c The Author 2009. Published by Oxford University Press
on behalf of Japanese Society of Microscopy. All rights reserved.
For permissions, please e-mail: [email protected]
Journal of Electron Microscopy 58(3): 199–212 (2009)
doi: 10.1093/jmicro/dfp016
.........................................................................................................................................................................................................................................................
Special number: Applications
Evolution of gold structure during thermal treatment
of Au/FeOx catalysts revealed by aberration-corrected
electron microscopy
Lawrence F. Allard1,∗ , Albina Borisevich1 , Weiling Deng2 , Rui Si2 ,
Maria Flytzani-Stephanopoulos2 and Steven H. Overbury3
1
Materials Science and Technology Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831-6064,
Department of Chemical and Biological Engineering, Tufts University, Medford, MA 02155,
and 3 Chemical Sciences Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831-6201, USA
∗
To whom correspondence should be addressed. E-mail: [email protected]
2
................................................................................................................................................................................................
Abstract
High-resolution aberration-corrected electron microscopy was performed
on a series of catalysts derived from a parent material, 2 at.% Au/Fe2 O3
(WGC ref. no. 60C), prepared by co-precipitation and calcined in air at
400◦ C, and a catalyst prepared by leaching surface gold from the parent
catalyst and exposed to various treatments, including use in the water–gas
shift reaction at 250◦ C. Aberration-corrected JEOL 2200FS (JEOL USA,
Peabody, MA) and Vacuum Generators HB-603U STEM instruments
were used to image fresh, reduced, leached, used and re-oxidized catalyst samples. A new in situ heating technology (Protochips Inc., Raleigh,
NC, USA), which permits full sub-Ångström imaging resolution in the
JEOL 2200FS was used to study the effects of temperature on the behavior of gold species. A remarkable stability of gold to redox treatments
up to 400◦ C, with atomic gold decorating step surfaces of iron oxide was
identified. On heating the samples in vacuum to 700◦ C, it was found that
monodispersed gold began to sinter to form nanoparticles above 500◦ C.
Gold species internal to the iron oxide support material was shown to
diffuse to the surface at elevated temperature, coalescing into discrete
nanocrystals. The results demonstrate the value of in situ heating for understanding morphological changes in the catalyst with elevated temperature treatments.
................................................................................................................................................................................................
Keywords aberration-corrected, high-resolution microscopy, gold, iron oxide,
water–gas shift catalyst, in situ heating
................................................................................................................................................................................................
Received
8
January 2009, accepted 5 March 2009, online 1 April 2009
................................................................................................................................................................................................
Introduction
Catalysis by Au has attracted considerable interest, in part
because of the high activity of supported nanoscale Au
for catalyzing certain oxidation reactions [1,2]. About two
decades of research have explored and demonstrated how
the activity and stability of Au catalysts depend on the type
of support and sample preparation methods used [3]. On
the other hand, there is still no consensus on the active
sites in Au catalysts and in particular the role of nanoparticle size and oxidation state of the active Au species. One
novel approach that has been tried to address the latter is to
modify the catalyst by leaching the surface of weakly bound
Au species by sodium cyanide solutions at pH 12, and the
resulting support is then expected to contain only strongly
bound or embedded gold species [4]. The catalytic properties and surface chemistry of the leached catalyst can
thus be studied in the absence of effects from metallic gold
nanoparticles. Using this approach, Deng et al. [5] recently
reported that well-dispersed, reduced gold clusters on ceria
or iron oxide supports provide the active sites for the lowtemperature CO oxidation reaction, while fully dispersed
oxidized gold species, strongly bound on these oxide supports, catalyze the water–gas shift (WGS) reaction. Also the
evolution of these species with time-on-stream, heating, etc.
can be investigated, again in the absence of masking effects
from neighboring particles. Deng et al. [5,6] reported that the
200
J O U R N A L O F E L E C T R O N M I C R O S C O P Y , Vol. 58, No. 3, 2009
.........................................................................................................................................................................................................................................................
Table 1. Description of samples and treatmenta
Sample
Parent 2.0 at.% Au/Fe2 O3
Leached 0.7 at.% Au/Fe2 O3
Fresh
Used
Reduced
Re-oxidized
a All
Catalyst; Pretreatment
World Gold Council reference catalyst; used as
received (had been calcined at 400◦ C)
Parent sample leached by aq. NaCN, washed and
air dried at 100◦ C
Leached sample calcined in air at 400◦ C for 2 h
‘Fresh’ sample after testing in WGS reactor for 1 h
‘Fresh’ sample after heating in 20% H2 /N2 at
10◦ C min−1 to 400◦ C, held for 0 h and cooled to
RT in H2
‘Used’ catalyst after re-oxidation in 20% O2 –He at
350◦ C for 1 h
Catalytic behavior
High activity for CO oxidation; moderately high
activity for WGS [5]
N/A
Low activity for CO oxidation; same activity for WGS
as the parent [5]
∼30% loss of activity at T = 250◦ C
High activity for CO oxidation, less active for WGS
Recovered activity for WGS, higher stability than
‘fresh’ catalyst
samples derived from WGC Au/Fe2 O3 catalyst (gold ref. catalyst no 60C).
WGS stability of Au–FeOx was superior to that of Au–CeOx .
Thus, the former catalyst may be used for high-temperature
(300–400◦ C) WGS reaction applications. The leached Au–
FeOx catalyst was active for the WGS reaction, but inactive
for CO oxidation that required pre-reduction of the sample
to light off. Pre-reduction in H2 up to 400◦ C produced reduced gold clusters on Fe3 O4 , in a very active state for CO
oxidation [5].
In the present work, we have probed the structural
changes of gold species in Au–FeOx catalysts similar to those
used by Deng et al. after various stages of preparation, including the unleached parent catalyst, leached catalyst, the
calcined active catalyst and the catalyst after being used for
the WGS reaction. For this purpose, we have used the JEOL
2200FS and VG HB-603U aberration-corrected microscopes
at Oak Ridge National Laboratory. To further monitor catalyst changes, we have also investigated their thermal evolution by in situ hot-stage microscopy.
