The Key Role of Nanocasting in Gold-based Fe2O3 Nanocasted Catalysts for Oxygen Activation at the Metal-support Interface

Article abstract of DOI:10.1002/cctc.201900210

The total oxidation of propane, a representative Volatile Organic Compound, has been studied using gold-based α-Fe₂O₃ catalysts. Catalysts consisting of gold nanoparticles confined in nanostructured Fe₂O₃ prepared by a nanocasting route present the highest catalytic activity for propane total oxidation, and the activity is significantly greater than those of gold-based catalysts where iron oxide supports are prepared by other conventional methods, such as calcination. Detailed characterisation and Density-functional theory (DFT) studies have been undertaken in order to explain the enhancement in catalytic properties. The presence of confined gold nanoparticles on the nanocast Fe₂O₃ facilitates the production of highly reactive oxygen vacancies at the metal-support interface, increasing the catalyst performance. Both the development of a microporous/mesoporous structure in the iron oxide support and the presence of a mixed surface phase of Si and Fe oxides, seem to be key parameters, being both features inherent in the nanocasting process from silica templates. Additionally, the catalytic activity is enhanced due to other positive effects, which are closely related to the nanocasting preparation method: i) a higher contact surface area between partially confined small gold nanoparticles in the internal mesoporosity of the nanostructured support and the metal oxide and; ii) a more reducible support due to the presence of more active surface lattice oxygen.

Full text of DOI:10.1002/cctc.201900210

Accepted Article
Title: The key role of nanocasting in gold-based Fe2O3 nanocasted
catalysts for oxygen activation at the metal-support interface
Authors: Tomás García, Jose Manuel López, Benjamín Solsona, Rut
Sanchis, David J Willock, Thomas Davies, Li Lu, Qian He,
Christopher J Kiely, and Stuart H. Taylor
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: ChemCatChem 10.1002/cctc.201900210
Link to VoR: http://dx.doi.org/10.1002/cctc.201900210
A Journal of
www.chemcatchem.org

ChemCatChem
10.1002/cctc.201900210
FULL PAPER
The key role of nanocasting in gold-based Fe
2
O nanocasted
3
catalysts for oxygen activation at the metal-support interface
[a]
[a]
[b]
[b]
[c]
Tomás García,* José M. López, Benjamín Solsona,* Rut Sanchis, David J. Willock, Thomas E.
Davies, Li Lu, Qian He, Christopher J. Kiely, and Stuart H. Taylor*[c]
[
c]
[d]
[c]
[c,d]
Abstract: The total oxidation of propane, a representative Volatile
Organic Compound, has been studied using gold-based α-Fe
catalysts. Catalysts consisting of gold nanoparticles confined in
nanostructured Fe prepared by a nanocasting route present the
been driven by new developments in catalyst design, and by the
O
2 3
greater availability of gold compared to platinum and palladium.
For gold catalysts, the oxidation state has been reported to
be a key factor in VOC oxidation activity, although there is a
surprising lack of agreement on precise details from studies in the
literature. From a number of reports cationic gold is assumed to
be the active site,[3], [4] whereas others propose that metallic gold
2 3
O
highest catalytic activity for propane total oxidation, and the activity is
significantly greater than those of gold-based catalysts where iron
oxide supports are prepared by other conventional methods, such as
calcination. Detailed characterization and Density-functional theory
δ+
[5]-[7]
is more reactive than Au species.
In several studies, the co-
existence of Auδ+ and Au has been postulated as the most active
combination;[8]-[10] with metallic gold involved in the adsorption and
activation of the organic compound, whilst the excess oxygen
associated with cationic gold participates in oxygen activation. A
positive feature of adding gold is to improve the redox properties
of metal oxides, leading to an enhanced reactivity compared to
the support alone. However, in order to achieve satisfactory
results, the preparation method and the nature of the support
have to be controlled appropriately.
