Formation and Thermal Stability of Au
2
O
3
on Gold Nanoparticles: Size and Support
Effects
Luis K. Ono and Beatriz Roldan Cuenya*
Department of Physics, UniVersity of Central Florida, Orlando, Florida 32816
ReceiVed: NoVember 28, 2007; In Final Form: January 9, 2008
Gold nanoparticles with two different size distributions (average sizes of ∼1.5 and ∼5 nm) have been
synthesized by inverse micelle encapsulation and deposited on reducible (TiO
2
) and nonreducible (SiO
2
)
supports. The thermal and chemical stability of oxidized gold species formed upon cluster exposure to atomic
oxygen have been investigated in ultrahigh vacuum using a combination of temperature-, time- and CO dosing-
dependent X-ray photoelectron spectroscopy (XPS), as well as temperature-programmed desorption (TPD).
Our work demonstrates that (a) low-temperature (150 K) exposure to atomic oxygen leads to the formation
of surface as well as subsurface gold oxide on Au nanoparticles, (b) the presence of the reducible TiO
2
substrate leads to a lower gold oxide stability compared to that on SiO
2
, possibly because of a TiO
2
oxygen
vacancy-mediated decomposition process, (c) heating to 550 K (Au/SiO
2
) and 300 K (Au/TiO
2
) leads to a
near-complete reduction of small (∼1.5 nm) NPs while a partial reduction is observed for larger clusters
(∼5 nm), and (d) the desorption temperature of O
2
from preoxidized Au clusters deposited on SiO
2
depends
on the cluster size, with smaller clusters showing stronger O
2
binding.
Introduction
Bulk gold is known as one of the most inert metals in the
periodic table.
1
This trait is attributed to the lack of interaction
between the orbitals of adsorbates and the filled d states of
gold.
2,3
The high value of the enthalpy of oxygen chemisorption
by gold to form its oxide Au
2
O
3
(∆H )+19.3 kJ/mol) also
indicates its chemical inertness.
4
However, pioneering work by
Haruta et al. has demonstrated that highly dispersed Au
nanoparticles (NPs) (<10 nm) supported on TiO
2
are consider-
ably active for the low-temperature oxidation of CO
5
and C
3
H
6
epoxidation.
6
Since then, the topic of gold in catalysis has
received significant attention. Despite the large number of
reports currently available, the microscopic origin of gold’s
heterogeneous catalytic activity is still poorly understood, and
lack of consensus prevails with respect to the nature of the active
sites,
7-9
the chemical state of the active gold species (metallic
vs ionic gold),
10-12
as well as the relative importance of the
different oxygen species (chemisorbed oxygen, surface oxide,
subsurface oxygen) that might be present on these catalysts
under realistic reaction conditions.
13,14
Several authors attribute the enhanced chemical reactivity of
gold NPs to the presence of ionic gold; however, whether
anionic or cationic (Au
δ-
,Au
δ+
) gold species are preferable is
still a matter of debate.
11,12,14-17
The presence of cationic
interfacial gold species has been inferred on catalytically active
systems such as Au deposited on reduced TiO
2
.
18
This is
contrary to the traditional picture involving electron transfer
from oxide supports to Au NPs. Positively charged atoms at
the metal-support interface have been observed, and their
presence has been correlated to the enhanced catalytic activity
of the system.
19,20
The active role of these cationic gold species
on the water-gas shift reaction (WGS) has been demonstrated
recently by Fu et al.
21
In that work, similar activities for the
WGS reaction were obtained on gold NPs supported on La-
doped CeO
2
before and after the removal of the metallic gold
species.
21
This result suggested a strong interaction of ionic gold
with the ceria support, and the authors claimed that the presence
of metallic NPs was not necessary for this catalytic reaction.
22
On the theoretical side, several density functional theory
(DFT) works have investigated the strength of the binding of
CO,
23,24
O
2
,
25,26
propene,
27
and methanol
28
to nonmetallic Au
clusters. Here, a common trend was found, with reactant
molecules showing stronger binding energy to anionic and
cationic gold species, as compared to metallic gold. Recent
hybrid DFT calculations carried out by Okumura et al.