Experimental section
The Au/Fe2 O3 (WGC ref. no. 60C) catalyst was obtained
from the World Gold Council; according to the manufacturer it was prepared by co-precipitation, it contained
2.0 at.% Au and had been calcined to 400◦ C. This parent catalyst was subjected to various pretreatments or exposures to
catalytic conditions prior to microscopy. First, it was leached
in an aqueous solution of 2% NaCN under O2 gas sparging at
room temperature (RT) and high pH (12). The ‘leached’ catalyst was washed with deionized water, and then dried for
12 h at 100◦ C. The residual gold in the leached sample was
0.7 at.% as measured by inductively coupled plasma spectroscopy (ICP). A portion of the sample was calcined in static
air from RT to 400◦ C with a heating rate of 2◦ C min−1 , and
then held at 400◦ C for 2 h, creating the ‘fresh’ catalyst. The
activity of this catalyst in the WGS reaction was measured
in a micro-reactor using 1% CO–3% H2 O–He at 250◦ C for
1 h before removing the ‘used’ sample for microscopy [5].
A portion of the fresh catalyst was heated in 20% H2 /N2 up
to 400◦ C and then cooled back to RT in H2 /N2 , creating a
reduced catalyst. Finally, a portion of the used sample was
re-oxidized by exposure to oxygen at 350◦ C. The six samples
described above and their post-treatments are summarized
in Table 1.
Catalyst microstructure was characterized at ORNL’s
Advanced Microscopy Laboratory, using two aberrationcorrected electron microscopes. A JEOL 2200FS STEM/TEM
instrument equipped with a CEOS GmbH corrector on the
probe-forming lenses was used to obtain most images discussed and presented below. A Vacuum Generators HB603U dedicated STEM fitted with a Nion Co. aberration corrector was used to obtain images shown for the fresh and reduced catalysts. Most images from both microscopes shown
here were recorded in a high-angle annular dark-field mode
(HA-ADF), which shows contrast relative to atomic number,
e.g. the Au species appear in bright contrast in comparison
to the iron oxide support. Choice between these two microscopes was generally based on operational convenience, and
images from both microscopes should provide equivalent information. However, the heating stage capability was only
available on the JEOL. Each sample was deposited, either
dry or via an alcohol suspension, onto holey carbon films
supported on copper TEM grids.
Some samples for in situ heating studies on the JEOL
2200FS were supported on novel heater devices manufactured using MEMS technology by Protochips Inc. (Raleigh,
NC, USA) [7,8]. The devices are 3-mm silicon chips fabricated with 500 μm2 thin ceramic membranes in the center,
through which current is passed to heat the membrane. A
carbon film having 1-μm-diameter holes is supported over
3-μm-diameter holes in the ceramic membrane, and the catalyst powder is deposited onto this carbon film. The heater
chip is clamped into a special-made specimen rod, with electrical leads to provide current from a Keithley 2611 power
supply. Membrane temperatures are calibrated by the manufacturer, using an optical pyrometer, and a table of current versus temperature is provided, permitting estimation
of the temperature of the sample deposited upon it. The
outstanding characteristic of this in situ heating technology
is the rapid response of the MEMS device to changes in
L. F. Allard et al. Evolution of gold structure in Au/FeOx catalysts
201
current. For example, membranes are typically heated to
temperatures of 1000◦ C or higher in 1 ms. When the current is removed, the membrane returns to RT (or a pre-set
starting temperature) in about the same time. The devices
are very stable while heating; the major effect is simply a
change in the sample height due to expansion of the membrane with increasing temperature. In some heating experiments, the sample was taken to a desired temperature for a
desired time, and then returned to an original temperature
for image recording, allowing 1 min for complete stability to
return.
Results
Micrographs of the ‘fresh’ catalyst are shown in Fig. 1 to
help understand the structure of the catalyst when it is in its
leached and active form. Figure 1a is typical of lower magnification (×1 M) micrographs that show that the iron oxide
support particles are oblong or oval shaped, 20–60 nm in
diameter, and with rounded edges and embedded dark features. The dark contrast features are consistent with internal voids in the oxide crystallites; these voids were found in
most oxide particles. In no cases were they observed to intersect the surface of a crystallite, implying that they are not
simply dimples on the surface, but instead are internal. The
Au particles appear in bright contrast compared to the loweratomic-number oxide support crystallites in the HA-ADF
image. We confirmed that the bright particles were Au
particles using energy dispersive spectra and associated elemental maps (not shown). These HA-ADF images are projections through the oxide support particles, so features may be
on the front or back surface, or embedded within the oxide
particle. It is apparent from Fig. 1a (and other micrographs)
that very few Au nanoparticles with sizes in the range of
2–4 nm exist on or within the iron oxide particles. Much
more common are highly dispersed, individual Au atoms
(probably ions) which are distributed across the oxide crystals as clearly seen in Fig. 1b. Highly dispersed, cationic Au is
consistent with XAS data described elsewhere for a sample
of this catalyst at the same stage of preparation [5]. XANES
spectra demonstrate a high fraction of cationic Au [5] and
EXAFS data show a strong contribution of Au–O shell, and
the fitted coordination number (CN) is 2.3 ± 0.5 consistent with highly dispersed cationic Au [9]. The microscopy
demonstrates that the distribution of the Au ions is not completely uniform. The increased Au density around the edges
of the voids, illustrated in Fig. 1b, suggests that much of
the Au is supported on the internal walls of the voids. The
sub-surface location of Au or encapsulation of gold inside
the matrix of iron oxide is consistent with XPS results, reported elsewhere for this sample [5]. The absence of gold
from the surface and sub-surface layers probed by XPS indicates that, for the ‘fresh’ sample, the highly dispersed as well
as the occasional Au nanoparticles are dispersed throughout the bulk of the iron oxide lattice in voids, grain bound-
Fig. 1. HA-ADF images of the fresh sample. (a) Typical area of
FeOx support particles, showing dark features consistent with internal voids; (b) higher magnification image showing nearly monodispersed Au species in bright contrast; note the decoration of voids
indicative of some Au species within the body of the support rather
than on the surface.
aries, at the interface between fused crystallites or within the
lattice.
Another view of the fresh sample is shown in Fig. 2 for
a different iron oxide crystallite. In this case, the highly
crystalline nature of the iron oxide support is evident from
its clearly delineated lattice fringes and atom columns.