0
(
DFT) studies have been undertaken in order to explain the
enhancement in catalytic properties. The presence of confined gold
nanoparticles on the nanocast Fe facilitates the production of
2 3
O
highly reactive oxygen vacancies at the metal-support interface,
increasing the catalyst performance. Both the development of a
microporous/mesoporous structure in the iron oxide support and the
presence of a mixed surface phase of Si and Fe oxides, seem to be
key parameters, being both features inherent in the nanocasting
process from silica templates. Additionally, the catalytic activity is
enhanced due to other positive effects, which are closely related to
the nanocasting preparation method: i) a higher contact surface area
between partially confined small gold nanoparticles in the internal
mesoporosity of the nanostructured support and the metal oxide and;
ii) a more reducible support due to the presence of more active
surface lattice oxygen.
The characteristics of gold catalysts differ significantly to
those of Pt and Pd, as it appears that the role of the support is
much more important for gold catalysts.[
11],[12]
The primary role of
the metal oxide supports for Pt and Pd-based catalysts has been
considered to be related to their capacity to impart the noble metal
particles with the right crystallite size and oxidation state.[13] In
contrast for gold-based catalysts, the situation may be
considerably more complex, with catalytic performance being
dictated by both the properties of the support and the gold
components, which often lead to activity enhancements due to
synergistic effects.[ For instance, the high activity observed for
Introduction
2]
Most of the efficient catalysts for the total oxidation of
volatile organic compounds are based on noble metals, primarily
palladium and platinum. However, gold is scarcely used in
industrial VOC catalyst formulations, due to its lower reactivity and
lower stability at high reaction temperatures. Interestingly,
research undertaken for the past 5-10 years has led researchers
to re-consider the possible contribution of Au nanoparticles in
catalysts for VOC emission control by oxidation.[1],[2] This has
x
Au/FeO catalysts in the oxidation of a series of VOCs has been
related to the presence of highly dispersed gold. Small gold
nanoparticles have been proposed to modify the characteristics
of the iron oxide support by decreasing the strength of the Fe-O
bonds, thus increasing the mobility of the lattice oxygen.[14] The
presence of gold can also in principle distort the iron oxide lattice
which may positively influence the catalytic performance.[15]
In the present work, iron oxide has been selected as a
support for gold nanoparticles. Iron oxide, apart from being
inexpensive, environmentally friendly and readily available, is
[
[
a]
b]
Dr. T. García, Dr. J.M. López. Instituto de Carboquímica (CSIC),
C/Miguel Luesma Castán 4, 50018 Zaragoza, Spain.
E-mail: tomas@icb.csic.es
Dr. B. Solsona, Dr. R. Sanchis. Departament d´Enginyeria Química,
Universitat de València, C/ Dr. Moliner 50, 46100 Burjassot,
Valencia, Spain.
conveniently synthesised, and it is also reasonably active for the
total oxidation of VOCs including propane.[
16]-[19]
For bulk iron
oxide catalysts, it has been observed that high surface area and
reducibility of the iron species are the key parameters needed to
achieve high catalytic activity. This is in accordance with the redox
Mars-van Krevelen mechanism that controls the oxidation of
propane on iron oxide. For supported iron oxide catalysts, the
extent of iron oxide dispersion determines the catalytic activity. In
particular, more highly dispersed iron oxide species have been
E-mail: benjamin.solsona@uv.es
[
[
c]
d]
Dr. D.J. Willock, Dr. T.E. Davies, Dr. Q. He, Prof. C.J. Kiely, Prof.
S.H. Taylor. Cardiff Catalysis Institute, School of Chemistry, Cardiff
University, Main Building, Park Place, Cardiff, CF10 3AT, UK.
E-mail: taylorsh@cardiff.ac.uk
L. Lu, Prof. C.J. Kiely. Department of Materials Science and
Engineering, Lehigh University, 5 East Packer Avenue, Bethlehem,
PA 18015-3195, USA.
reported to be more reducible than highly aggregated species,
and consequently this leads to higher catalytic activity.[20]
Supporting information for this article is given via a link at the end of
the document.
For bulk oxide catalysts, the nano-architecture of the iron
oxides does not seem to be of paramount importance for total
This article is protected by copyright. All rights reserved.

ChemCatChem
10.1002/cctc.201900210
FULL PAPER
oxidation of VOCs. Ordered mesoporous iron oxides can be
highly active and stable, but do not seem to offer additional
slightly higher than nanocrystalline Fe
either the soft templating route (Fe
iron(II) nitrate precursor (Fe -C).