9
suggested that although O
2
activation occurs on anionic Au
cationic Au atoms show stronger bonding to CO. The authors
proposed a model of dynamic charge polarization in which a
strong heterojunction between Au clusters and their support is
indispensable for the activation of oxygen species. The presence
of negatively charged atoms in the perimeter region of Au NPs
was attributed to localized Coulomb blockade effects. Further-
more, this surface negative charge was found to increase with
decreasing cluster size,
29
in agreement with the known enhanced
catalytic activity observed for small Au clusters.
Several experimental studies have been dedicated to the
electronic and chemical characterization of oxidized gold species
on Au single crystals,
30-32
polycrystalline films,
33,34
and the
surface of Au NPs.
35,36
The instability of Au
2
O
3
formed upon
exposing a gold film (100 nm) to O
2
plasma has been reported
by Tsai et al.
37
In this work, an activation energy of 57 kJ/mol
was extracted for the decomposition of gold oxide from the
correlation of electrical resistance measurements, conducted in
air at room temperature (RT), and X-ray photoelectron spec-
troscopy (XPS) investigations after several time intervals. A
nonmonotonic interaction of size-selected gold clusters with
atomic oxygen was found by Boyen et al.
35
More specifically,
1.4 nm clusters (55 atoms, closed-shell electronic structure) were
found to be resistant to oxidation. However, when larger (>1.6
* To whom correspondence should be addressed. E-mail:
roldan@physics.ucf.edu.
4676 J. Phys. Chem. C 2008, 112, 4676-4686
10.1021/jp711277u CCC: $40.75 © 2008 American Chemical Society
Published on Web 03/04/2008
nm) and smaller clusters (<1.3 nm) were exposed to an identical
treatment, Au(core)/Au
2
O
3
(shell) structures with oxide shell
thicknesses on the order of 0.7 nm were obtained.
Recently, a number of experimental studies discussing the
chemical reactivity of gold surfaces precovered with atomic
oxygen have been published.
38-41
Min et al.
42
reported the
existence of three types of oxygen species on Au(111): (i)
chemisorbed oxygen (oxygen bound to gold that is not part of
an ordered phase), (ii) oxygen in surface oxide (well-ordered
two-dimensional phase), and (iii) subsurface oxygen or bulk
oxide (three-dimensional phase). On the basis of CO oxidation
studies, the authors found the following relation among the
reactivity of the different oxygen species: chemisorbed oxygen
> oxygen in surface oxide > oxygen in bulk gold oxide.
42
The
role played by these species in the thermodynamics and kinetics
of oxidation/reduction reactions is a major challenge for the
microscopic understanding of gold catalysis. This understanding
becomes highly challenging when small NPs are considered
because of the added complexity of the presence of different
facets, kinks, and steps
43
and substantial interactions with the
support.
44
As an example, on ∼6 nm Au NPs deposited on
highly ordered pyrolytic graphite (HOPG), Lim et al.
36
observed
by XPS the formation of a single oxygen species upon exposure
to atomic oxygen under ultrahigh-vacuum (UHV) conditions.
This species was ascribed to chemisorbed oxygen because it
was found to readily react with CO producing CO
2
. Interestingly,
two different oxygen species were identified on similarly treated
but smaller gold NPs (∼3 nm). In this case, one of the oxygen
species was assigned to chemisorbed oxygen, which decomposes
rapidly upon CO exposure, and the other to subsurface oxygen,
inert toward reaction with CO.
Our work intends to address the following questions: (1)
under which conditions is gold oxide stable on a gold nano-
particle, (2) how are the chemical kinetics of gold oxide
decomposition affected by the size of the particles and the nature
of their metal oxide support, and (3) is more than one gold oxide
species (surface and subsurface) stable on NPs? In order to gain
insight into these topics, we have used micelle encapsulation
methods
43-48
to synthesize size- and shape-selected gold nano-
clusters supported on thin SiO
2
and TiO
2
films. Morphological
characterization was conducted by atomic force microscopy
(AFM), and the decomposition of oxidized gold species, formed
upon in situ O
2
plasma exposure, was investigated in UHV by
XPS. Temperature-programmed desorption (TPD) measurements
provided information on the reaction order and activation energy
for molecular oxygen desorption from Au
2
O
3
(shell)/Au(core)
NPs supported on SiO
2
. Finally, CO exposure experiments were
conducted to distinguish surface oxides from subsurface oxides.