Figure 2a displays an aggregated Au particle that is a discrete
single crystal, as evident from the lattice fringes seen in the
particle. A diffractogram of the Au particle, inset in Fig. 2a,
shows d-spacings of 2.281 and 2.248 Å as indicated. These
202
J O U R N A L O F E L E C T R O N M I C R O S C O P Y , Vol. 58, No. 3, 2009
Fig. 2. (a) Fresh catalyst, showing discrete crystals of Au; inset
diffractogram was computed from the higher magnification image of the particle, also shown as inset. The circled ‘reflection’ results from the supporting Fe2 O3 lattice, and is consistent with the
1.48 Å {214} interplanar spacing. The computed Au periodicities
are 2.281 and 2.248 Å as shown, indicating a constriction of the
bulk Au lattice [Au (111) = 2.35 Å]. Outlined area shows region of
panel (b).
spacings were calibrated using the Fe2 O3 lattice of the support particle as an internal standard, in this and several other
high-resolution images showing strong (hexagonal) Fe2 O3
spacings. The spot circled in Fig. 2a corresponds to the 1.48 Å
(hkl = 214 in the three-index notation) d-spacing in Fe2 O3 .
The primary interplanar spacings expected in lattice images
of Au particles are typically 2.35 and 2.04 Å, corresponding to the (111) and (200) (hkl) reflections, respectively. We
consistently observed an apparent decrease in the Au (111)
lattice spacings measured from the high-resolution images
from particles that were both small (≤3 nm) and likely to
be internal to the iron oxide particle. Comparison with a
diffractogram of the oxide lattice demonstrates that the Au
is not topotactic with the oxide. In contrast to the distinct
crystal lattice seen in the Au particle in Fig. 2a, the feature
circled in Fig. 2b (from the white rectangle area of panel a)
does not show lattice features, and instead appears to be an
amorphous aggregate of Au cations. The bright patch nearby
to the upper left of the circle is out of focus in Fig. 2b, but
an image with this patch in focus (not shown) revealed that
it too is an amorphous aggregate. It is evident that reduced
metallic Au nanoparticles co-exist with oxidized Au on this
sample.
It is of interest to compare the fresh catalyst with the parent catalyst from which it was derived. Micrographs of the
parent catalyst are shown in Fig. 3, and from this and many
other images, we note five important observations. First,
there is a high concentration of Au nanoparticles in the size
range of 1–5 nm visible in Fig. 3a, higher than what is observed on the fresh catalyst (compare with Fig. 1a). This is
entirely consistent with the higher Au loading in the parent catalyst prior to the leaching process used to prepare
the fresh catalysts. Second, on the parent catalysts there are
many Au particles on the outside surfaces of the oxide crystallites, as is proven by the existence of particles seen along
the profile of the oxide crystallites. Particles such as those
seen in the profile in the example of Fig. 3b are unmistakably on the surface of the oxide support particle rather than
being embedded within the oxide matrix. In contrast to the
parent catalyst, particles visible in the profile were not observed in the leached and fresh catalysts. In Fig. 1a, no particles are seen in the profile, although there are only a few
Au particles visible. Inspection of many micrographs from
other oxide crystallites did not reveal particles in the profile
for the fresh catalyst. Similarly, micrographs were also obtained for the leached catalyst (prior to calcination), and in
this case also there were no particles seen in the profile that
could prove them to be located on the surface. These comparisons demonstrate that the leaching had the desired effect
of removing the majority, if not all, of the small (1–5 nm)
metallic Au particles found in profusion on the surface of
the oxide particles of the parent catalyst. The third observation is that although the parent catalyst has many Au
nanoparticles, it also has highly dispersed Au atoms (or ions)
present. Evidently, the last calcination step of the preparation of the parent catalyst leaves highly dispersed Au, either
because it does not sinter into Au particles or because the
oxidizing conditions (air) of the calcination creates or preserves surface-bound Au cations. Fourth, the larger particles
on the surface show crystal lattices with spacings consistent
with bulk Au, as opposed to the decreased lattice spacings
L. F. Allard et al. Evolution of gold structure in Au/FeOx catalysts
203
Fig. 4. Au nanoparticle in the leached catalyst. Lattice structure
demonstrates the presence of metallic Au particles within the oxide particles.
Fig. 3. (a) HA-ADF image of a typical area of the parent catalyst.
Note the distribution of particles 2–5 nm in size, with some [such
as the inset in (b)] seen in the profile at the edge of the particles.
In addition to the discrete particles, the parent catalyst also shows a
uniform dispersion of single atoms and small clusters, as seen in the
higher magnification HA-ADF image of panel (b).
seen in the small, internal particles. Finally, the voids observed on the fresh sample are also observed in the parent
catalyst. This demonstrates that these voids are not created
by the leaching process.
Since the fresh catalyst had been subjected to leaching to
remove (essentially) all surface Au crystallites, it is of interest to know if the few Au nanoparticles evident on the fresh
catalysts (Figs. 1 and 2) were present on the leached sample
or whether they grew as a result of the subsequent 400◦ C
calcination. To explore this question, images of the ‘leached’
sample were obtained and studied as a function of heating
(in the microscope vacuum) at increasing temperatures and
durations. After recording initial images of the leached sample as introduced, the sample was heated to 250◦ C. Next,
images were recorded at a few different magnifications and
locations, and the sample was then heated to 500◦ C first for
2 min and then for five additional minutes. After each heating cycle, the sample was returned to 250◦ C and images
were recorded from the same locations imaged prior to heating. We note that these treatments for short times in vacuum
are not directly comparable with the samples calcined in air
for longer times, but we hypothesize that these experiments
should provide qualitative information about the processes
that occur during the extended calcination.