2 3
O supports, prepared by
-SC) or by calcination of the
2
O
3
advantages when compared to other high surface area FeO
materials.[
x
2 3
O
18], [19]
Gold supported on nanostructured iron oxide has previously
been prepared and tested as a catalyst for CO oxidation, and they
A
show higher activities than other Au/FeO
x
catalysts.[21] This
improved performance, in addition to factors such as small
particle size and high surface area, is also related to the presence
of a large amount of hydroxylated iron species. In other work, it
was reported that the catalytic activity for CO oxidation can be
2 3
enhanced using nanostructured α-Fe O with relatively small
pores that are of suitable dimensions to accommodate gold
nanoparticles inside.[22] The importance of accommodating
particles into the pores has also been observed with other
supported noble metal catalysts. For example, Pt deposited into
2
microscale mesoporous CeO presented higher activity for the
oxidation of benzene, than when Pt was deposited onto ceria
nanocubes.[23] This enhancement was proposed to be related to
the formation of active interface sites between and the support.
Following this line of reasoning, the present work investigates
B
2 3
catalysts consisting of gold deposited on mesoporous Fe O
prepared by a nanocasting route for the total oxidation of propane,
a representative VOC. These catalysts have been synthesized so
that gold particles are partially confined within the mesopores of
the nanostructured iron oxide. Our study probes how these
restrained gold particles affect the surface of the iron oxide and
the gold/support interface and consequently the catalytic
performance. For comparative purposes, gold supported on both
a non-mesoporous iron oxide and a mesoporous iron oxide
prepared with oxalic acid as a swelling agent were also
investigated.
Results and Discussion
Catalytic activity.
C
2 3
Gold-based Fe O catalysts were tested for the total
oxidation of propane. Conversion as a function of temperature is
shown in Figure 1A for the catalysts based on the iron oxide
prepared by
a hard template nanocasting method. For
comparison gold-free and gold containing iron oxides, in which
the iron oxide has been prepared by other methods, have also
been tested (Figures 1B, 1C and Table 1). For all catalysts the
2
main reaction product was CO . In some cases, low selectivity to
propylene was noted (but only at very low conversion) and
propylene selectivity decreased rapidly as conversion increased.
The propylene yield never exceeded 1%. Carbon monoxide
was not detected, but due to the detection limits of the GC thermal
conductivity detector a very low selectivity to this product cannot
be ruled out. Table summarizes the catalytic reaction
1
temperatures of T , T50 and T90 (corresponding to the
temperature for 10, 50 and 90% propane conversion respectively).
The total propane oxidation activity of the supports alone
10
Figure 1. Propane conversion as a function of temperature over gold-free and
gold-containing iron oxide catalysts. The supports have been prepared using A)
hard template nanocasting method, B) soft chemistry method and C) simple
calcination procedure. Reaction conditions: 8,000 vppm propane in air, GHSV
has also been studied. In agreement with previously published
data,[
18]
the mesoporous Fe
2
O
3
support prepared by the
-
1
=
100,000 h .
2 3
nanocasting route (Fe O -HT) showed a light-off temperature
This article is protected by copyright. All rights reserved.

ChemCatChem
10.1002/cctc.201900210
FULL PAPER
Interestingly, it can be observed from Figure 1 that the
efficacy of the catalytic combustion of propane over the
corresponding gold-based Fe catalysts greatly depended on
the support characteristics. Whilst both the 3AuFe -HT and
AuFe -HT catalysts showed an enhanced activity compared
with the parent Fe -HT nanocast support, a similar positive
effect was not observed for either the gold-based Fe -C
catalyst or the gold-based Fe -SC catalyst. The mass
normalised order of activity followed the sequence:
AuFe -HT > 3AuFe -HT > Fe -SC > Fe
-C > 5AuFe -SC > Fe -HT.