Experimental Section
Size-selected Au nanoparticles were synthesized by inverse
micelle encapsulation on polystyrene-block-poly(2-vinylpyri-
dine) diblock copolymers [PS(x)-b-P2VP(y), Polymer Source
Inc.]. When the PS-P2VP polymers are dissolved in toluene,
inverse micelles are formed with the polar units (P2VP)
constituting the core and the nonpolar polystyrene (PS) tails
extending outward. Subsequently, chloroauric acid (HAuCl
4
‚
3H
2
O) is added to the polymeric solution and AuCl
4
-
com-
pounds attach to the pyridine groups in the P2VP core. The
micelles containing Au NPs are deposited onto different
substrates by dip-coating at a speed of 1 µm/min. The NP size
can be tuned by changing the molecular weight of the polymer
head (P2VP) as well as the metal salt/P2VP concentration ratio.
The length of the polymer tail (PS) determines the interparticle
distance. In this study, the following two polymers have been
used: PS(53000)-P2VP(43800) and PS(8200)-P2VP(8300).
Naturally oxidized Si(111) wafers and ultrathin Ti films (15
nm) electron-beam deposited on Si(111) have been used as NP
supports. Further details on this preparation method can be found
in refs 43-48. A summary of the synthesis parameters used is
given in Table 1.
The characterization of the sample morphology was per-
formed ex situ by AFM in tapping-mode (Digital Instruments,
Multimode). The ex situ prepared samples were mounted on a
molybdenum sample holder with a K-type thermocouple located
directly underneath the sample and transferred into a modular
UHV system (SPECS GmbH) for polymer removal and
electronic/chemical characterization. The system is equipped
with an hemispherical electron energy analyzer (Phoibos 100,
SPECS GmbH) and a dual-anode (Al KR, 1486.6 eV and Ag
LR, 2984.4 eV) monochromatic X-ray source (XR50M, SPECS
GmbH) for XPS, and a differentially pumped quadrupole mass
spectrometer (QMS, Hiden Analytical, HAL 301/3F) with an
electron-beam sample heating system connected to a PID
temperature controller (Eurotherm, 2048) for TPD experiments.
The base pressure in this chamber is 1-2 × 10
-10
mbar.
Polymer removal from the Au NP’s surface was conducted
by O
2
plasma exposure (Oxford Scientific, OSMiPlas) at low
temperature (150 K) at a pressure of 5.5 × 10
-5
mbar for 100
min. The polymer is removed during the first 20-30 min of
this treatment, and further O
2
plasma exposure results in the
formation of Au-O compounds. The XPS measurements,
conducted immediately after the O
2
plasma treatments, were
done with the sample at a temperature lower than 200 K. For
the temperature-dependent decomposition of gold oxide (XPS),
a linear heating ramp with β ) 3 K/s was used. Because all of
our annealing treatments were conducted in vacuum, lower
decomposition temperatures for gold oxide are expected in our
case as compared to similar studies conducted under higher
partial O
2
pressure.
37
The XPS binding energy (BE) scale has been calibrated using
the Ti-2p
3/2
peak on Ti (454.2 eV) and Si-2p
3/2
peak on Si (99.3
eV) substrates as references. We can see the Ti
0
and Si
0
peaks
because of the ultrathin nature of our TiO
2
[TiO
2
(6 nm)/Ti(9
nm)/SiO
2
/Si(111)] and SiO
2
[SiO
2
(4 nm)/Si(111)] support films.