An image of the leached, unheated sample (Fig. 4) exhibits reduced Au nanoparticles similar to those seen on the
fresh catalysts, therefore indicating that such particles do remain after the leaching and are not only the result of the
400◦ C calcination. Also, their presence after leaching implies that they are within the iron oxide particle. Many images were recorded after heating to 250◦ C as typified by the
image sequence in Figs. 5–7. A number of Au particles at
250◦ C showed strong lattice fringe contrast, as for example,
the sample area shown in Figs. 5a, 6a, b and 7, reinforcing
the conclusion that reduced Au particles were present in the
oxide crystals. Subsequent heating to 500◦ C for 2 min within
the microscope had observable effects on the size, location
or thickness of the observed Au particles, but such changes
204
J O U R N A L O F E L E C T R O N M I C R O S C O P Y , Vol. 58, No. 3, 2009
Fig. 5. HA-ADF images of the leached catalyst showing changes as
a result of in situ heating. Very little change was seen with time at
250◦ C. Panel (a) shows a starting image recorded after a few minutes at temperature. Only slight changes were seen after 2 min at
500◦ C (panel b), but an additional 5 min at 500◦ C showed void
shrinkage, coalescence and growth of Au nanoparticles, and the diffusion of Au species to the surface to form discrete nanocrystals
(arrowed). Inset box shows the area of Fig. 6.
were more pronounced after heating for 7 min (Fig. 5b and
c). This treatment yielded clear changes in the spatial distribution of Au particles, the number and size distribution of
Au particles, the apparent sharpness of the particles and in
the visible void structure of the iron oxide crystallites. The
voids change shape, shrink, or in some case disappear altogether. Many of the Au atoms and clusters of less than
∼1 nm are lost after the 7-min anneal, presumably ending
up in the 2–4 nm Au particles that remain. This verifies that
sintering occurs. Loss of small Au particles is more clearly
seen in Fig. 6 which shows one region at two different magnifications, before and after heating to 500◦ C.
Two additional effects of the heating are noted. First, in
the initial sample (Fig. 5a), none of the discrete Au nanoparticles are visible along the edges of the oxide crystallites.
However, after the longer anneal at 500◦ C, several Au particles are visible in the profile around the edges of the oxide particles (Fig. 5c). Particles thus observed must be on the
outside of the oxide particles. This suggests that the Au particle growth process may involve not only surface diffusion
but also diffusion of Au to the outer surface of the oxide particles.
Secondly, this re-distribution of Au between inside and outside of the oxide particles is reinforced by experiments in
which the focal point of the electron beam is swept through
the support crystallite as illustrated by the two micrographs
shown in Fig. 7. Two Au particles 3–4 nm in diameter are
visible and appear to be adjacent on the oxide support particle. However, the lattice fringes observable are not in focus for both particles simultaneously. By sweeping the focus
(via objective lens current change) from in front of the oxide crystallite to behind it in steps of 1 nm, it is possible to
bring the Au lattice of either of the nanoparticles into focus.
From the focus parameters of the microscope, it was determined that the nanoparticle on the lower left is ∼28 nm farther back than its neighboring particle, as viewed down the
electron beam. The particle behind appears to be contained
within a bulk void after heating to 500◦ C (Fig. 6d). Further
evidence was obtained by collecting a sequence of images
at the location of the image shown in Fig. 6b (after annealing to 250◦ C), obtained by varying the focal position of the
beam stepwise through the particle. From a video produced
by stacking equivalent sequential images, it was apparent
that several pairs of particles that appear to be adjacent or
overlapping are offset vertically. Therefore, for the leached
catalysts, some Au nanoparticles must be embedded within
the oxide.
The two particles described above, and shown in appropriate focus in Fig. 7, have another interesting property.
Although they are unconnected, they are both likely embedded in the oxide matrix, because the crystal lattices of
the particles are aligned. This seemingly improbable coincidence can only occur because the particles are topotactically
aligned with the iron oxide support. Indeed the Au lattice is
aligned with the lattice planes visible in the support, as seen
in Fig. 7a. From further analysis of Fig. 7 and its companion images, it was also apparent that the void edges seen for
L. F. Allard et al. Evolution of gold structure in Au/FeOx catalysts
205
Fig. 6. Details of the area of the leached sample shown as inset in Fig. 5b. (a) and (b) show the area as annealed at 250◦ C; (c) and (d) show
the area after annealing at 500◦ C for 7 min. Images were recorded at approximately the same level of focus. The loss of the highly dispersed
Au species and shrinkage of the voids is clearly seen after the 7-min anneal. The central Au nanoparticle shows an alignment with the lattice
of the FeOx support.
the particle in focus in Fig. 7a were sharpest, but not for the
particle in focus in Fig. 7b. It is likely that the first particle is
located at the bottom surface of the upper void visible in figure (bottom being relative to the image as displayed), while
the second particle is likely located at the back surface of the
lower void (back relative to the beam direction).
High-resolution images were also recorded from the ‘used’
catalyst that had been previously exposed to reaction conditions at 250◦ C for over an hour. This sample was heated
in the microscope at a sequence of increasing temperatures,
holding for 1 min at each temperature before cooling and
performing the microscopy. A set of micrographs from the
same area is shown for six annealing temperatures in Fig. 8.
There are ‘large’ Au particles of ∼2 nm diameter that are accompanied in the micrographs by smaller diffuse and weakly
scattering patches with a width of ∼1 nm. (XPS detected the
presence of gold on the Fe3 O4 surface of the used sample
[5].) The latter are interpreted as small Au clusters that are
not in optimal focus. Micrographs of the as-inserted, used
sample (Fig. 8a) prior to heating in the microscope exhibit
the void structure and Au size distribution characteristic of
the leached (compare Fig. 5a) or ‘fresh’ catalyst (compare
Fig. 2). Apparently the calcination and the subsequent use
in a catalytic reactor did not cause substantive changes in
206
J O U R N A L O F E L E C T R O N M I C R O S C O P Y , Vol. 58, No. 3, 2009
Fig. 7. Images of the leached catalysts heated to 250◦ C sample,
at two different focal positions. Particle 1 is in focus in panel (a)
and particle 2 is in focus in panel (b). From measured focus parameters, it is estimated that particle 1 is ‘in front’ of particle 2 by
∼28 nm. Particle 1 appears to be positioned at the bottom of the
upper void, while particle 2 is apparently located at the back of
the lower void (based on examination of a number of images in
the focal sequence). The image also shows that the lattices of these
two nanoparticles are aligned, which occurs because they are both
topotactic with the Fe2 O3 host lattice.
the catalyst morphology or the mean observed Au particles.
However, as suggested by the through-focus experiments
described above, the distribution of Au between the surface
and the bulk of the oxide particles may vary.