2
O
3
2
O
3
6
2 3
O
2
O
3
2
O
3
2
O
3
6
2
O
3
2
O
3
2
O
3
2 3
O -C ≈
3
AuFe
2
O
3
2
O
3
2 3
O
Table 1. Summary of catalytic activity for the various Au/Fe
2 3
O -HT
catalysts expressed as the temperature for 10, 50 and 90% propane
conversion to CO
propane oxidation
2
(T10, T50 and T90), and mass normalised rates of
Propane oxidation[
a]
T10 / ºC
T50 / ºC
T90 / ºC
Catalytic
activity[b]
Fe
AuFe
Fe
AuFe
Fe
2
O
3
-C
-C
-SC
-SC
-HT
300
305
295
300
317
279
268
350
355
350
355
370
339
334
385
400
385
385
418
387
370
71
90
3
2
O
3
2
O
3
107
81
5
2
O
3
2
O
3
34
3
6
AuFe
2
O
3
-HT
-HT
146
198
AuFe
2
O
3
[
a] Reaction conditions: 8000 vppm propane in air, GHSV = 100,000
-
1
h . [b] Catalyst activity determined at 300 ºC and expressed in gpropane
kgcat h .
-
1
-1
Notably, the T10 and T50 values decreased from 317 ºC to
68 ºC and from 370 ºC to 334 ºC, respectively, after deposition
Figure 2. Propane conversion for the 6AuFe
2
O
3
-HT catalyst as a function of the
2
A) reaction temperature during cyclic operation; B) time on line. Reaction
-
1
of 6 wt% gold onto the Fe
nanocasting route. The stability of the most active catalyst for the
propane total oxidation, 6AuFe -HT, was also assessed. It can
2
O
3
-HT support prepared by a
conditions: 8,000 vppm propane in air, GHSV = 100,000 h , temperature = 350
ºC.
2
O
3
During the nanocasting process, the KIT-6 mesochannels
and their connecting microchannels are filled and then the silica
wall material between them is etched away, generating pores of
be observed that this catalyst was not only showing an
outstanding stability with the time on line experiment but also
during cyclic operation (see Figure 2).
approximately
4 nm in dimension. However, varying the
hydrothermal conditions during the KIT-6 synthesis can lead to a
lower proportion of KIT-6 microchannels. Consequently, the
simultaneous occupation of the KIT-6 microchannels is not fully
achieved and a broad mesopore size distribution centred at 18
nm is obtained,[25] which is consistent with observations in the
Catalyst characterisation.
2 3
The gold loadings for different gold-based Fe O catalysts
were measured by XEDS analysis and experimental values are
reported in Table 2. It was observed that measured gold loadings
were in good agreement with the expected nominal values,
implying that most of the gold precursor was successfully
deposited on the catalyst support. Nitrogen adsorption isotherms
are presented in Figure 3A for the catalysts prepared on the hard
and soft templated iron oxide supports. The isotherms confirmed
the mesoporous character of the solid support prepared by the
2 3
current study. Additionally, it is worth noting that for the Fe O -HT
support there is also some evidence of macropore formation,
arising from larger interstitial pores created between the hard-
template particles. The presence of this macroporosity could be
beneficial, as it can potentially improve the diffusion of the gold
nanoparticle precursor into the inner porosity during the
deposition-precipitation preparation process. After loading gold
nanocasting route (denoted as Fe
2
O
3
-HT). The nanostructured
2
support showed a markedly high specific surface area of 185 m
2 3
nanoparticles, at 3 wt. % and 6 wt. %, onto the Fe O -HT support,
-1
3
-1
the BET surface area decreased slightly to 173 m g and 177 m2
2
-1
g , with a total pore volume of 0.49 cm g . The BJH method
applied to the adsorption branch of the isotherm (see Figure 3C),
showed a broad mesopore size distribution centred at 18 nm. KIT-
-1
g , respectively, whilst total pore volume decreased more
significantly from 0.49 cm3 g-1 to 0.38 cm g and 0.37 cm g
respectively. These observations can be linked to the partial
blocking of the large mesopores on the Fe support by Au
3
-1
3
-1
6
consists of two interpenetrating mesoporous channels linked by
microporous channels.[24], [25]
O
3
2
This article is protected by copyright. All rights reserved.

ChemCatChem
10.1002/cctc.201900210
FULL PAPER
nanoparticles. Accordingly, BJH pore size distributions of the
2 3
gold-based Fe O -HT catalysts showed that the maxima of the
distribution decreased from 18 to 12 nm after Au deposition
(
Figure 3C).