From cross-sectional TEM measurements (not shown), the
thicknesses of the TiO
2
and SiO
2
films are 6.0 ( 0.5 and 3.8 (
0.5 nm, respectively. Because Au
2
O
3
is known to decompose
under intense X-ray exposure within hours,
31
a control experi-
ment was conducted to ensure that no gold oxide decomposition
occurred during our XPS acquisition time (∼10 min). A
maximum decrease in the Au
3+
signal of 2% was observed under
TABLE 1: Summary of the Parameters Tuned and Average Height and Diameter of Au Nanoparticles Deposited on SiO
2
and
TiO
2
sample substrate
PS/P2VP molecular
weight (g/mol)
HAuCl
4
/
P2VP ratio
height
(nm)
diameter
(nm)
interparticle
distance (nm)
no. 1 SiO
2
53 000/43 800 0.4 4.9 ( 1.6 16 ( 350( 6
no. 2 TiO
2
53 000/43 800 0.4 5.4 ( 1.2 19 ( 551( 18
no. 3 SiO
2
8200/8300 0.05 1.7 ( 0.8 12 ( 424( 4
no. 4 TiO
2
8200/8300 0.05 1.4 ( 0.5 10 ( 459( 40
Formation and Thermal Stability of Au
2
O
3
J. Phys. Chem. C, Vol. 112, No. 12, 2008 4677
our measurement conditions (Al KR radiation, 1486.6 eV at a
power of 300 W). For the analysis of peak positions, line widths,
and relative areas of the Au
0
and Au
3+
components, the raw
XPS spectra were fitted with two (Au
0
) or four (Au
0
,Au
3+
)
Gaussian functions after linear background subtraction. During
the fitting, the intensity ratio between the Au-4f
7/2
and Au-4f
5/2
peaks was fixed to 0.75, and the full width at half-maximum
(fwhm) of the different components was 1.3 ( 0.3 eV (Au
0
)
and 1.7 ( 0.5 eV (Au
3+
).
34
Prior to the TPD measurements, the polymer-free Au NPs
were flash-annealed to 700 K. This procedure reduced the
residual gas background significantly at high temperatures
without inducing any significant changes in the NP size
distribution. Subsequently, the samples were exposed to atomic
oxygen at a pressure of 2.3 × 10
-5
mbar for 15 min. For the
TPD studies, the samples were kept at RT during O
2
plasma
exposure because a cold plasma treatment resulted in an increase
of the background of the TPD spectra. Subsequently, the samples
were placed ∼3 mm away from the mass spectrometer glass
shield opening (5 mm aperture) and heated at a rate of 5 K/s.
In order to investigate if for a given particle size and support
combination, both surface and subsurface gold oxide species
were formed upon exposure to atomic oxygen at low temper-
ature (150 K), the samples were dosed with CO (P
CO
) 1.0 ×
10
-5
mbar for 10 min, 4500 L), and the Au
3+
XPS signal was
monitored before and after CO exposure. All samples were
dosed with CO at RT except sample no. 4, for which the CO
dosing was conducted at ∼150 K (and XPS measured at ∼200
K) in order to prevent significant thermal decomposition of the
relatively unstable oxide formed on this sample.
Results and Discussion
(a) Morphological Characterization (AFM). Figure 1
displays AFM micrographs of Au NPs with two different size
distributions synthesized using diblock copolymers with different
molecular weights: PS(53000)-P2VP(43800) (Figure 1a-d),
and PS(8200)-P2VP(8300) (Figure 1e and f). The particles were
deposited on SiO
2
, Figure 1a and e (samples 1 and 3) and on
TiO
2
, Figure 1c and f (samples 2 and 4), and all images were
taken after polymer removal using O
2
plasma. In addition, the
influence of annealing in UHV to 700 K (sample no. 1) and
500 K (sample no. 2) on the nanoparticle size was monitored,
Figure 1b and d, respectively. No significant size changes were
observed in any of the samples upon annealing. The sizes of
the Au NPs estimated by AFM after O
2
plasma are given in
Table 1. Particles with similar size distributions (4.9-5.4 nm
height for samples 1 and 2, and 1.4-1.7 nm for samples 3 and
4) were found on both substrates when the same encapsulating
polymer was used in the synthesis. Because of the AFM tip-
convolution effects (tip radius < 7 nm), the average NP height
was used as the characteristic size parameter.
