domínguez , Cristian Ledesma , Ricardo J. Chimentão , Francisco Medina ,
b
c
c
c
Jesús Sueiras , Inmaculada Angurell , Miquel Seco , Oriol Rossell
a
Institut de Tècniques Energètiques, Universitat Politècnica de Catalunya, Diagonal 647, ed. ETSEIB, 08028 Barcelona, Spain
Departament d’Enginyeria Química, Universitat Rovira i Virgili, 43007 Tarragona, Spain
Departament de Química Inorgánica, Universitat de Barcelona, Martí i Franquès 1-11, 08028 Barcelona, Spain
b
c
a r t i c l e i n f o
a b s t r a c t
Article history:
Propene oxidation to propene oxide (PO) was performed with N O in the temperature range of 473–673 K
2
Received 25 April 2008
Revised 5 June 2008
Accepted 5 June 2008
Available online 11 July 2008
using TiO2-supported Au and Au–Cu alloy nanoparticles synthesized from pre-formed thiol-capped
nanoparticles of controlled composition and size. Catalysts were activated by calcination at different
temperatures in the range 573–873 K and characterized by HRTEM, XPS, and TPR. Among a series of
catalysts with different Au/Cu ratio and metal loading, the Au1Cu3/TiO2 with 1.2 wt% exhibited the best
catalytic performance for epoxidation, both in terms of propene conversion rate and selectivity towards
Keywords:
−1
−1
PO (0.25 mol PO gM
h at 573 K). The highest TOF was obtained over a catalyst calcined at 673 K.
Propene epoxidation
Au–Cu alloy
Nanoparticles
HRTEM
At this calcination temperature, HRTEM revealed a large perimeter interface between the nanoparticles
and the support, which was accompanied by an intense TPR hydrogen uptake at low temperature. At
increasing calcination temperature, the surface of Au–Cu alloy nanoparticles was progressively decorated
with oxidized Cu species, which were detrimental for epoxidation and favored allylic oxidation products.
Isolation effects and control of the extent of Cu oxidation by Au in the alloy nanoparticles as well as the
perimeter interface between Au–Cu alloy nanoparticles and TiO2 are imagined to play pivotal roles in the
epoxidation of propene.
©
2008 Elsevier Inc. All rights reserved.
1
. Introduction
direct epoxidation of propene with O2/H2 mixtures has attracted
considerable industrial and academic interest [8–16]. The mode of
operation of Au–Ti–O catalysts for this reaction remains unclear,
but the role of the support, the Au particle size, and the periph-
ery between the support and the Au particles have been recog-
nized [11,16,17]. In particular, only the anatase form of TiO2 leads
to the formation of propene oxide, whereas the rutile form and
amorphous TiO2 cause the complete oxidation of propene into CO2
When dispersed as fine particles of less than ∼10 nm in di-
mension over selected metal oxides, gold exhibits exceptionally
high activity in a variety of reactions such as CO oxidation, hy-
drogenation of CO and CO2, water gas shift, reduction of NO to
N2, hydrogenation of unsaturated substrates, oxidation of alco-
hols and aldehydes, hydrochlorination of ethyne, and epoxidation
of propene, among others [1–6]. Current industrial methods used
for the epoxidation of propene, the chlorohydrin and organic hy-
droperoxide processes, have important disadvantages in terms of
environmental pollution and by-products. However, propene ox-
ide is one of the most important feedstocks for the production
of many useful chemicals, such as polyurethane, polyester resins,
and surfactants [7], and a process based on the direct epoxidation
of propene in the gas phase using only oxygen or any other oxi-
dant, resulting in benign side products, is highly desirable. In this
context, the high catalytic activity of gold supported on TiO2 and
titanium-containing materials (such as Ti-MCM-41 and TS-1) in the
[18]. As regards the Au particle size, there is a strong particle size
dependence and only small nanoparticles (1–5 nm) are selective
towards the formation of propene oxide, whereas larger Au parti-
cles increase the selectivity to combustion [9,16,19,20]. However,
the high reactivity of surface-activated oxygen species generated
from O2 often leads to consecutive oxidation and, consequently, to
a loss of propene oxide selectivity. For this reason, recent attention
has been given to the use of nitrous oxide as a mild oxidant in the
propene epoxidation reaction [21–29]. Usually N2O is activated at
metal sites, leading to N2 and atomically adsorbed oxygen with a
mild electrophilic character, suitable for the transformation of the
vinyl group of propene into an oxirane ring [21,26]. In addition,
N2O is a greenhouse warming gas which is produced abundantly
in industrial processes (i.e. in the production of precursors of ny-
lon and of nitric acid) and its use in propene epoxidation, with
Corresponding author. Fax: +34 39 401 71 49.
E-mail address: jordi.llorca@upc.edu (J. Llorca).
*
0
021-9517/$ – see front matter © 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2008.06.010

188
J. Llorca et al. / Journal of Catalysis 258 (2008) 187–198
the only byproduct N2, could offer an interesting solution to the
environmental problems associated with this gas [28]:
solvent was evaporated slowly overnight and the samples were cal-
cined at 573, 673, 773, or 873 K for 2 h. No organic residues were
detected by thermogravimetric analysis, infrared spectroscopy, and
XPS after calcination at any of these temperatures. All samples at-
tained a pale pink color typical of the presence of gold nanoparti-
cles over TiO2. Au and Cu contents were measured by wet methods
by inductively coupled plasma.
CH3CH=CH2 + N2O → CH3 CH–CH2 + N2.
O
In addition to nanosized Au, it is well known that Cu-based cata-
lysts are particularly effective for alkene epoxidation when allylic
hydrogen atoms are present because Cu favors metallacycle forma-
tion instead of allylic hydrogen stripping [30,31], which is the main
obstacle for epoxidation. Several examples have been reported in
the literature concerning direct propene epoxidation with O2 [32–
2.2. Characterization of catalysts
High resolution transmission electron microscopy (HRTEM) was
35]. However, at sufficiently high oxygen coverage, nucleation and
carried out using a JEOL JEM 2010F electron microscope equipped
with a field emission source at an accelerating voltage of 300 kV.
Electron energy-loss spectroscopy (EELS) was performed in STEM
mode with a Gatan detector. In the case of thiol-capped solutions,
the sol was directly dropped onto carbon-coated grids. For Au–
Cu/TiO2 catalysts, powders were suspended in methanol for about
growth of copper oxide occurs, which inhibits epoxidation [30].
One way to control the extent of Cu oxidation under reaction con-
ditions involves the use of alloys. It has been reported recently that
Au–Cu/TiO2 bimetallic catalysts prepared by impregnation methods
perform better than Au/TiO2 and Cu/TiO2 for propene epoxidation
[36]. However, precise control of particle size was not achieved due
1
min under ultrasonic treatment before they were deposited on
to the preparation route, and particles in the range 2–20 nm were
encountered, which prevented accurate interpretation of the alloy-
ing effect, taking into account the dependence of propene oxide
selectivity with particle size.
holey carbon-coated grids. The microscope was calibrated at differ-
ent magnifications before and after measurements using appropri-
ate standards. The point-to-point resolution achieved was 0.19 nm
and the resolution between lines was 0.14 nm. A minimum of 250
particles were measured in each sample for particle size determi-
nation. The size limit for the detection of nanoparticles on the
support was about 1 nm. The average particle diameter was cal-
culated from the mean diameter frequency distribution with the
Several approaches have been used to prepare gold nanoparti-
cles on metal oxide supports including incipient wetness impreg-
nation, co-precipitation, deposition–precipitation, ion exchange,
gas-phase grafting, co-sputtering, organic capping, and dendrimer
and micelle encapsulation [1,2,9,37–45]. Impregnation and precip-
itation methods are very simple and scalable, but normally suffer
from precise particle size control. In contrast, organic capping and
encapsulation methods produce size-controlled gold nanoparti-
cles whose particle size is established before deposition on the
metal oxide support. Moreover, the coordinating ligands in the
precursor solution and on the oxide surface prevent aggregation
of the nanoparticles [12,41,46–48]. In this work, we have pre-
pared a variety of well-defined Au and Au–Cu alloy nanoparticles
with different composition and size supported on anatase from
dodecanethiol-capped nanoparticles. These have been character-
ized by high resolution transmission electron microscopy (HRTEM),
X-ray photoelectron spectroscopy (XPS), and temperature pro-
grammed reduction (TPR), and tested in the direct epoxidation of
propene by N2O. We have attempted to study the role of alloy
composition as well as the contact structure between TiO2 and the
alloy nanoparticles in the propene epoxidation reaction by study-
ing the effect of calcination temperature, particle size, and surface
composition.