Figure 8 shows the results of heating the samples to as
high as 700◦ C. Heating to 400◦ C causes loss of many of
the diffuse particles as seen in Fig. 8 and from micrographs
at lower magnification (not shown). As seen from the micrographs of other crystallites, this loss occurs generally between 400 and 500◦ C, although a precise temperature range
cannot be specified. As the temperature increases, the larger
particles gradually sharpen, becoming what appear to be
thin ‘rafts’ of reduced Au, and Au lattice planes become visible on several of the particles. The ability to observe the
Au lattice structure in several Au clusters simultaneously indicates that they are all within the same focal plane. It is
probable that the focal plane corresponds to the surface of
the oxide particle. In some cases, pre-existing large Au particles (e.g. 4 nm across) are seen to change aspect and also
crystallographic orientation as illustrated by the example of
Fig. 9. The lattice structure of the underlying oxide support
remains essentially unchanged between the two images. The
two particles (one at a significantly different focus level) appear to change position and morphology. The proximity to
the particle edge and the dramatic change in morphology
of the central particle suggests that the particles may reside
on the surface of the support.
Figure 10 shows micrographs obtained for the reduced
catalyst. Although this catalyst had been reduced in H2 at
400◦ C and cooled in H2 , it had necessarily been exposed
to air at RT during the lengthy period prior to microscopy.
This sample exhibits obvious faceted Au particles in the size
range of 3–4 nm, with some smaller particles seen in the
profile indicating that they reside on the surface of the particle. On a finer scale (Fig. 10b), many Au clusters 0.5–1 nm
are observed, most of which show lattice structure indicating
discrete crystallites rather than disordered clusters. There is
also evidence of atomically dispersed Au remaining on the
sample. In Fig. 10c, these species are observed to form rows
apparently decorating step edges or other structural features
of the oxide support. Voids still exist in the oxide particles.
The distribution of the Au differs from the fresh catalyst
(e.g. compare with Figs. 1 or 2) in that there are surface Au
particles in evidence only on the reduced catalyst, while
atomically dispersed Au is present on both samples but is
apparently a minority species on the reduced catalyst. The
distribution of Au on the latter, more closely resembles that
of the used catalyst (compare to Fig. 8a) or of the leached
sample after heating to 500◦ C (compare to Fig. 5c).
Finally, experiments were also performed on a re-oxidized
sample, that is, a sample that had been used for catalysis and
then removed from the reactor and heated in a 20% O2 /He
stream at 350◦ C for 1 h. From XANES analysis (not shown
here) we found that the reduced gold species after the WGS
reaction were partially re-dispersed and cationic after this
re-oxidation process [6]. High-resolution HA-ADF images
(Fig. 11) from this sample showed that there is a distribution
between isolated Au ions, thin irregular patches of Au (ions
or atoms) and predominantly Au nanoparticles (presumably metallic). Many of the Au particles are on the outside
L. F. Allard et al. Evolution of gold structure in Au/FeOx catalysts
207
Fig. 8. HA-ADF images of the ‘used’ catalyst as received (at RT), and after in situ annealing sequentially to increasingly higher temperatures.
Slight offset occurs between micrographs, so the same particle is circled in the first four micrographs. Images were recorded at 10 Mx, showing
the remarkable stability of the heating holder.
208
J O U R N A L O F E L E C T R O N M I C R O S C O P Y , Vol. 58, No. 3, 2009
Fig. 9. HA-ADF images showing the behavior of Au nanocrystals
in the ‘used’ catalyst (a) after annealing to 300◦ C and (b) 550◦ C.
The lattice structure of the support particle suggests that its position
is essentially unchanged between the two images. The positions of
the two particles and the change in morphology clearly seen in the
central particle suggest that they are on the surface of the sample.
surface of the oxide particles since they are found in the profile on the edges of the oxide crystallites as typified by the
images in Fig. 11a. The isolated Au atoms were not as prolific as on the fresh sample, but they were present as shown
in Fig. 11b that shows highly dispersed Au on a region of
the Fe2 O3 crystal that happened to be oriented such that the
Fe columns align with the beam direction. Interestingly, the
Fig. 10. HA-ADF images of the reduced catalyst show faceted crystals 3–4 nm in size, with many finer clusters dispersed throughout.
In panel (a), small clusters are seen in the profile at the edge of the
particle, indicating that they reside on the surface. In panel (b), small
clusters show lattice structure, indicating that they are discrete crystallites. In panel (c), Au species form rows suggesting the decoration
of steps on the surface of the oxide support
L. F. Allard et al. Evolution of gold structure in Au/FeOx catalysts
209
void structure seen on the leached sample was still visible
on the re-oxidized sample, Fig. 11c, but it is not possible to
determine with certainty that the voids contain Au particles.
Discussion
Fig. 11. HA-ADF images of the re-oxidized catalyst. There is a distribution between isolated Au ions, thin irregular patches of Au (ions
or atoms) and crystalline nanoparticles seen on the surface as in
panel a. Panel b shows highly dispersed Au species on an oxide particle oriented with the beam parallel to discrete Fe-rich columns.
Panel c shows that the void structure still exists in the re-oxidized
catalyst.
Catalytic activity of Au–Fe2 O3 catalysts has been reported
previously for CO oxidation [10–14], selective CO oxidation
[10,15] and for the WGS reaction [5,16,17]. Considerable
variation in the activity for CO oxidation has been noted and
has been attributed to differences in synthesis conditions, especially the pH used during preparation by co-precipitation
and subsequent calcination temperatures [13]. For both CO
oxidation and WGS, uncalcined catalysts dried at 90–120◦ C
are typically quite active, and calcination may diminish their
activity [12,13]. There are conflicting reports about how
strongly the CO oxidation activity depends upon particle size
[10,11,19]. This discrepancy is possibly attributable to subtle
differences in preparation conditions and difficulty in assessing both size and activity when a distribution of metal particle sizes, including atomically dispersed Au atoms or ions
and 0.5 nm bi-layer rafts of Au, coexists on the support [19].
Deng et al. [5] have performed activity measurements for
both CO oxidation and WGS reactions on the parent and the
‘fresh’ catalysts reported here, and they find different trends.