(b) Electronic Characterization (XPS). The thermal de-
composition of oxidized gold species in core(Au
0
)/shell(Au
3+
)
nanoparticles was monitored in situ (UHV) by XPS. Figure 2
shows XPS spectra from the Au-4f core level region of our four
samples as a function of the annealing temperature. Spectra (i)
were measured at ∼200 K directly after a low-temperature
(∼150 K) O
2
plasma treatment. Spectra ii and iii were acquired
after isothermal sample annealing for 10 min at 400 and 500 K
respectively, followed by a fast cool down to room temperature
using liquid nitrogen flow. The two doublets observed with
maxima at (84.6 ( 0.3, 88.4 ( 0.3 eV) and (86.9 ( 0.2, 90.6
( 0.2 eV) were assigned to the 4f
7/2
and 4f
5/2
core levels of
Au
0
and Au
3+
in Au
2
O
3
.
35,36,44
The vertical reference lines in
Figure 2 indicate the binding energies of bulk metallic gold (84.0
and 87.7 eV, solid lines) and Au
3+
(85.8 and 89.5 eV, dashed
lines).
34,49
The different Au-O species cannot be distinguished
based on XPS spectra from the Au-4f region. Previous studies
by Friend’s group
42
on preoxidized gold single crystals dem-
onstrated the presence of distinct Au-O species (chemisorbed
oxygen, surface, and bulk gold oxide) based on the appearance
of multiple peaks in their O-1s XPS spectra. A similar analysis
of our samples is more difficult because our substrates (SiO
2
and TiO
2
) already contain oxygen and the overlap between the
binding energies of the different oxide species makes their
individual detection challenging.
In Figure 2, positive shifts in BE were observed, and in
agreement with previous literature reports,
43,44,50
their magnitude
was found to strongly depend on the size of the NPs and the
nature of the substrate. In particular, BE shifts of +0.3 ( 0.1
eV (sample no. 1), +0.2 ( 0.1 eV (sample no. 2), +0.9 ( 0.1
eV (sample no. 3), and +0.8 ( 0.1 eV (sample no. 4) were
measured on our samples after annealing at 350 K. By
comparing the BE values of samples with two distinct size
distributions, deposited on the same substrate (SiO
2
), a clear
size effect is observed with the smallest particles (1.7 nm,
sample no. 3) displaying larger BE shifts (+0.6 eV) than the
4.9 nm clusters in sample no. 1. The same conclusion is true
when differently sized clusters are deposited on TiO
2
(samples
2 and 4). Positive BE shifts observed for small clusters are
Figure 1. Tapping-mode AFM images of size-selected Au nanopar-
ticles supported on SiO
2
[samples 1 (a) and 3 (e)] and on TiO
2
[samples
2 (c) and 4 (f)] taken after an in situ O
2
plasma treatment (90 W, 5.5
× 10
-5
mbar, 100 min) at 150 K. The images shown in b and d
correspond to samples 1 and 2 after a subsequent flash anneal in UHV
at 700 K (b) and 500 K (d), respectively. The particles were synthesized
by encapsulation in two different diblock copolymers: PS(53000)-
P2VP(43800) (samples 1 and 2) and PS(8200)-P2VP(8300) (samples
3 and 4). The height scales are (a) z ) 20 nm, (b) z ) 20 nm, (c) z )
30 nm, (d) z ) 30 nm, (e) z ) 12 nm, and (f) z ) 8 nm.
4678 J. Phys. Chem. C, Vol. 112, No. 12, 2008 Ono and Roldan Cuenya
commonly attributed to initial
51,52
and final state effects.
44,53
In
addition, DFT calculations by Yang et al.
54
suggested that
positive core-level shifts measured for Au NPs supported on
MgO(001) and TiO
2
(110) could be related to the presence of
oxygen vacancies in the supports.