ꢀ
ꢀ
formula: d =
nidi/
ni, where ni is the number of particles
with particle diameter di in a certain range. For surface analy-
sis, X-ray photoelectron spectroscopy (XPS) was performed with
a Perkin–Elmer PHI-5500 instrument equipped with an Al X-ray
source operated at 12.4 kV and a hemispherical electron analyzer.
Temperature programmed reduction (TPR) experiments were per-
formed with a Thermo Finnigan TPDRO 1100 apparatus equipped
with a thermal conductivity detector and coupled to a Omnistar
−1
QMS 422 mass spectrometer. A 20 mL min
gas flow of 5% H2 in
Ar was used as a reducing agent and the temperature was raised
−1
from 323 to 1073 K at a constant rate of 20 K min
.3. Catalytic tests
Propene epoxidation was carried out in a fixed-bed reactor
.
2
at atmospheric pressure. Experiments were carried out typically
with 0.15 g of catalyst and a total flow rate between 30 and
−1
9
0 mL min . The gas mixture consisted of 10% propene and 10%
N2O balanced in Ar. The reaction was monitored at increasing
temperature from 473 to 673 K. Reaction products were analyzed
on-line with two gas chromatographs equipped with a molecular
sieve and Porapak T columns using TCD and FID detectors. Trans-
fer line and valves were maintained at 373 K in order to prevent
condensation of products.
2. Experimental
2
.1. Preparation of Au–Cu/TiO2 catalysts
Bimetallic Au–Cu nanoparticles with Au:Cu molar ratios of 3:1,
:1, and 1:3 as well as monometallic Au nanoparticles encap-
1
sulated with dodecanethiol monolayer shells were synthesized
following the two-phase method described for the synthesis of
dodecanethiol-capped monometallic Au nanoparticles [49–51] and
3. Results and discussion
3.1. Characterization of catalysts
−
2+
species were first trans-
metal alloy clusters [52]. AuCl4 and Cu
ferred from aqueous HAuCl4 and Cu(NO3)2 solutions (30 mM) to
toluene solution using tetraoctylammonium bromide as a phase
transfer reagent. Dodecanethiol was then added to the solution at
a dodecanethiol:(Au + Cu) = 3:2 mole ratio, and an excess of aque-
ous NaBH4 was slowly added to reduce the metal salts. Bimetallic
particles of different sizes were obtained by changing the concen-
tration of metal salt solutions from 30 to 15, 6, and 3 mM. The
resulting dodecanethiol-capped metallic nanoparticles were sub-
jected to solvent removal and cleaned using ethanol. The nanopar-
ticles were then dissolved in toluene and impregnated onto com-
Table 1 compiles the composition of catalysts investigated in
this work. Two series of samples were prepared. In one of them,
the total metal loading was maintained at ca. 1.2 wt% while the
ratio Au:Cu was varied from 1:0 to 1:3 on a molar basis. In the
other series, the nominal ratio Au:Cu was fixed at 1:3 and the
total metal loading was varied from ca. 0.5 to 1.2 wt%. All these
catalysts were characterized in detail by transmission electron mi-
croscopy techniques and XPS after calcination at different tempera-
tures. The corresponding precursor nanoparticles were also studied
by HRTEM. Details of particle size distribution of these samples are
summarized in Table 2.
2
−1
mercial anatase (Sigma–Aldrich, 99.8+%, SBET = 10 m g ). The

J. Llorca et al. / Journal of Catalysis 258 (2008) 187–198
189
Table 1
Composition of catalysts calcined at 673 K and calculated alloy composition using
d111 and d200 lattice spacings determined by HRTEM
Catalyst
Wt%
M total
Wt%
Au
Wt%
Cu
Au/Cu
atomic
Alloy
composition
Au/TiO2
Au3Cu1/TiO2
Au1Cu1/TiO2
Au1Cu3/TiO2
Au1Cu3/TiO2 (1%)
Au1Cu3/TiO2 (0.8%)
Au1Cu3/TiO2 (0.5%)
1.21
1.23
1.17
1.14
0.98
0.81
0.52
1.21
1.12
0.93
0.61
0.49
0.42
0.25
0
–
Au100
0.11
0.24
0.53
0.49
0.39
0.27
3.29
1.25
0.37
0.32
0.35
0.30
Au77Cu23
Au56Cu44
Au27Cu73
Au24Cu76
Au26Cu74
Au23Cu77
3
.1.1. HRTEM of nanoparticles as prepared
Fig. 1 shows a representative TEM image of nanoparticles as
prepared. Homogeneous and well-dispersed particles were ob-
tained in all cases with narrow size distributions (Fig. 1a and Ta-
ble 2). It is interesting to note that, at the same metal loading, the
incorporation of Cu resulted in slightly larger particles with respect
to pure Au clusters (2.2–2.3 nm and 1.8 nm, respectively). Further-
more, nanoparticles obtained under increasingly diluted conditions
exhibited progressively smaller size and nanoparticles of about
1.6 nm were obtained with a total metal loading of ca. 0.5% by
weight. HRTEM images of individual nanoparticles as prepared are
shown in Figs. 1b and 1c. Irrespectively of the Au:Cu ratio and size
of nanoparticles, all of them showed lattice fringes corresponding
to single face-centered cubic (fcc) phases. Lattice constant val-
ues were comprised between those of pure bulk phases of gold
(
aAu = 0.4079 nm) and copper (aCu = 0.3615 nm), indicating that
AuxCu1−x intermetallic alloys were formed.
3
3
.1.2. Au/TiO2 and Au–Cu/TiO2 catalysts calcined at 673 K
.1.2.1. HRTEM After deposition on TiO2 and calcination at 673 K,
nanoparticles increased in size, as expected, but a narrow size dis-
tribution in each case was maintained (Table 2). Particle growth
showed a strong dependence on metal loading. The mean particle
size measured in catalysts with a total metal loading of ca. 1.2%
increased about 2.5–2.8 times after calcination at 673 K with re-
spect to the size of nanoparticles as prepared, irrespective of the
Au:Cu ratio. In contrast, this increase was progressively lower in
those samples with lower metal content, from 2.6 and 2.5 in
Au1Cu3/TiO2 (1.2 and 1%) to 1.9 in Au1Cu3/TiO2 (0.8%) and 1.3 in
Au1Cu3/TiO2 (0.5%). This resulted in TiO2-supported Au–Cu alloy
nanoparticles with the same composition, but with different par-
ticle size. Fig. 2 shows low-magnification HRTEM images of the
different Au1Cu3/TiO2 catalysts prepared with different metal load-
ings after calcination at 673 K. A range in particle size from 5.7 nm
Fig. 1. (a) Bright-field TEM image of dodecanethiol-capped nanoparticles as pre-
pared. (b, c) HRTEM images of Au Cu3 alloy precursor nanoparticles of catalysts
1
Au1Cu3/TiO2 (0.5%) and Au1Cu3/TiO2 (1.2%), respectively. Lattice fringes at about
2
.2 Å correspond to fcc Au–Cu alloy (111) crystallographic planes.
Detailed EELS and HRTEM investigations were carried out in
order to fully characterize the Au–Cu alloy nanoparticles present
in each catalyst. In all cases, EEL spectra recorded over individ-
ual nanoparticles in bimetallic catalysts showed the presence of a
strong signal located at 24 eV, which corresponds to the Au–Cu
alloy [36]. As an example, Fig. 3 shows Z-contrast STEM images in
bright and dark field modes (Figs. 3a and 3b) corresponding to cat-
alyst Au1Cu1/TiO2 and EEL spectra obtained at different locations
(Figs. 3c–3f). The EEL spectrum depicted in Fig. 3c was recorded
over the TiO2 support, as corroborated by energy electron-loss sig-
nals at about 5, 12, 25, 39, and 47 eV, whereas EEL spectra in
Figs. 3d–3f were recorded over individual nanoparticles labeled
according to Fig. 3b. In order to gain further insight into the na-
ture of the Au–Cu alloys present in the different catalysts, HRTEM
(
Au1Cu3/TiO2, 1.2%) to 2.1 nm (Au1Cu3/TiO2, 0.5%) was obtained. In
contrast, virtually the same particle size distribution (5.6–5.8 nm)
was encountered in the bimetallic catalysts with the same total
metal loading (1.2%) and different Au:Cu ratios after calcination
at 673 K (Table 2), thus providing a suite of valuable samples for
comparison of propene epoxidation.