For the CO oxidation reaction, the parent catalyst is active
but the ‘fresh’ catalyst is much less active. The latter comprises only gold cations as identified by XANES [5]. However, reduction of the fresh catalyst re-establishes the activity of the parent catalyst to within a factor of 2 when normalized for the decreased Au loading [5]. For the WGS reaction,
the fresh catalyst has the same activity as the parent catalyst when rates are normalized by catalyst surface area or by
catalyst weight. After reduction, the activity of both the parent and the fresh catalysts is similar [5,6]. Evidently, the two
reactions are catalyzed by different sites. Therefore, the key
issue is to correlate the structure of the catalysts with their
activity.
These catalytic results can be explained by the structural
observations. Consider first the CO oxidation reaction. The
parent catalyst is very active for this reaction and is characterized by many reduced 1–5 nm Au particles on the
surface of the iron oxide crystallites. These small Au
nanoparticles are the most striking Au features on the parent catalyst when viewed at the relatively low magnification used in Fig. 3a. However, the surface is also covered
with Au atoms/cations (Fig. 3b). Which of these is associated with high activity for CO oxidation? Leaching the catalyst, followed by calcination in air, removes most of the
1–5 nm Au nanoparticles (Fig. 1a) from the catalyst, leaving many monodispersed Au atoms (Fig. 1b) and rendering
it largely inactive for CO oxidation. This comparison suggests that some fraction of the gold nanoparticles removed
by leaching were the active sites for CO oxidation and that
the atomically dispersed gold is not active for this reaction.
By comparison, the WGS activity is relatively unchanged by
210
J O U R N A L O F E L E C T R O N M I C R O S C O P Y , Vol. 58, No. 3, 2009
leaching suggesting that it is the atomically dispersed Au that
catalyzes this reaction. Metallic particles are re-established
on the surface of the fresh catalyst upon annealing in vacuum (not done in catalytic experiments), or after reaction
up to 350◦ C or by heating in hydrogen up to 400◦ C. Either
of the latter two treatments renders the catalysts once again
active for CO oxidation [5]. Therefore, comparison of the
catalytic results with the present microscopy associates the
catalytic activity for CO oxidation with metallic Au nanoparticles and the WGS activity with highly dispersed cationic
Au.
The size of the CO-oxidation active gold nanoparticles supported on iron oxide was recently examined by
Herzing et al. [19]. These authors have described microscopy
of Au–FeOOH catalysts prepared by co-precipitation to compare two catalysts that, although prepared similarly, demonstrated significant differences in the CO oxidation activity.
By high-resolution, aberration-corrected microscopy, they
were able to discriminate monolayer structures (0.2–0.3 nm)
and bi-layer structures (0.5 nm) from atomically dispersed
Au and Au nanoparticles (>1 nm particle size) and conclude
that the bi-layer structures most closely correlated with the
CO oxidation activity.
In the present work, the purpose of leaching is intended to
remove metallic Au nanoparticles from the support surface,
leaving only atomically dispersed cationic Au. The extent to
which this is achieved is evident from comparison of Fig. 3
(parent) and Fig. 5a (leached). Leaching removes most, but
not all, of the larger Au particles, and the remaining size
distribution is altered. However, analysis of the microscopy
results suggests that the metallic Au particles remaining on
the fresh catalyst are trapped in internal voids, included in
the matrix or at the interface between oxide crystallites. We
speculate that these particles are inhibited from removal by
the leaching procedures and from participation in the catalytic activity. Besides the Au particles, the leached (and
fresh) catalyst also has large quantities of individual Au ions
present, visible in the micrographs (Fig. 3a). The atomic dispersion is evident from the images of both the leached catalyst (Fig. 3a) and the ‘fresh’ catalyst (Figs. 1 and 2), and the
cationic nature is indicated from EXAFS and XANES. These
species are not detected by XPS [5], which may be due to a
combination of two factors: (i) they are fairly low in concentration, and (ii) many of them also may be located in internal
voids, or trapped at interfaces, or occupy defect sites inside
the oxide matrix.
A key feature demonstrated by the electron microscopy
results is the presence of Au particles throughout the oxide particles. The evidence is from through-focus measurements and the in situ heating experiments. The z-difference
between neighboring particles in Fig. 7, as measured in a
through-focus image series, is smaller than the diameter of
the oxide crystallite indicating that (at least) one of the particles is embedded within the crystallite. In situ heating experiments provide evidence that these Au particles diffuse
outward during heating in vacuum. The heating sequence
in Fig. 5 shows visually how the Au particles are moving
as a result of heating, and the increase in numbers of Au
particles observed at the margins of the oxide particle
demonstrates that new particles are growing on the exterior surface where none had existed prior to heating. The
increasing clarity of the Au particles as the sample is heated
in Fig. 8 at a fixed focal position is also attributed in part to
the migration of the Au particles to the surface of the oxide, rendering more of them to be in sharp focus since they
would be in approximately the same plane of focus. This sequence also demonstrates the increasing ordering and improved faceting of the Au particles. These conclusions could
not be obtained by comparing two differently heated samples because it would be impossible to accurately compare
representative areas. They demonstrate the value of in situ
heating as a technique in microscopy.
The presence of reduced Au particles within the oxide
particles may not be widely appreciated. We assume that
the nature of the co-precipitation process used for the synthesis of the parent catalyst is responsible for this threedimensional mixing between the nascent iron precursors
with the cationic Au in the synthesis solution. The present
results indicate that not just cationic Au but even reduced
Au nanoparticles are formed inside the oxide particles either
during the synthesis, drying or first calcination. Therefore,
this internal Au is not only related to the leaching process
used to prepare the catalysts discussed here, but must also
exist in the unleached parent catalyst. Following synthesis,
the redistribution of Au between the bulk and the surface is
activated by thermal treatments and competes with surface
processes including growth and oxidation. These redistribution processes may be expected to be altered by conditions
of the pretreatments including the nature of the ambient
(oxidizing, reducing, water concentration, vacuum) temperature or even whether the ambient gas is static or flowing
[19]. Using the present techniques, it is not possible to locate the highly dispersed Au cations to determine if they
are on the surface or within the oxide crystallite. Nor is
it possible to locate whether any given nanoparticle is on
the front or back surface or inside the particle, unless the
particle happens to be clearly on the profile of a support
particle. It is notable that the computed diffractogram data
suggest that the encapsulated Au particles have decreased
d-spacings compared to the 2.35 Å (111) d-spacing of bulk
Au. Decreased Au–Au bond distances have been observed
by EXAFS measurements for small (<2 nm) nanoparticles of
Au [20].