(c) Temperature Dependence of the Thermal Decomposi-
tion of Au
2
O
3
. Size Effects. It is known that the particle size
affects the reduction rate of metal nanocatalysts, and two models
based on geometrical and electronic effects have been pro-
posed.
55,56
In the geometrical model, different oxygen adsorption
sites are available for differently sized clusters. In the electronic
model, size-dependent changes in the electronic structure of
small clusters are believed to play a role in the stability of metal
oxide cluster shells. For our large NPs (∼5 nm), geometrical
effects should dominate, while electronic effects may also play
a role in the reduction of our small clusters (∼1.5 nm).
The influence of the particle size on the thermal stability of
Au
2
O
3
can be inferred by comparing XPS spectra taken on
samples with two different particle size distributions supported
on the same substrate (Figure 2a and c for Au/SiO
2
and Figure
2b and d for Au/TiO
2
). In order to estimate the thickness of the
gold oxide formed upon low-temperature O
2
plasma exposure,
the model described by Nanda et al. and Wu et al. in refs 57
and 58 was used. Following this model, the NP shape is assumed
to be spherical and composed of a metallic core (Au) and an
oxidized shell (Au
2
O
3
). The ratio of the intensities of the
photoelectron peaks (4f
7/2
in our analysis) from the Au
0
core
and Au
3+
shell is given by
where R
1
is the radius of the NP core and R
2
is the total radius
of the NP (measured by AFM). The inelastic mean free path
(IMFP) of electrons in metallic Au (λ
1
) is 1.781 nm.
59
Using
NIST software
59
and a gold oxide density of 13.675 g/cm
3
(see
ref 60 for details on the structure of Au
2
O
3
), we estimated an
IMFP (λ
2
) of 1.937 nm for Au
2
O
3
. The constants K
Au
and K
Au
2
O
3
are related to the distinct elemental sensitivities and instrumental
factors. A value of K
Au
2
O
3
/K
Au
) 0.32 was used in our studies.
57
Equation 1 was evaluated numerically, and the R
1
value was
varied until the I
Au
2
O
3
(θ)/I
Au
(θ) ratio matched the intensity ratio
measured by XPS.
Figure 3 shows the calculated Au
2
O
3
shell thicknesses as a
function of temperature. As described above, all samples were
annealed for 10 min at the respective temperatures and the XPS
spectra were measured subsequently at room temperature. The
maximum thicknesses of the Au
2
O
3
shell formed on the large
NPs deposited on SiO
2
(sample no. 1) and TiO
2
(sample no. 2)
after O
2
plasma were 0.79 ( 0.02 nm and 0.83 ( 0.02 nm,
respectively. For the small clusters, the initial maximum Au
2
O
3
thicknesses were 0.38 ( 0.02 nm (sample no. 3, SiO
2
) and 0.31
( 0.01 nm (sample no. 4, TiO
2
). For NPs deposited on both
substrates, a clear size-dependence of the stability of Au
2
O
3
can
be inferred from Figure 3.
For the Au/TiO
2
system, the Au oxide shell was found to be
more stable on the large NPs (sample no. 2), with a 50%
Figure 2. XPS spectra (Al KR)1486.6 eV) corresponding to the
Au-4f core level of Au nanoparticles with two different average sizes:
(a and b) ∼5 nm and (c and d) ∼1.5 nm supported on SiO
2
(a and c)
and TiO
2
(b and d). The temperature-dependent spectra shown follow
the decomposition of Au oxide after UHV annealing from 200 up to
500 K (10 min).
Figure 3. Temperature dependence of the decomposition of gold oxide
studied for two different gold particle sizes ∼5 nm (open and closed
circles) and ∼1.5 nm (open and closed triangles) supported on SiO
2
(samples 1 and 3) and TiO
2
(samples 2 and 4).