Table 2
Metal particle size distribution determined by HRTEM of nanoparticles as prepared and after deposition on TiO2. Values in parenthesis indicate the range of particle size
corresponding to 95% of all particles. All values in nm
Catalyst
Precursor
nanoparticles
Calcination temperature (K)
73
5
673
773
873
Au/TiO2
Au3Cu1/TiO2
Au1Cu1/TiO2
Au1Cu3/TiO2
Au1Cu3/TiO2 (1%)
Au1Cu3/TiO2 (0.8%)
Au1Cu3/TiO2 (0.5%)
1.8 (1.5–2.1)
2.2 (1.7–2.5)
2.3 (2.1–2.6)
2.2 (1.6–2.4)
2.0 (1.7–2.2)
1.7 (1.4–1.9)
1.6 (1.4–1.8)
–
–
–
5.1 (4.1–6.3)
5.6 (4.4–6.5)
5.8 (4.5–6.6)
5.7 (4.4–6.8)
4.9 (3.6–6.0)
3.3 (2.0–3.6)
2.1 (1.6–2.8)
–
–
–
–
–
–
5.3 (4.1–6.2)
4.2 (3.3–5.8)
3.0 (2.4–3.5)
2.0 (1.7–2.4)
7.6 (5.2–11.3)
6.8 (4.7–10.9)
4.5 (3.4–8.8)
3.1 (2.6–8.0)
9.8 (6.7–14.5)
9.2 (6.3–15.0)
7.0 (5.5–12.4)
5.8 (5.1–9.8)

190
J. Llorca et al. / Journal of Catalysis 258 (2008) 187–198
Fig. 4. HRTEM and FT images of catalysts Au3Cu1/TiO2 (a) and Au1Cu3/TiO2 (b)
calcined at 673 K. Au–Cu alloy nanoparticles are oriented along the [110] crystal-
lographic direction. Lattice fringes correspond to (111) and (200) planes at 2.17–2.31
and 1.88–1.99 Å, respectively.
Fig. 2. HRTEM images of Au1Cu3 nanoparticles supported on TiO2 and calcined at
73 K. (a) Au1Cu3/TiO2 (1.2%), (b) Au1Cu3/TiO2 (1.0%), (c) Au1Cu3/TiO2 (0.8%), and
d) Au1Cu3/TiO2 (0.5%).
6
(
Fig. 5. Lattice parameters of fcc Au–Cu alloy nanoparticles of different composition
supported on TiO2 calculated from HRTEM (111) and (200) lattice fringe values. The
dashed line corresponds to Vegard’s law.
spectively. In Fig. 4b, spots at 2.30–2.31 and 1.99 Å correspond
to fcc (111) and (200) planes of an Au–Cu alloy nanoparticle of
catalyst Au3Cu1/TiO2. HRTEM images indicated that most particles
in all samples were monocrystalline and round-shaped (Fig. 4b),
whereas well-faceted particles were rarely encountered (Fig. 4a).
We noted that prolonged exposure of particles under the electron
beam caused structural changes, in particular, round-shaped parti-
cles became well-faceted. In order to minimize the contribution of
such beam-driven effects, HRTEM images were acquired as fast as
possible. Similarly to what was observed for the Au–Cu nanopar-
ticles as prepared, TiO2-supported alloy nanoparticles showed fcc
lattice constant values intermediate to those of pure Au and Cu,
indicating that the alloy structure was maintained after deposition
and calcination. It is noteworthy that by this preparation route no
tetragonally ordered Au–Cu phases occur, as commonly observed
in Au–Cu alloys prepared by sequential implantation [53] or TiO2-
supported Au–Cu alloys prepared by conventional impregnation
methods [36].
Fig. 3. Bright-field (a) and dark-field (b) STEM images of catalyst Au1Cu1/TiO2 cal-
cined at 673 K and EEL spectra recorded over areas labeled (c)–(f).
For each catalyst, a set of d111 and d200 values were compiled
from all HRTEM images and the lattice constant for a cubic fcc
structure was calculated. The values obtained and the correspond-
ing error bars are plotted in Fig. 5 with respect to catalyst com-
position. Vegard’s law (aalloy = xaAu + (1 − x)aCu + 0.01198x(1 − x))
images were obtained for nearly 30 individual nanoparticles for
each sample and analyzed by Fourier transform (FT) methods. As
an example, Fig. 4 shows HRTEM and FT images obtained from
different catalysts. Fig. 4a corresponds to catalyst Au1Cu3/TiO2.
Lattice fringes of the anatase support are visible at 3.52 Å, and
two Au–Cu alloy particles exhibit spacings at 2.17–2.18 and 1.88 Å
corresponding to fcc (111) and (200) crystallographic planes, re-
[
54] has also been included in the plot for comparison (dashed
line). Lattice constant values decreased as the Cu content in-
creased, according to the formation of solid solutions. The calcu-
lated alloy compositions from HRTEM measurements are included

J. Llorca et al. / Journal of Catalysis 258 (2008) 187–198
191
Table 3
Binding energies and surface atomic ratios determined by XPS over catalysts cal-
cined at 673 K
Catalyst
BE Au 4f7/2
eV)
BE Cu 2p3/2
(eV)
Au/Ti
Cu/Ti
Au/Cu
(
Au/TiO2
Au3Cu1/TiO2
Au1Cu1/TiO2
Au1Cu3/TiO2
Au1Cu3/TiO2 (1%)
Au1Cu3/TiO2 (0.8%)
Au1Cu3/TiO2 (0.5%)
83.7
83.6
83.7
83.5
83.5
83.3
83.2
–
0.049
0.033
0.023
0.009
0.009
0.007
0.004
0.000
0.012
0.030
0.050
0.070
0.075
0.058
–
932.3
932.5
932.5
932.7
932.8
932.7
2.793
0.763
0.183
0.131
0.097
0.065
in Table 1. Two additional trends merit highlighting. First, the
higher the copper content of the catalysts, the larger the error in
the measurement of the lattice constant. Second, the higher the
copper content of the catalyst, the larger the deviation from Veg-
ard’s law. These trends indicate that the structure of nanoparticles
is increasingly deficient in Cu as the total Cu content of the cat-
alysts increases. This could imply that part of the copper atoms
do not alloy, but remain as small clusters that escape HRTEM de-
tection over TiO2 or are segregated on the surface of the alloy
particles. Deviations from Vegard’s law have also been reported in
other Au–Cu systems [36,53,54].
Fig. 6. Comparison of bulk and XPS atomic Au/Cu ratios recorded over catalysts with
different composition calcined at 673 K. The dashed line indicates the same Au/Cu
ratio for both axes.
3
.1.2.2. XPS X-ray photoelectron spectroscopy was used to inves-
tigate the oxidation states of Au and Cu and the elemental surface
composition of alloy nanoparticles following synthesis and after
deposition on anatase and calcination. No significant differences in
elemental Au/Cu ratios and oxidation states were detected before
or after deposition of nanoparticles on TiO2 and calcination at the
lowest temperature, except for the presence of sulfur, which was
observed by XPS only before calcination. Table 3 compiles binding
energy values and surface elemental ratios obtained over catalysts
prepared with different bulk Au:Cu ratios and metal loading after
calcination at 673 K. For the monometallic Au/TiO2 sample, two
sharp peaks at 87.2 and 83.7 eV were recorded, which are charac-
teristic of spin–orbit splitting (Au 4f5/2 and 4f7/2, respectively) of
0
Au core level spectra consistent with the Au oxidation state [12,
4
9,55]. The Au 4f7/2 binding energies recorded for the bimetallic
catalysts with different bulk Au:Cu ratios, but with the same total
metal loading (ca. 1.2 wt%) were virtually identical (83.5–83.7 eV),
indicating that Au was in the reduced oxidation state, too. How-
ever, a slight decrease in the Au 4f7/2 binding energy was obtained
over catalysts Au1Cu3/TiO2 with decreasing metal loading, and val-
ues as low as 83.2 eV were obtained (Table 3). Taking into account
the fact that Au is more electronegative than Cu, this could be
explained by the existence of electron density transfer from cop-
per to gold, which would result in the shifting of Au core levels
towards lower binding energies [36]. This is supported by a si-
multaneously slight increase in Cu 2p binding energy values, from
Fig. 7. Bright-field conventional TEM images of catalyst Au1Cu3/TiO2 calcined at
573 (a), 673 (b), 773 (c), and 873 K (d).
strong Cu segregation on the surface as the metal loading of the
sample decreases. Therefore, surface enrichment in copper progres-
sively increases as the particle size decreases.