Dark features consistent with internal voids in the iron
oxide crystallite supports were observed in all catalysts if
the sample was not further heated to high temperatures.
These voids do not result from the leaching procedures because they are also seen on the parent catalyst. However,
these voids were not stable to high-temperature annealing in
vacuum. From a comparison of sets of micrographs obtained
from different areas after heating at sequentially higher temperatures (not shown), it was found the voids decreased
L. F. Allard et al. Evolution of gold structure in Au/FeOx catalysts
in size, sharpened and then disappeared at temperatures in
the range of 450–500◦ C. It is assumed that the voids collapse by diffusion of material from the oxide crystallite into
the voids. Such a void collapse can occur without readily
visible changes in the particle diameter. For example, Fig. 3
suggests that typical oxide crystallites may be 60 nm in diameter with voids of ∼10 nm. A collapse of a single spherical
void by densification within a spherical crystallite with these
diameters can be calculated to cause <0.2% decrease in the
crystallite diameter. Besides the void collapse, other changes
related to oxide diffusion also occur. Changes occur in the
shape of the edges of the crystallites and these are especially apparent at the interfaces between particles that
close up and sinter together as the heating temperature
increases.
From powder X-ray diffraction analysis, we found that a
phase transformation for the iron oxide took place during
the WGS reaction. The ‘fresh’ catalyst support is α-Fe2 O3
(hematite, hexagonal), while the ‘used’ or the reduced catalyst were found to have converted to Fe3 O4 (magnetite,
cubic). Each of these FeOx structures can be converted to the
other via ‘topotactic transformation’ [21] under controlled
oxidation and/or reduction processes [22]. Microscopy
results show that diffusion and transformation of the Au
particles (heated under vacuum conditions) occur without a substantial transformation of the oxide crystallites,
as demonstrated in the sequence of Fig. 8. A phase transformation of the oxide might be expected to disrupt the
Au particles and lead to changes in the shape of the particles. Extended heat treatments or exposure during reactions
may lead to such a transformation of the oxides but the
oxide transformation is not necessary to promote the outward diffusion or the observed changes observed in the Au
distributions.
There have been many previous studies of Au catalysts by
high-resolution electron microscopy. Akita et al. examined
ceria-supported Au catalysts and monitored the growth of
Au particles during heat treatments [23]. Their samples were
made by Au vapor deposition onto a thinned polycrystalline
CeO2 disk. Samples could be externally heated in air or in H2
but were exposed to air during transfer into the microscope.
They observed the Au particle growth by Ostwald ripening
during heating in air, and the growth was inhibited when
the samples were heated in H2 . Where particle orientation
could be obtained, Au particles were aligned with their (111)
faces growing on (111) faces or on (100) faces of the CeO2
[23,24]. Akita et al. also noted that beam exposure during
TEM could cause shrinking of the Au particles [24]. They attributed this to either beam-induced temperature rise or enhanced diffusion caused by beam-induced oxygen defects on
the CeO2 . Our measurements were made in the dark-field
STEM mode rather than the bright-field TEM mode used
by Akita et al. The differences in these two imaging techniques and the different beam energies used (200 keV in the
present STEM versus 300 keV for TEM by Akita et al.) make
it difficult to predict which is more sensitive to beam heat-
211
ing and damage effects. However, in our measurements we
did not observe significant beam-induced effects. Repeated
images of Au nanoparticles were obtained without observing changes. In contrast, in situ heating did induce obvious
changes in the Au particles shape, as well as void collapse
and Au diffusion.
Summary
High-resolution aberration-corrected electron microscopy
was performed on a series of catalysts derived from a parent Au/Fe2 O3 (WGC ref. no. 60C) catalyst prepared by coprecipitation and a catalyst prepared by leaching surface Au
from the parent catalyst and exposed to various treatments.
The following are the major findings:
r
r
r
r
r
r
r
The parent catalyst shows a mixture of atomically dispersed Au atoms or cations and large number of welldefined crystalline Au nanoparticles.
After leaching, the catalyst contains a high proportion of
atomically dispersed Au atoms or cations, strongly bound
to iron oxide. However, it still also contains 2–4 nm Au
particles that appear to be incorporated throughout the
oxide particles, including within internal voids in the
oxide.
Small Au particles incorporated within the oxide matrix
have a contracted lattice compared to bulk Au.
The fresh catalyst, calcined to 400◦ C and active for the
WGS reaction, closely resembles the leached catalyst.
Although calcination at 400◦ C does not sinter the atomically dispersed Au present in the leached Au, heating to
500–700◦ C (in vacuum) caused sintering and growth of
Au nanoparticles, at least some of which are on the surface of the oxide particles.
Used and reduced samples show similar features; the distribution of gold is predominantly in 2–4 nm particles,
many clearly lying on the surface of the iron oxide, but a
minority of atomically dispersed gold is also present, decorating step surfaces.
The re-oxidized sample shows a distribution between isolated Au ions, thin irregular patches of Au (ions or atoms)
and Au nanoparticles.
Thus, this technique not only allows us to follow structural
evolution of gold as a function of thermal treatment in situ,
but also to identify and distinguish structural effects produced ex situ via redox, and catalytic treatments of this type
of catalyst with atomic-scale resolution.
Funding
U.S. Department of Energy (DE-AC05-00OR22725 to L.A.,
A.B. and S.O; DE-FG02-05ER15730 to W.D., R.S. and
M.F.-S.).
Acknowledgements
Research sponsored by the Division of Chemical Sciences, Geosciences,
and Biosciences, Office of Basic Energy Sciences, U.S. Department of
Energy, under contract DE-AC05-00OR22725 with Oak Ridge National
212
J O U R N A L O F E L E C T R O N M I C R O S C O P Y , Vol. 58, No. 3, 2009
Laboratory, managed and operated by UT-Battelle, LLC. The High Temperature Materials Laboratory microscopy facility at ORNL is supported
by the Assistant Secretary for Energy Efficiency and Renewable Energy,
Office of Vehicle Technologies, U.S. Department of Energy. Research
at Tufts University sponsored by the Office of Basic Energy Sciences,
Hydrogen Fuel Initiative Program, grant no. DE-FG02-05ER15730, U.S.