I
Au
2
O
3
(θ)
I
Au
(θ)
)
K
Au
2
O
3
∫
R
1
R
2
∫
0
π
exp
(
r cos θ -
x
R
2
2
- r
2
sin
2
(θ)
λ
2
)
r
2
sin(θ)dθ dr
K
Au
∫
0
R
1
∫
0
π
exp
(
r cos θ -
x
R
2
2
- r
2
sin
2
(θ)
λ
1
)
r
2
sin(θ)dθ dr
(1)
Formation and Thermal Stability of Au
2
O
3
J. Phys. Chem. C, Vol. 112, No. 12, 2008 4679
decomposition of the Au
2
O
3
shell obtained at ∼310Kas
compared to ∼265 K for the smaller clusters (sample no. 4).
The higher surface/volume ratio present in the small clusters
should contribute to their faster reduction. Nearly complete
disappearance of the Au
3+
signal (<14% of the initial Au
2
O
3
thickness remaining) was observed at 300 K for the small
clusters on TiO
2
(sample no. 4), while a substantial Au
3+
signal
(∼66% of the initial Au
2
O
3
thickness) could still be detected
on the surface of the large clusters (sample no. 2) at 300 K and
∼25% Au
3+
signal at 500 K. A similar size-dependent trend
was observed by our group during the reduction of surface Pt-
oxides (PtO and PtO
2
) on Pt NPs supported on anatase TiO
2
powders.
61
These results are also in agreement with data
obtained by Suhonen et al.
62
on oxidized Rh clusters supported
on different oxide powders, where faster reduction rates were
measured for smaller clusters.
For the Au/SiO
2
system, the Au
2
O
3
shell on smaller NPs
displayed a slightly higher thermal stability at low temperature
than the large NPs, with a 50% decay in the Au
3+
signal on
small clusters (sample no. 3) at 430 K as compared to a similar
decay at 410 K for the large clusters (sample no. 1). On this
system, a nearly complete decomposition of the Au
2
O
3
shell
(within the experimental error margin) was observed for the
small clusters after 10 min annealing at 550 K (sample no. 3),
while a ∼0.2 nm-thick gold oxide layer was still present at that
temperature on the large clusters (sample no. 1). The dramatic
differences in the stability of Au
2
O
3
on small NPs deposited
on SiO
2
and TiO
2
will be discussed in the next section, where
the role of the cluster support will be considered.
It is noteworthy that on both substrates only a partial reduction
of the Au
2
O
3
layer is observed for our large NPs (samples 1
and 2) up to an annealing temperature of ∼550 K (10 min),
where a ∼0.2-nm-thick Au
2
O
3
shell remains on both SiO
2
and
on TiO
2
. Further annealing (>600 K, not shown) resulted in a
slow decomposition of this oxide component. However, for the
smaller NPs supported on both substrates, complete Au
3+
reduction is observed below 550 K, Figure 3. This difference
can be attributed to the presence of more than one gold oxide
species (surface and subsurface oxide) in these NPs and to a
distinct thermal stability of such species. This is discussed in
more detail in Section e. For bulk systems, typical values for
the decomposition temperature of gold oxide Au
2
O
3
49,60
are in
the range of 360-450 K. However, the existence of a more
stable form of gold oxide on Au(111), stable up to 1073 K, has
also been reported by Chesters et al.
63
Support Effects. In addition to size effects, the influence of
the oxide support on the decomposition of surface oxides on
metal nanoparticles cannot be neglected. As an example,
Schalow et al.
14
attributed the more facile reduction of preoxi-
dized, small (<3 nm) Pd NPs supported on Fe
3
O
4
/Pt(111) as
compared to larger clusters (10-100 nm) because of the stronger
metal-support interactions expected for small clusters.