932.5 to 932.8 eV for the Cu 2p3/2 signal (Table 3). As regards sur-
face atomic ratios, there is a good correlation between Au/Ti and
Cu/Ti values and the respective Au and Cu contents of catalysts.
However, there is a substantial enrichment of copper at the sur-
face of catalysts with respect to bulk values. This is evident from
the Au/Cu elemental surface ratios reported in Table 3. A similar
segregation of copper at the surface has been reported for air-
exposed silica supported gold-copper clusters of about 6–7 nm
3
3
.1.3. Catalysts calcined at different temperatures
.1.3.1. HRTEM Fig. 7 shows low-magnification, bright-field con-
ventional TEM images obtained at the same magnification of cat-
alyst Au Cu /TiO (1.2%) calcined at different temperatures. It is
1
3
2
evident that the calcination temperature had a strong effect on
the nanoparticle size, which increased from 5.3 nm in the sample
calcined at 573 K up to 9.8 nm in the sample calcined at 873 K.
A more pronounced effect was observed for catalysts Au Cu /TiO
[56]. In Fig. 6 the Au/Cu XPS ratios are plotted against the Au/Cu
bulk ratios for the different catalysts (Table 1). Surface enrichment
in copper increases as the gold content of the sample decreases
1
3
2
(
solid line). This is in accordance to HRTEM studies and lattice
with lower metal loading, where the relative increase of parti-
cle size was progressively larger as the metal loading decreased
(Table 2), probably due to the corresponding smaller starting par-
ticle size. Another important effect of calcination temperature con-
constant calculations reported above, where the deviation from Ve-
gard’s law increased as the Au content of catalysts decreased. On
the other hand, for a given composition (Au1Cu3/TiO2), there is a

192
J. Llorca et al. / Journal of Catalysis 258 (2008) 187–198
Fig. 8. HRTEM and FT images of catalyst Au1Cu3/TiO2 calcined at 573 K.
Fig. 10. HRTEM and FT images of catalyst Au1Cu3/TiO2 calcined at 773 K.
Fig. 9. HRTEM and FT images of catalyst Au1Cu3/TiO2 calcined at 673 K.
cerned the morphology of the alloy nanoparticles and the con-
tact structure between nanoparticles and the TiO2 support. Alloy
nanoparticles in catalysts calcined at 573–773 K exhibited spher-
ical or hemispherical morphologies (Figs. 7a–7c), whereas alloy
particles in the sample calcined at 873 K showed strong flatten-
ing and aggregation (Fig. 7d). Finally, the surfaces of TiO2 crystals
showed additional roughness in catalysts calcined at 673 and 773 K
Fig. 11. HRTEM and FT images of catalyst Au1Cu3/TiO2 calcined at 873 K.
area has increased considerably with respect to samples calcined at
lower temperatures and the diameter of the contact area is about
5 nm, while the mean particle size is centered at 7.6 nm. As a
consequence of the stronger alloy–support interaction, the contact
◦
angle between the alloy nanoparticles and TiO2 is greater than 90 .
(
Figs. 7b and 7c). From Fig. 7b it is evident that the roughness
Lattice fringes of the alloy nanoparticles have increased slightly,
2.19–2.20 for fcc (111) planes, which may be indicative of an Au
enrichment of the bulk of the alloys and Cu segregation onto the
surface after calcination at 773 K. This trend is confirmed by the
HRTEM images recorded over the catalyst calcined at the high-
est temperature, 873 K, where (111) lattice fringes up to 2.21 Å
are encountered (Fig. 11). At this calcination temperature, how-
ever, the most outstanding effects are the increase in particle size
and the flattened pattern exhibited by the alloy particles. Parti-
cles with axes larger than 10 nm are common, with contact areas
of about 8 nm in diameter. All particles imaged by HRTEM were
single-domain particles, indicating that the strong coalescence of
alloy nanoparticles observed after calcination at 873 K resulted in
the global rearrangement of the atoms in the structure. The FT im-
age depicted in Fig. 11b even suggests the possibility of an epitaxial
relationship between (111) planes at 2.21 Å of the Au–Cu alloy and
(101) planes at 3.52 Å of TiO2 support.
originated from the contact between the alloy nanoparticles and
the anatase support, which is an indication of the existence of a
strong metal–support interaction.
The interaction between alloy nanoparticles and the support as
well as the alloy structure was investigated for each of these cat-
alysts by HRTEM. Fig. 8 corresponds to the catalyst calcined at
the lowest temperature, 573 K. Round-shaped particles of about
5
2
.3 nm are well dispersed over the TiO2 support. Lattice fringes at
.18–2.19 and 1.85–1.88 Å correspond to fcc (111) and (200) planes
of Au–Cu, respectively, in accordance with values reported for the
alloy nanoparticles as synthesized. At this calcination temperature,
the physical interaction between alloy nanoparticles and TiO2 is
minimum, and the contact area has a diameter of about 1.6 nm.
The contact angle between nanoparticles and the support is usually
◦
<
45 . Fig. 9 shows profile views of different alloy nanoparticles
corresponding to the catalyst calcined at 673 K. Particles have in-
creased their size slightly, from 5.3 to 5.7 nm, but now a firm
contact between the alloy nanoparticles and the anatase support
exists. The diameter of the contact area has notably increased to
3.1.3.2. XPS X-ray photoelectron spectroscopy was carried out
over catalyst Au Cu /TiO (metal loading 1.2 wt%) calcined at
1
3
2
◦
ca. 2.6 nm and the contact angle is about 60–90 . It is interest-
different temperatures. The corresponding binding energies and
elemental surface ratios are reported in Table 4. No differences
in oxidation states were encountered in samples calcined at 573,
673 and 773 K, and binding energies for Au (83.5–83.6 eV) and
Cu (932.5–932.8 eV) corresponded well to those of the reduced
metals. In contrast, the binding energies recorded for the sample
calcined at 873 K shifted towards higher values, which is a mani-
festation of charging during calcination. The binding energy value
ing to note in Fig. 9b that the contact region between the support
and the alloy does not exhibit lattice fringes, whereas these are
clearly seen both in the alloy nanoparticle as well as in the sup-
port, suggesting that the interface between them is poorly ordered
at the atomic level due to a strong alloy–support interaction. The
alloy nanoparticles are perfectly crystalline, as depicted in the en-
larged area of Fig. 9a. From the spots in the FT images it is deduced
that the lattice parameter of the Au–Cu alloy remains unchanged
after calcination at 673 K. Fig. 10 corresponds to the catalyst cal-
cined at 773 K. At this temperature the perimeter of the contact
of the Au 4f7 peak at 84.1 eV is intermediate between Au0 and
/2
+
Au species, but the binding energy of Cu 2p3/2 at 933.3 eV is
close to oxidized copper [57], thus indicating that significant cop-

J. Llorca et al. / Journal of Catalysis 258 (2008) 187–198
193
Table 4
Binding energies and surface atomic ratios determined by XPS over catalyst
Au1Cu3/TiO2 (1.2%) calcined at different temperature
Calcination
temperature (K)
BE Au 4f7/2
(eV)
BE Cu 2p3/2
(eV)
Au/Ti
Cu/Ti
Au/Cu
5
73
83.5
83.5
83.6
84.1
932.7
932.5
932.8
933.3
0.013
0.009
0.008
0.006
0.066
0.050
0.046
0.055
0.191
0.183
0.174
0.105
673
773
873
Fig. 13. TPR profiles of catalyst Au1Cu3/TiO2 (1.0%) calcined at 573 (a), and 673 K
(b).
534 K (Fig. 12b). The general shape of the TPR profiles and the
total hydrogen consumption recorded over catalyst Au1Cu3/TiO2
−1
calcined at 573 and 673 K are similar (0.21 vs 0.23 mmol H2 g
,
respectively). However, the substantial increase of hydrogen con-
sumption at low temperature merits pointing out. HRTEM and
XPS data reported above showed that the increase in calcination
temperature from 573 to 673 K resulted in discrete nanoparticle
growth (from 5.3 to 5.7 nm), but in a strong alloy–support inter-
action. The remarkable intensity increase of the TPR peak at 445 K
is thus likely to be explained in terms of better reducibility of
the contact area between the alloy nanoparticles and TiO2 due
to a strong alloy–support interaction. The appearance of a low-
temperature hydrogen consumption signal was also observed in
the TPR profiles of catalysts Au1Cu3/TiO2 with lower total metal
loading calcined at this temperature. Fig. 13 shows the TPR profiles
recorded over catalyst Au1Cu3/TiO2 (1.0% metal loading) calcined at
Fig. 12. TPR profiles of catalyst Au1Cu3/TiO2 (1.2%) calcined at 573 (a), 673 (b),
73 (c), and 873 K (d).