Department of Energy.
References
........................................................................................................................
1 Bond G C, Louis C, and Thompson D T (2006) Catalysis by Gold
(Imperial College Press, London).
........................................................................................................................
2 Bond G C, and Thompson D T (1999) Catalysis by gold. Catal. Rev.-Sci.
Eng. 41: 319.
........................................................................................................................
3 Kung M C, Davis R J, and Kung H H (2007) Understanding Aucatalyzed low-temperature CO oxidation. J. Phys. Chem. C 111:
11767.
........................................................................................................................
4 Fu Q, Saltsburg H, and Flytzani-Stephanopoulos M (2003) Active
nonmetallic Au and Pt species on ceria-based water-gas shift catalysts. Science 301: 935.
........................................................................................................................
5 Deng W, Carpenter C, Yi N, and Flytzani-Stephanopoulos M (2007)
Comparison of the activity of Au/CeO2 and Au/Fe2 O3 catalysts for
the CO oxidation and the water-gas shift reactions. Top. Catal. 44:
199.
........................................................................................................................
6 Deng W (2008) Ph.D. Thesis (Tufts University, Massachusetts).
........................................................................................................................
7 Allard L F, Bigelow W C, Nackashi D P, Damiano J, and Mick S E
(2008) A new paradigm for ultra-high-resolution imaging at elevated temperatures. Microsc. Microanal. 14: 792.
........................................................................................................................
8 Allard L F, Bigelow W C, Jose-Yacaman M, Nackashi D P,
Damiano J, and Mick S E (2009) A new MEMS-based system for
ultra-high-resolution imaging at elevated temperatures. Microsc. Res.
Tech. 72: 208.
........................................................................................................................
9 Si R, Deng W, and Flytzani-Stephanopoulos M EXAFS measurements of Au catalysts. Unpublished data.
........................................................................................................................
10 Schubert M M, Hackenberg S, Van Veen A C, Muhler M, Pizak V,
and Behm R J (2001) CO oxidation over supported gold catalysts –
‘inert’ and ‘active’ support materials and their role for the oxygen
supply during reaction. J. Catal. 197: 113.
........................................................................................................................
11 Haruta M, Tsubota S, Kobayashi T, Kageyama H, Genet M J,
and Delmon B (1993) Low-temperature oxidation of CO over
gold supported on TiO2 , alpha-Fe2 O3 and Co3 O4 . J. Catal. 144:
175.
........................................................................................................................
12 Moreau F, and Bond G C (2006) CO oxidation activity of gold catalysts supported on various oxides and their improvement by inclusion of an iron component. Catal. Today 114: 362.
........................................................................................................................
13 Al-Sayari S, Carley A F, Taylor S H, and Hutchings G J (2007)
Au/ZnO and Au/Fe2 O3 catalysts for CO oxidation at ambient temperature: comments on the effect of synthesis conditions on the
preparation of high activity catalysts prepared by coprecipitation.
Top. Catal. 44: 123.
........................................................................................................................
14 Hutchings G J, Hall M S, Carley A F, Landon P, Solsona B E,
Kiely C J, Herzing A, Makkee M, Moulijn J A, Overweg A et al.
(2006) Role of gold cations in the oxidation of carbon monoxide
catalyzed by iron oxide-supported gold. J. Catal. 242: 71.
........................................................................................................................
15 Schubert M M, Venugopal A, Kahlich M J, Pizak V, and Behm R J
(2004) Influence of H2 O and CO2 on the selective CO oxidation in
H2 -rich gases over Au/alpha-Fe2 O3 . J. Catal. 222: 32.
........................................................................................................................
16 Andreeva D, Idakiev R V, Tabakova T, and Andreev A (1996) Lowtemperature water–gas shift reaction over Au/a-Fe2 O3 . J. Catal. 158:
354.
........................................................................................................................
17 Andreeva D, Tabakova T, Idakieva V, Christov P, and Giovanoli R
(1998) Au/a-Fe2 O3 catalyst for water-gas shift reaction prepared by
deposition-precipitation. Appl. Catal. A Gen. 169: 9.
........................................................................................................................
18 Hodge N A, Kiely C J, Whyman R, Siddiqui M R H, Hutchings G
J, Pankhurst Q A, Wagner F E, Rajoram R R, and Golunski S E
(2002) Microstructural comparison of calcined and uncalcined gold/
iron-oxide catalysts for low-temperature CO oxidation. Catal. Today
72: 133.
........................................................................................................................
19 Herzing A A, Kiely C J, Carley A F, Landon P, and Hutchings G J
(2008) Identification of active gold nanoclusters on iron oxide supports for CO oxidation. Science 321: 1331.
........................................................................................................................
20 Miller J T, Kropf A J, Zha Y, Regalbuto J R, Delannoy L, Louis C,
Bus E, and van Bokhoven J A (2006) The effect of gold particle size
on Au-Au bond length and reactivity toward oxygen in supported
catalysts. J. Catal. 240: 222.
........................................................................................................................
21 Becker P P, Heizmann J J, and Baro R (1977) Relations topotaxiques entre des cristaux naturels d’hématite et le magnétite qui en
est issue par réduction à basse température. J. Appl. Crystallogr. 10:
77.
........................................................................................................................
22 Jia C-J, Sun L-D, Yan Z-G, Pang Y-C, You L-P, and Yan C-H (2007)
Iron oxide tube-in-tube nanostructures. J. Phys. Chem. C 111: 13022.
........................................................................................................................
23 Akita T, Tanaka K, and Kohyama M (2008) TEM and HAADF-STEM
study of the structure of Au nano-particles on CeO2 . J. Mater. Sci. 43:
3917.
........................................................................................................................
24 Akita T, Okumura M, Tanaka K, Ohkuma K, Kohyama M, Koyanagi
T, Date M, Tsubota S, and Haruta M (2005) TEM observation of gold
nanoparticles deposited on cerium oxide. J. Mater. Sci. 40: 3101.