In the current study, a drastic difference in the stability of
Au
2
O
3
over NPs with similar size distributions but deposited
on different substrates was observed. As an example, the Au
2
O
3
thicknesses obtained from the analysis of Figure 2a(ii) and 2b-
(ii) (large particles, ∼5 nm height) after annealing at 400 K for
10 min were 0.56 ( 0.04 nm for Au/SiO
2
and 0.22 ( 0.04 nm
for Au/TiO
2
. Similarly, after annealing the small NPs (∼1.5
nm) at 350 K for 10 min, gold oxide shells with thicknesses of
0.28 ( 0.04 nm for Au/SiO
2
(not shown) and 0.0 ( 0.04 for
Au/TiO
2
[Figure 2d(iii)] were obtained. As can be seen in Figure
3, for both particle sizes, Au
2
O
3
is more easily decomposed (at
lower temperature) when the NPs are supported on reducible
TiO
2
substrates. A tentative explanation for this distinct behavior
involves oxygen spillover from the oxidized Au NP shell to
the TiO
2
/Ti substrate. This model is based on the well-known
facile reduction of the TiO
2
support at low temperature. Oxygen
vacancies created in the TiO
2
upon sample annealing might be
replenished by oxygen from the gold oxide NP shell. Such a
mechanism is not possible in the case of SiO
2
because it only
becomes reduced well above the temperatures employed here
(maximum annealing temperature of 600 K). Oxygen vacancies
in the support are known to be favorable sites for strong
interactions with Au NPs,
8,64,65
becoming preferential sites for
the nucleation of metal clusters.
66
Furthermore, a stronger
binding of Au clusters to reduced TiO
x
films as compared to
bulk TiO
2
has been measured,
67
and charge-transfer phenomena
from oxygen vacancies in the reduced TiO
2
supports to metal
clusters have been discussed.
18,68
In our studies, it appears that
a strong metal-support interaction for the Au/TiO
2
system could
be responsible for the decomposition of gold oxide on TiO
2
at
a lower temperature than that on SiO
2
.
Interestingly, Chang et al.
69
observed that the thermal
decomposition of Ta
2
O
5
was faster when it was deposited
directly on Ti as compared to structures where an intermediate
oxygen diffusion barrier (TiN) was present (Ta
2
O
5
/TiN/Ti). They
concluded that oxygen diffuses from the Ta
2
O
5
film to the
underlying Ti layers and that the Ti contributes to its reduction.
Furthermore, when the Ta
2
O
5
film was in direct contact with
Si, no enhanced reduction was detected. Finally, it should be
noted that the influence of a reducible support on oxide
decomposition depends on the relative stability of the oxides
involved. For example, there are reports indicating a slower
decomposition rate of RhO
x
clusters when deposited on partially
reduced Ce-Zr powders.
62
Here, the reducible oxide support
is believed to act as a local supplier of oxygen, slowing down
the metal oxide decomposition. This is, however, a different
experimental system because, contrary to the case of Au on
TiO
2
, where the enthalpy of formation of Au
2
O
3
is very high,
Rh forms a much more stable oxide.
Because the size-dependent Au
2
O
3
decomposition behavior
is different for similarly sized Au clusters deposited on our two
different substrates, a distinct underlying mechanism is expected.
Assuming that oxygen vacancies in the oxide support play a
role in the stability of the Au oxide in the Au/TiO
2
system, the
higher stability of Au
2
O
3
in large Au NPs (sample no. 2) as
compared to smaller NPs (sample no. 4) could indicate that a
rate-limiting step for the decomposition of Au
2
O
3
is the diffusion
of atomic oxygen to the Au/TiO
2
interface. Following this
model, larger diffusion barriers to the NP/support interface must
be present in the case of large Au clusters.
A schematic summary of the mechanisms proposed here for
the decomposition of the Au oxide shell, formed on size-selected
Au NPs supported on SiO
2
and TiO
2
upon atomic oxygen
exposure, is given in Figure 4. In the SiO
2
case, desorption of
atomic (labeled as process 1) and molecular oxygen (upon
recombination, process 2) from the oxide shell are depicted.
Furthermore, for the large NPs, the diffusion of subsurface
oxygen to the NP surface is also considered (3). On TiO
2
, one
additional decomposition pathway (4) is proposed, where O
vacancies created in the TiO
2
substrate during annealing are
continually replenished by oxygen atom spillover from the Au
oxide NP shell.
In order to demonstrate that a relatively large number of
oxygen vacancies were present in our TiO
2
support after
annealing at low temperature, we have conducted in situ XPS
studies on the uncoated TiO
2
support after O
2
plasma exposure
4680 J. Phys. Chem. C, Vol. 112, No. 12, 2008 Ono and Roldan Cuenya