7
per oxidation occurred at the calcination temperature of 873 K.
Besides the charging effect, other phenomena took place at in-
creasing calcination temperature, which are clearly visible in the
relative Au/Cu ratios of catalysts reported in Table 4. A progressive
segregation of copper on the catalyst surface was observed as the
calcination temperature increased, particularly in the sample cal-
cined at 873 K. This is in accordance with HRTEM characterization
results, where the lattice parameters obtained after calcination at
573 and 673 K. Following calcination at 573 K, hydrogen consump-
tion occurred at 530, 556, and 630 K (Fig. 13a), whereas calcination
at 673 K resulted in a new, well-resolved hydrogen uptake signal at
773 and 873 K indicated a progressive bulk enrichment of the alloy
in Au.
5
00 K. Hydrogen uptake for these samples were in the range 0.16–
−1
0.20 mmol H2 g , much higher than the theoretical value if all
3
.1.3.3. TPR Temperature-programmed reduction profiles obtained
−1
metal was oxidized (0.11 mmol H2 g ). Given the smaller size of
nanoparticles in sample Au1Cu3/TiO2 (1.0%) with respect to sample
Au1Cu3/TiO2 (1.2%), this large difference between the theoretical
hydrogen uptake and the measured values can be explained by a
large perimeter interface between the alloy nanoparticles and the
support.
from catalyst Au1Cu3/TiO2 (1.2% metal loading) calcined at differ-
ent temperatures are shown in Fig. 12. The TPR profile recorded
over the sample calcined at 573 K (Fig. 12a) showed a maximum
hydrogen consumption peak at 523 K and minor consumption
peaks at about 450, 585, and 645 K. The TPR profile of monometal-
lic Cu/TiO2 usually presents two reduction peaks centered at about
TPR profiles of sample Au1Cu3/TiO2 (1.2% metal loading) ob-
tained after calcination at 773 and 873 K (Figs. 12c and 12d)
exhibit similar patterns, but different to those reported at lower
calcination temperature, with hydrogen consumption peaks located
at about 530–535 and 575–590 K, and an additional peak at 652 K
for the sample calcined at 873 K. The most remarkable difference
between samples calcined at 673 and 773 K deduced from HRTEM
and XPS studies is an increase of particle size, from 5.7 to 7.6 nm
4
60 and 630 K [36], which are ascribed to highly dispersed CuO
clusters and isolated Cu ions, respectively [34,58]. Taking into ac-
count the presence of Au in our bimetallic samples and XPS data
presented above, we tentatively explain the TPR peaks at 450 and
6
45 K of catalyst Au1Cu3/TiO2 as mainly due to highly dispersed
Cu-rich ensembles. In contrast, the main hydrogen consumption
peak centered at 523 K is ascribed to dissociative adsorption of hy-
drogen on Au–Cu alloy nanoparticles and reduction of the support
surface by hydrogen spillover. Considering the metal loading, the
hydrogen uptake expected if all metal was oxidized is 0.14 mmol
H2 g , whereas the real uptake was 0.21 mmol H2 g . Taking
into account that both HRTEM and XPS measurements showed
the presence of metallic particles, this means that most of the
hydrogen uptake in the TPR experiment correspond to hydrogen
adsorbed on the alloy nanoparticles and the support. Hydrogen
spillover has been demonstrated to occur over nanometer-sized
gold nanoparticles supported on TiO2 [59].
(
Table 2), whereas similar alloy–support interaction and surface
Au/Cu ratios have been reported. Therefore, the shift of the TPR
signals at higher temperature could be explained in terms of size
effect. In fact, it is known that the spillover of hydrogen over Au
nanoparticles is progressively favored as the nanoparticle size de-
creases [60]. On the other hand, the increase in particle size is
accompanied by a large diminution of the total number of atoms at
the periphery of alloy nanoparticles, thus explaining in the disap-
pearance of the low-temperature hydrogen consumption ascribed
to the perimeter interface. Accordingly, the hydrogen uptake mea-
−1
−1
After calcination at 673 K, the TPR profile of catalyst Au1Cu3/
TiO2 shows two main hydrogen consumption peaks at 445 and
−1
sured over these samples decreased to 0.16–0.17 mmol H2 g
.

194
J. Llorca et al. / Journal of Catalysis 258 (2008) 187–198
Fig. 14. Temperature dependence of the catalytic performance of Au–Cu/TiO2 (1.2%) catalysts calcined at 673 K in propene epoxidation by N2O. (a) Au/TiO2, (b) Au3Cu1/TiO2,
(
c) Au1Cu1/TiO2, (d) Au1Cu3/TiO2. (!) Propene conversion, (") propene oxide selectivity, (2) acrolein selectivity. GHSV = 9000 h−1
.
3
.2. Propene epoxidation
nanoparticles with similar size (from 5.1 to 5.8 nm, Table 2). In
fact, catalyst Au1Cu3/TiO2, which was about four times more ac-
tive than Au/TiO2 at 573 K, for example, had a lower metal loading
and larger particles than Au/TiO2 (1.14 vs 1.21% and 5.7 vs 5.1 nm,
respectively). Therefore, it is confirmed that Au–Cu alloy particles
supported on TiO2 are much more active in the transformation of
propene than Au/TiO2. Our organic-capped nanoparticle prepara-
tion route did not allow a monometallic Cu/TiO2 sample to be
prepared, but in a previous work [36] it was demonstrated that
bimetallic Au–Cu/TiO2 systems perform better in propene epoxi-
dation than Cu/TiO2. Concerning catalyst selectivity, it is interest-
ing to note that catalyst composition also had a strong influence
on product distribution. In particular, the selectivity towards the
desired product, propene oxide, increased strongly as the Cu con-
tent in the catalyst formulation increased (Fig. 14), and values as
high as 50% at 2.5% propene conversion were reached over cata-
lyst Au1Cu3/TiO2 at 473 K. The increase in propene oxide formation
was accompanied by a concomitant decrease in acrolein formation,
which may result from the direct reaction between oxygen species
with H atoms in the allylic position of propene or from propene
oxide transformation [21,28,33]. These results suggest that site iso-
lation effects in the Au–Cu alloy nanoparticles favor the epoxida-
tion route or, alternatively, overcome the problem of consecutive
oxidation of propene oxide. Again, samples can be ordered in terms
of selectivity towards propene oxide in the following manner:
The direct epoxidation of propene with N2O in the temperature
range 473–673 K over the Au–Cu alloy nanoparticles supported
on TiO2 resulted in the formation of propene oxide, propanal,
dimethylketone, acrolein and carbon oxides. As expected, an in-
crease of the reaction temperature resulted in higher propene con-
version, but also in lower selectivity towards the desired product,
propene oxide, at the expense of total oxidation products. Total
deactivation of catalysts was observed after 10 h of reaction; how-
ever, the deactivation was reversible and the activity was com-
pletely restored after a regeneration procedure at 573 K under
an oxygen/helium atmosphere. We first carried out catalytic tests
over Au–Cu/TiO2 samples with different Au/Cu content and the
same total metal loading (1.2%) in order to evaluate the effect of
nanoparticle alloy composition on the reaction. We then continued
with another set of catalytic tests over the best alloy composi-
tion with the aim of deciphering the role of catalyst calcination
temperature, which, as has been demonstrated in the preceding
sections, resulted in different alloy–support interaction and surface
composition. Finally, we performed a similar study with samples
containing lower metal loadings. No activity for the epoxidation of
propene by N2O was detected in a blank experiment carried out
over TiO2 under the same experimental conditions.
3.2.1. Effect of catalyst composition and preparation method
Au1Cu3/TiO2 > Au1Cu1/TiO2 > Au3Cu1/TiO2 > Au/TiO2.
Fig. 14 shows the performance of catalysts containing different
Interestingly, catalytic tests conducted over catalyst Au1Cu3/TiO2
under different gas hourly space velocities (GHSV) did not sig-
nificantly alter product selectivity. When the residence time was
varied from 0.4 to 0.8 and 1.2 s, propene conversion at 473 K in-
creased from 2.5 to 3.1 and 5.1%, whereas propene oxide selectivity
varied slightly from 53.2 to 54.3 and 51.2%, respectively.