Table 2. Textural parameters of the KIT-6 silica template and hard templated
Fe O supports and the corresponding gold containing catalysts.
2 3
Au][a]
S
[b]
V
[c]
d
[d]
0
[
BET
2
T
-1
3
-1
/
wt %
/ m ·g
/ cm ·g
/ nm
KIT-6
Fe -C
AuFe -C
Fe -SC
AuFe -SC
Fe -HT
-
-
809
24
0.94
0.05
0.04
0.42
0.44
0.49
0.38
0.37
5.5
47
46
21
28
12
8.4
9.1
2
O
3
3
2
O
3
2.9
-
28
2
O
3
82
5
2
O
3
4.1
-
60
2
O
3
185
173
177
3
6
AuFe
2
O
3
-HT
-HT
2.6
4.9
AuFe
2
O
3
[
a] Percentage of gold determined by XEDS. [b] Specific surface area
deduced from the isotherm analysis in the relative pressure range of 0.05-
.25. [c] Total pore volume at relative pressures 0.95. [d] average pore
0
diameter calculated from the adsorption branch of the isotherm using the
BJH method.
Additionally, the BJH distribution revealed that there was a
loss of the macroporosity originally available for the bare Fe O -
2 3
HT support. In agreement with this, TEM analysis (presented later
in Figure 5), shows that ordered interstitial gaps that make up
much of the macroporosity were mostly missing from the gold-
based Fe
result, the preparation method creates gold particles that are
partially confined by the internal mesoporosity of the Fe -HT
2 3
O -HT catalysts as compared to the bare support. As a
2
O
3
support, which, in turn, may increase the metal-support contact
area compared to gold nanoparticles supported on the external
surface of the more conventional metal oxide support. This
hypothesis was further confirmed by non-local density functional
theory (NL-DFT) analysis of the adsorption isotherm branch
(
Figure 3B). Interestingly, NL-DFT pore size distributions clearly
showed the existence of some microporosity in the Fe -HT
2
O
3
sample, with a narrow peak centered at 1.4 nm. Furthermore, this
peak, which had almost disappeared after 3 wt. % gold was added,
was completely absent after 6 wt. % gold addition. Therefore, it is
2 3
possible that micropores intersecting the Fe O -HT surface could
act as anchoring points for small gold nanoparticles. It should also
be pointed out that the NL-DFT model could accurately reproduce
Figure 3. (A). N
2
adsorption isotherm data, (B). NL-DFT pore size distribution
the N
proposed model. The mesoporous support formed by aggregation
of nanoparticles (Fe -SC) showed lower specific surface area
2
adsorption isotherm data, which strongly supports the
and (C) BJH pore size distribution for the various gold-free and gold-containing
Fe O
2 3
catalysts prepared by hard and soft templating methods.
2
O
3
2
-
than the nanostructured support (Fe
2
O
3
-HT), at 82 and 185 m g
Accordingly, the pore size distribution of this sample is
1
respectively, but only a slightly lower total pore volume, at 0.42
and 0.49 cm3 g-1 respectively (Table 2). Comparing both
isotherms (Figure 3A) it can be observed that the increase in
slope at low partial pressures (ca. 0.4), typical of mesoporous
materials with intra-particle pore systems, is more noticeable for
shifted to higher values and centered at about 30 nm (Figure 3C),
with an average pore dimension of 21 nm compared with the 12
nm presented in the 3D network nanostructured support. After
2 3
loading gold nanoparticles at 5 wt.% onto the Fe O -SC support,
2
-1
the BET surface area decreased significantly to 60 m g , whilst
the total pore volume remained at a comparable value of 0.44
2 3
the nanostructured support, whilst the Fe O -SC sample presents
a more visible increase of the slope at higher relative pressures,
indicating a more substantial contribution from inter-particle
porosity.
3
-1
cm g . This behavior can be linked to the deposition of gold
nanoparticles onto the mesoporous external surface of the Fe
SC nanoparticles, as shown in Figure 3. In line with this, it should
2
3
O -
This article is protected by copyright. All rights reserved.