These catalytic results can be compared with those obtained
under similar experimental conditions over Au and Au–Cu cata-
lysts obtained by conventional impregnation and reduction in H2
Au:Cu ratios and calcined at 673 K in the propene epoxidation re-
action by N2O. Among these samples, strong differences appear
both in terms of propene conversion and product selectivity. Re-
garding propene conversion, catalysts can be ordered as follows:
Au1Cu3/TiO2 > Au1Cu1/TiO2 > Au3Cu1/TiO2 > Au/TiO2.
It should be recalled that all these catalysts contained virtually
the same total metal amount (from 1.14 to 1.23 wt%, Table 1) and

J. Llorca et al. / Journal of Catalysis 258 (2008) 187–198
195
Table 5
Rate of propene transformation and selectivity to propene oxide at 573 K over cat-
alysts prepared from pre-formed nanoparticles or by conventional impregnation. In
all cases samples were treated at 673 K prior to reaction
−1
−1
h )
Catalyst
Rate (mmol gM
Impregnationa
PO selectivity (%)
Impregnationa
Pre-formedb
Pre-formedb
Au/TiO2
9
133
261
316
592
7.4
23.6
26.8
35.7
41.0
Au3Cu1/TiO2
Au1Cu1/TiO2
Au1Cu3/TiO2
22
45
49
12.4
23.6
26.3
a
Data from [36].
This work.
a
flow at 673 K [36]. The rate of propene transformation at 573 K
for samples prepared by impregnation [36] and those prepared
from thiol-capped nanoparticles (this work) are compiled in Ta-
ble 5, as well as propene selectivity values. In all cases, Au–Cu
catalysts prepared from pre-formed nanoparticles were about one
order of magnitude more active on a metal basis than their respec-
tive counterparts prepared by conventional impregnation. This can
be explained taking into account the fact that the samples used
in this work, prepared from pre-formed Au–Cu alloys and calcined
at 673 K, contained nanoparticles of about 4–7 nm in diameter
Fig. 15. Catalytic performance at 573 K of catalyst Au1Cu3/TiO2 (1.2%) after calci-
nation at 573, 673, 773, and 873 K. (!) Propene conversion, (") propene oxide
selectivity. GHSV = 9000 h−
1
.
Table 6
Catalytic performance at 573 K of Au1Cu3/TiO2 catalysts calcined at different tem-
perature
(Table 2), whereas samples prepared by impregnation contained
b
Catalyst
T a
χ
Selectivity (%)
particles with broad size distributions (up to 20 nm [36]). The
size effect was even more pronounced in Au/TiO2 samples. The
catalyst prepared by impregnation contained Au agglomerates in
the 5–200 nm range and its activity was approximately 15 times
lower than that of catalyst Au/TiO2 prepared from thiol-capped
nanoparticles, which contained well-defined Au nanoparticles of
(
K)
(%)
PO
PA
AC
ACR
COx
Au1Cu3/TiO2 (1.2%)
573
6
7
873
3.0
4.2
3.4
2.5
46.6
52.2
45.8
39.9
6.1
4.4
4.1
3.6
7.6
5.8
3.9
4.0
15.6
18.0
22.5
22.4
24.1
19.6
23.7
30.1
73
73
Au1Cu3/TiO2 (1%)
Au Cu /TiO (0.8%)
573
2.9
3.3
2.6
1.7
38.7
40.1
36.9
30.1
5.6
4.7
3.8
2.5
7.8
5.9
4.1
3.7
23.4
27.2
26.3
28.4
24.5
22.1
28.9
35.3
4–6 nm in size. Due to the wide range of particle size distribu-
6
73
773
73
tion in samples prepared by impregnation, it is not possible to
perform accurate turnover frequency (TOF) comparisons. In addi-
tion to catalyst activity, another point which deserves discussion
is product distribution. For all catalyst formulations, selectivity to-
wards propene oxide was considerably higher in those samples
prepared from pre-formed nanoparticles with respect to samples
prepared by conventional impregnation (Table 5). For example, cat-
alyst Au1Cu3/TiO2, which exhibited the best catalytic performance
in propene epoxidation, yielded 41% of propene oxide at 573 K,
whereas the same catalyst prepared by impregnation yielded, un-
der the same propene conversion (4.3 vs 4.2%) and temperature,
only 26.3% of propene oxide, but 32.2% of COx. For comparison,
8
1
3
2
573
2.2
1.7
0.9
0.3
29.2
32.1
24.1
19.7
7.4
6.4
3.6
2.1
9.8
7.2
4.7
2.3
35.4
35.3
41.1
41.8
18.2
19.0
26.5
34.1
6
7
8
73
73
73
Au1Cu3/TiO2 (0.5%)
573
6
7
873
1.4
0.9
0.4
0.1
23.2
25.6
21.2
19.8
9.4
5.7
4.6
2.4
11.8
6.1
4.2
3.3
39.9
45.8
41.1
37.7
15.7
16.8
28.9
36.8
73
73
Note. PO—propene oxide, PA—propanal, AC—dimethylketone, ACR—acrolein, COx =
−
1
CO + CO2. GHSV = 9000 h
.
a
Calcination temperature.
Propene conversion.
b
+
it has been reported very recently that a K -promoted 1 wt%
CuOx/SBA-15 catalyst yields 26% of propene oxide and 70% COx at
3
.2.2. Effect of calcination temperature and metal loading
Catalyst Au1Cu3/TiO2 (1.2%) was calcined at various tempera-
573 K (4.7% propene conversion) in the propene epoxidation with
O2 [35]. Therefore, catalysts prepared from pre-formed Au–Cu al-
loy nanoparticles are not only more active, but also more selective
towards the epoxidation of propene by N2O. In the epoxidation of
propene over Au/TiO2 in the presence of oxygen and hydrogen it
has been reported that Au particles larger than 2.0 nm in diame-
ter produce propene oxide, whereas smaller Au particles produce
propane [9,19,20]. In our case, where the oxygen donor is N2O and
no hydrogen is present in the reaction mixture, Au–Cu alloy parti-
cles of about 5–6 nm in diameter produce propene oxide, whereas
larger alloy particles (such as those obtained by impregnation) fa-
vor propene combustion (COx). Interestingly, the propene oxide
formation rate over Au–Cu/TiO2 alloy nanoparticles from the epox-
idation of propene by N2O is much higher than that over Au/TiO2
using O2/H2 mixtures, partly because the use of N2O as a mild ox-
idizing agent allows higher working temperatures (and, therefore,
higher propene conversion), whereas an increase in the reaction
temperature results in H2 combustion when O2/H2 mixtures are
employed [9,11].
tures and tested in the propene epoxidation reaction since it was
the most active and selective sample towards the formation of
propene oxide. Propene conversion and propene oxide selectivity
at 573 K of samples calcined at 573, 673, 773, and 873 K are shown
in Fig. 15 and a complete distribution of products is compiled in
Table 6. The performance of the catalyst clearly improved when the
calcination temperature was increased from 573 to 673 K, both in
terms of propene conversion and propene oxide selectivity. In ad-
dition, the amount of total combustion products decreased when
the calcination temperature was 673 K with respect to the sam-
ple calcined at 573 K. In contrast, an increase of the temperature
of calcination up to 773 and 873 K had a negative effect on cat-
alytic performance. Propene conversion progressively decreased at
higher calcination temperature, as well as propene oxide selectivity
(Fig. 15). Furthermore, an increase of the temperature of calcina-
tion from 673 to 873 K resulted in an increase of COx production
(Table 6). These results indicate that a strong transformation oc-
curred in the catalyst as the calcination temperature was varied

196
J. Llorca et al. / Journal of Catalysis 258 (2008) 187–198
and that the best catalytic performance was obtained over the
sample calcined at 673 K, both in terms of activity and selectiv-
ity towards propene oxide.
It has been shown from HRTEM that the calcination temper-
ature has a strong influence on the microstructure of catalysts
(Figs. 7–11). On one hand, the increase of the calcination temper-
ature caused a progressive increase of the nanoparticle size (Ta-
ble 2). On the other hand, higher calcination temperatures favored
a strong interaction between nanoparticles and the TiO2 support.
This has been already discussed in terms of particle shape, contact
angle, and support roughness. In particular, the sample calcined at
673 K, which performed better in the epoxidation of propene, ex-
hibited only a slight increase of particle size with respect to that
calcined at 573 K (5.7 vs 5.3 nm), whereas it showed enhanced
alloy–support interaction. This was also supported by TPR data
(Fig. 12), where the sample calcined at 673 K showed a strong
uptake of hydrogen at low temperature, which was ascribed to
the presence of more reactive sites, such as those expected from
the interface between nanoparticles and TiO2. At higher calcination
temperatures, alloy nanoparticles increased their size considerably,
from 5.7 nm for the sample calcined at 673 K to 7.6 and 9.8 nm
for samples calcined at 773 and 873 K, respectively. This large in-
crease in particle size for samples calcined at 773–873 K resulted
in a strong decrease of the number of atoms located at the inter-
face between the alloy particles and the support, which caused the
disappearance of the TPR signal at low temperature. The turnover
frequency (TOF) of propene oxide formation at 573 K for each cal-
cination temperature was calculated from HRTEM results, either
assuming that all surface atoms were active for propene epox-
idation (Fig. 16a), or that only the perimeter interfaces around
the alloy particles contributed to activity (Fig. 16b). TOF calcula-
tions strongly depend on particle size homogeneity and in some
cases can lead to confusing conclusions; in our case these calcu-
lations can be considered accurate since the preparation route us-
ing pre-formed alloy nanoparticles yielded catalysts with particles
exhibiting narrow size distributions and reasonably well-defined
perimeter interfaces. The highest TOF value per surface atom was
encountered over the Au1Cu3/TiO2 (1.2%) catalyst calcined at 673 K
Fig. 16. Propene oxide (PO) formation rate (TOF) at 573 K over Au1Cu3/TiO2 cata-
lysts calcined at different temperatures. (a) PO TOF per surface atom, (b) PO TOF
per atom at the perimeter interface. (") Au1Cu3/TiO2 (1.2%), (2) Au1Cu3/TiO2 (1%),
−1
(
Q) Au1Cu3/TiO2 (0.8%), (F) Au1Cu3/TiO2 (0.5%). GHSV = 9000 h
.
tween alloy nanoparticles and TiO2, as often postulated for Au/TiO2
catalysts.
−1
(0.010 s ).
As regards the samples calcined at 773 and 873 K, propene
oxide TOF values progressively decreased at increasing calcination
temperature, which cannot be explained in terms of a size increase
of the alloy nanoparticles since TOF values were calculated on the
basis of number of atoms exposed at the surface, or number of
atoms at the perimeter. As stated in Section 3.1.3.2, XPS data in-
dicated that the calcination temperature had a strong effect on
surface composition. As the calcination temperature increased, the
surface Au/Cu atomic ratio decreased and copper oxidized, that is,
the catalyst surface was decorated with positively charged copper
species, particularly after calcination at 873 K (Table 4). It has been
reported that propene epoxidation with O2 over Cu/SiO2 catalysts
is inhibited by Cu oxidation [30,32,33]. Therefore, our findings sug-
gest that at a calcination temperature of 773 K or higher, the sur-
face of Au–Cu alloy nanoparticles becomes increasingly enriched
in Cu and partly oxidized, thus resulting in progressively lower
TOF numbers for propene epoxidation. Fig. 17 summarizes all these
changes at the microstructural level. We conclude that the calcina-
tion temperature of 673 K resulted in the best catalyst for propene
epoxidation because a large alloy–support interaction took place
at this temperature, while surface decoration was avoided and a
small particle size was maintained, which means a large number
of active sites at the perimeter of the contact area between the
alloy nanoparticles and TiO2.
A slightly different pattern of TOF values vs temperature of cal-
cination was encountered when TOF values were calculated con-
sidering only the perimeter interfaces around the alloy nanopar-
ticles (Fig. 16b). Obviously in this case TOF values were much
higher than those calculated per surface atom, and the sample cal-
cined at 673 K yielded a TOF value for propene oxide formation of
−1
1
.33 s . The excellent match between the pattern of TOF values
calculated on a perimeter basis vs calcination temperature and that
of Fig. 15, which shows propene conversion and propene oxide se-
lectivity values obtained directly from the experiments, warrants
mention. This remarkable correspondence suggests that the active
sites for propene transformation are located at the contact area
(
interface) of alloy nanoparticles and TiO2. A similar conclusion has
been reported in the epoxidation of propene with O2/H2 mixtures
over several gold–titania catalysts, where the active sites were as-
cribed to the Au–Ti interface [4,11]. In addition, explanations for
the exceptional activity of Au nanostructures for CO oxidation of-
ten include a direct role played by the perimeter of the Au–support
interface [60–62]. Note that the increase of propene oxide yield be-
tween samples calcined at 573 and 673 K (approximately 1.5 times)
is comparable to that in propene oxide TOF values calculated con-
sidering only the perimeter interface atoms (ca. 1.2 times), but
not to the increase in TOF assuming all the surface atoms (about
2.3 times). These results suggest that the catalytic behavior may
The catalytic results attained at 573 K with the Au Cu /TiO
2
1
3
be related to the extent of perimeter interface around Au–Cu alloy
nanoparticles rather than the strength of the interfacial contact be-
catalysts prepared with lower metal loadings and calcined between
573 and 873 K are included in Table 6. These are Au1Cu3/TiO2

J. Llorca et al. / Journal of Catalysis 258 (2008) 187–198
197
supported on anatase and calcined at 673 K. These alloy nanoparti-
cles, with diameters of about 5–6 nm, are considerably more active
and selective towards the epoxidation of propene by N2O than
their monometallic counterparts, Au/TiO2 and Cu/TiO2, which may
be ascribed to site isolation effects caused by alloying. For sam-
ples calcined at lower temperature the interaction between alloy
nanoparticles and TiO2 is weak and low propene oxide TOF values
are obtained, suggesting that the complex synergy created at the
perimeter interface around Au–Cu alloy nanoparticles appears es-
sential for epoxidation capability. However, at calcination tempera-
tures higher than 673 K, alloy nanoparticles are progressively dec-
orated with oxidized Cu species and the catalytic performance to-
wards propene oxide decreases, in spite of strong particle–support
interaction. This decoration effect is more pronounced with de-
creasing alloy particle size from ca. 5 to 2 nm, and allylic oxidation
products are preferably obtained.
Acknowledgments
This work was supported by MEC grants ENE2006-06925 (J.Ll.),
CTQ2006-08196 (O.R.), and CTQ2006-02362 (F.M.).
References
[1] G.C. Bond, C. Louis, D. Thompson, Catalysis by Gold, Imperial College Press,
London, 2006.
[
[
[
[
[
[
2] M. Haruta, Gold Bull. 37 (2004) 27.
3] M. Haruta, Chem. Rec. 3 (2003) 75.
4] M. Haruta, CATTECH 6 (2002) 102.
5] G.C. Bond, D.T. Thompson, Appl. Catal. A Gen. 302 (2006) 1.
6] D.T. Thompson, Top. Catal. 38 (2006) 231.
7] T.A. Nijhuis, M. Makkee, J.A. Moulijn, B.M. Weckhuysen, Ind. Eng. Chem. Res. 45
(2006) 3447.
Fig. 17. Effect of calcination temperature on the structure of Au–Cu alloy nanopar-
ticles in catalyst Au1Cu3/TiO2 (1.2%). Initially, Au–Cu alloy nanoparticles are capped
with an organic envelop. After deposition on TiO2 and calcination at 573 K the or-
ganic envelop is removed but the Au–Cu alloy structure is preserved. At 673 K the
size of nanoparticles increases and a strong interaction between nanoparticles and
TiO2 occurs. At 773 K there is a slight enrichment of Cu on the surface, which is
more pronounced at 873 K. At this temperature the size of nanoparticles increases
strongly.
[8] M. Haruta, N. Yamada, T. Kobayashi, S. Iijima, J. Catal. 115 (1989) 301.
[9] T. Hayashi, K. Tanaka, M. Haruta, J. Catal. 178 (1998) 566.
[10] Y.A. Kalvachev, T. Hayashi, S. Tsubota, M. Haruta, J. Catal. 186 (1999) 228.
[11] E.E. Stangland, K.B. Stavens, R.P. Andres, W.N. Delgass, J. Catal. 191 (2000) 332.
[12] J. Chou, E.W. McFarland, Chem. Commun. (2004) 1648.
[13] E. Sacaliuc, A.M. Beale, B.M. Weckhuysen, T.A. Nijhuis, J. Catal. 248 (2007) 235.
[14] A.K. Sinha, S. Seelan, M. Okumura, T. Akita, S. Tsubota, M. Haruta, J. Phys. Chem.
B 109 (2005) 3956.
[15] B. Chowdhury, J.J. Bravo-Sárez, M. Daté, S. Tsubota, M. Haruta, Angew. Chem.
Int. Ed. 45 (2006) 412.
16] J. Lu, X. Zhang, J.J. Bravo-Suárez, K.K. Bando, T. Fujitani, S.T. Oyama, J. Catal. 250
(2007) 350.
(1%), Au1Cu3/TiO2 (0.8%), and Au1Cu3/TiO2 (0.5%). These samples
[
had the same bulk Au/Cu ratio but differed in their particle size
and surface Au/Cu values, as determined by HRTEM and XPS (Sec-
tion 3.1.2). Catalyst Au1Cu3/TiO2 (1%) exhibited a similar trend than
that of Au1Cu3/TiO2 (1.2%) discussed above, and the sample cal-
cined at 673 K was again the most active for propene epoxidation.
However, propene oxide TOF values were about 40% lower than
those of Au1Cu3/TiO2 (1.2%) catalyst (Fig. 16). Taking into account
the fact that samples Au1Cu3/TiO2 (0.8% ) and Au1Cu3/TiO2 (0.5%)
exhibited even lower activity for propene oxide formation and that
surface Au/Cu values progressively decreased as the metal loading
of the samples decreased (Table 3), we conclude that decoration
of the alloy nanoparticles with copper species is responsible for
the low catalytic performance in propene epoxidation of these
Au1Cu3/TiO2 catalysts with low metal loading. It is possible that
surface copper enrichment may be related to a nanoparticle size,
since the alloy nanoparticles became smaller as the metal loading
decreased. In fact, as the metal loading decreased and the surface
was progressively enriched in Cu, the selectivity towards acrolein
increased strongly at the expense of propene oxide (Table 6) and
it has been reported that copper oxide is particularly active for the
oxidation of propene to acrolein [32].
[17] C. Sivadinarayana, T.V. Chouhary, L.L. Daemen, J. Eckert, D.W. Goodman, J. Am.
Chem. Soc. 126 (2004) 38.
[
18] M. Haruta, B.S. Uphade, S. Tsubota, A. Miyamoto, Res. Chem. Intermed. 24
1998) 329.
19] M. Haruta, M. Daté, Appl. Catal. A Gen. 222 (2001) 427.
(
[
[20] A.K. Sinha, S. Seelan, S. Tsubota, M. Haruta, Angew. Chem. Int. Ed. 43 (2004)
1546.
[21] V. Duma, D. Hönicke, J. Catal. 191 (2000) 93.
[
[
[
22] E. Ananieva, A. Reitzmann, Chem. Eng. Sci. 59 (2004) 5509.
23] X. Wang, Q. Zhang, S. Yang, Y. Wang, J. Phys. Chem. B 109 (2005) 23500.
24] Q. Zhang, Q. Guo, X. Wang, T. Shishido, Y. Wang, J. Catal. 239 (2006) 105.
[25] A. Costine, T. O’Sullivan, B.K. Hodnett, Catal. Today 112 (2006) 103.
26] Y. Wang, W. Yang, L. Yang, X. Wang, Q. Zhang, Catal. Today 117 (2006) 156.
[
[27] T. Thömmes, S. Zürcher, A. Wix, A. Reitzmann, B. Kraushaar-Czarnetzki, Appl.
Catal. A Gen. 318 (2007) 160.
[28] B. Moens, H.D. Winne, S. Corthals, H. Poelman, R. De Gryse, V. Meynen, P. Cool,
B.F. Sels, P.A. Jacobs, J. Catal. 247 (2007) 86.
[29] S. Yang, W. Zhu, Q. Zhang, Y. Wang, J. Catal. 54 (2008) 251.
[30] R.M. Lambert, F.J. Williams, R.L. Cropley, A. Palermo, J. Mol. Catal. A Chem. 228
(
2005) 27.
31] D. Torres, N. Lopez, F. Illas, R.M. Lambert, Angew. Chem. Int. Ed. 46 (2007)
055.
[
2
[32] J. Lu, M. Luo, H. Lei, X. Bao, C. Li, J. Catal. 211 (2002) 552.
[33] O.P.H. Vaughan, G. Kyriakou, N. Macleod, M. Yikhov, R.M. Lambert, J. Catal. 236
(2005) 401.
[34] H. Chu, L. Yang, Q. Zhang, Y. Wang, J. Catal. 241 (2006) 225.
[35] Y. Wang, H. Chu, W. Zhu, Q. Zhang, Catal. Today 131 (2008) 496.
[36] R.J. Chimentao, F. Medina, J.L.G. Fierro, J. Llorca, J.E. Sueiras, Y. Cesteros, P. Sala-
gre, J. Mol. Catal. A Chem. 274 (2007) 159.
4
. Conclusions
High yields of propene oxide can be formed through Au–Cu–
Ti intimate contacting from well-defined Au–Cu alloy nanoparticles
[37] Q. Fu, H. Saltsburg, M. Flytzani-Stephanopoulos, Science 301 (2003) 935.

198
J. Llorca et al. / Journal of Catalysis 258 (2008) 187–198
[
[
38] C. Mohr, H. Hofmeister, J. Radnik, P. Claus, J. Am. Chem. Soc. 125 (2003) 1905.
39] B.L.V. Prasad, S.I. Stoeva, C.M. Sorensen, K.J. Klabunde, Langmuir 18 (2002)
515.
[51] J. Llorca, A. Casanovas, M. Domínguez, I. Casanova, I. Angurell, M. Seco, O. Ros-
sell, J. Nanoparticle Res. 10 (2008) 537.
[52] M.J. Hostetler, C.-J. Zhong, B.K.H. Yen, J. Anderegg, S.M. Gross, N.D. Evans,
M. Porter, R.W. Murray, J. Am. Chem. Soc. 120 (1998) 9396.
7
[
40] J.P. Spatz, S. Mössmer, C. Hartmann, M. Möller, T. Herzog, M. Krieger, H.-G.
Boyen, P. Ziemann, B. Kabius, Langmuir 16 (2000) 407.
41] J. Chou, N.R. Franklin, S.-H. Baeck, T.F. Jaramillo, E.W. McFarland, Catal. Lett. 95
[53] G. Battaglin, E. Cattaruzza, F. Gonella, G. Mattei, P. Mazzoldi, C. Sada, X. Zhang,
Nucl. Instrum. Methods Phys. Res. B 166–167 (2000) 857.
[54] G. De, C.N.R. Rao, J. Phys. Chem. B 107 (2003) 13597.
[
(2004) 107.
[
[
[
[
42] R. Zanella, S. Giorgio, C.R. Henry, C. Louis, J. Phys. Chem. B 106 (2002) 7634.
43] L.D. Menard, F. Xu, R.G. Nuzzo, J.C. Yang, J. Catal. 243 (2006) 64.
44] K. Mallick, M.J. Witcomb, M.S. Scurrell, Appl. Catal. A Gen. 259 (2004) 163.
45] J.-D. Grunwaldt, M. Maciejewski, O.S. Becker, P. Fabrizioli, A. Baiker, J. Catal. 186
[55] H.G. Boyen, G. Kästle, Science 297 (2002) 1533.
[56] G. Meitzner, G.H. Via, F.W. Lytle, J.H. Sinfelt, J. Chem. Phys. 83 (1985) 4793.
[
57] A.Q. Wang, J.H. Liu, S.D. Lin, T.S. Lin, C.Y. Mou, J. Catal. 233 (2005) 186.
58] S. Velu, K. Susuki, M. Okasaki, M.P. Kapoor, T. Osaki, F. Ohashi, J. Catal. 194
2000) 373.
[
[
[
(
1999) 458.
46] A.C. Templeton, W.P. Wuelfing, R.W. Murray, Acc. Chem. Res. 33 (2000) 27.
47] C.-J. Zhong, M.M. Maye, Adv. Mater. 13 (2001) 1507.
(
[
[
[
[
59] S. Schimpf, M. Lucas, C. Mohr, U. Rodemerck, A. Brückenr, J. Radnik, H. Hof-
meister, P. Claus, Catal. Today 72 (2005) 63.
60] S. Tsubota, T. Nakamura, K. Tanaka, M. Haruta, Catal. Lett. 56 (1998) 131.
48] M.M. Maye, J. Luo, L. Han, N.N. Kariuki, C.-J. Zhong, Gold Bull. 36 (2003) 75.
49] M. Brust, M. Walker, D. Bethell, D.J. Schiffrin, R. Whyman, J. Chem. Soc. Chem.
Commun. (1994) 801.
[61] J.-D. Grunwaldt, C. Kiener, C. Wögerbauer, A. Baiker, J. Catal. 181 (1999) 223.
[62] M.S. Chen, D.W. Goodman, Science 306 (2004) 252.
[
50] C.J. Kiely, J. Fink, J.G. Zheng, M. Brust, D. Bethell, D. Schiffrin, J. Adv. Mater. 12
(
2000) 640.