ChemCatChem
10.1002/cctc.201900210
FULL PAPER
be noted that the NL-DFT model did not show any features related
well-ordered structure, in accordance with it being an inverse
replica of the KIT-6 template, composed of uniform nanoparticles
linked by nanocrystalline bridges. The average primary particle
size in the Fe O -HT sample (determined from measurements on
2 3
more than 200 particles) was found to be 7 ± 1 nm.
to the presence of microporosity in the Fe
observed for the nanostructured Fe
contrast, N adsorption analysis of the catalyst derived from the
iron(II) nitrate salt revealed a low porosity with a specific surface
2
O
3
-SC sample, as was
2
3
O -HT support. In direct
2
2
-1
2
-1
area of 22 m g (Table 2). A comparable value of 28 m g was
obtained after 3 wt% gold deposition, indicating that gold
nanoparticles have been deposited exclusively on the external
surface of the support.
Previously published selected area electron diffraction data
[
19]
confirmed the nanocast support and Fe
2 3
O -C materials to have
a Fe structure (JCPDS: 85-0599) with space group R-3c. After
2
O
3
the incorporation of gold into the mesoporous iron oxide through
deposition precipitation, the ordered structure derived from the
template appears to be lost, as shown in Figure 5(b).
Nevertheless, the iron oxide primary particle size remains small
(~7 nm), as compared to the non-porous iron oxide sample
(Figure 5(c)), which has a particle size distribution ranging from
Powder X-ray diffraction (XRD) was used to identify the
crystalline phases present in the different materials. Figure 4
shows the XRD patterns of the iron oxide obtained by the hard-
templating route (Fe
supported Au catalysts (3AuFe
samples only showed two tiny broad peaks at 2θ = 33.2 and 35.7 , about 20 to 50 nm.
2
O
3
-HT) and the corresponding iron-oxide-
-HT and 6AuFe -HT). These
2
O
3
2 3
O
º
º
which could be tentatively indexed to 104 and 110 reflections from
2 3
O
.[26] On the other hand, both nanocrystalline iron oxide
α-Fe
supports (Fe
supported on them (5AuFe
diffraction peaks corresponding to reflections indexed to
rhombohedral α-Fe
.[26] A smaller mean iron oxide crystallite
size was obtained for the Fe -SC support compared with the
Fe -C counterpart, (i.e., 12 nm and 24 nm respectively).
a
b
c
2
O
3
-SC and Fe
2
O
3
-C), and those catalysts with Au
2
O
3
-SC and 3AuFe -C) showed nine
2 3
O
2
O
3
2
O
3
2
O
3
5
0 nm
50 nm
50 nm
Furthermore, the deposition of gold nanoparticles onto these
supports did not significantly change the mean crystallite sizes.
Moreover, none of the gold-loaded catalysts, irrespective of the
support, exhibited peaks corresponding to reflections from gold,
which indicates the presence of highly dispersed gold
nanoparticles. Low angle XRD patterns (Figure S1) showed a
Figure 5. Representative bright field (BF) TEM images of (a) the Fe
support material, (b) the 6AuFe -HT catalyst and (c) the Fe -C support
material. Remnants of the ordered template network can be clearly seen in
Fe -HT material, but this has essentially collapsed after depositing Au, as
3
shown in (b). The primary particle size of the iron oxide support in 6AuFe O -
2 3
O -HT
O
2 3
2 3
O
2 3
O
2
2 3
mesoporous structure for the Fe O -HT sample with a decrease
HT is significantly smaller than that of the conventional iron oxide support shown
in (c).
in the intensity after gold deposition and with no evidence of
crystalline order. These result matches with the TEM
observations presented below.
The gold nanoparticles can be distinguished from the iron
oxide support using atomic number (Z) contrast in a scanning
transmission electron microscope (STEM) using the high angle
annular dark field (HAADF) imaging mode. As shown in Figures
104
110
012
113
024 116
214 300
018
5
AuFe O -SC
2
3
6(a-c), gold nanoparticles from 2 to 6 nm in size can be found in
the 6AuFe -HT sample. Occasionally larger particles of 20 nm
2
O
3
can be also found (Figure S2). A comparable particle size
distribution of Au particles, ranging mainly from 2 to 8 nm, and
occasionally larger particles above 30-40 nm could be found in
Fe O -SC
2
3
3
AuFe O -C
2 3
2 3
the 5Au Fe O -SC catalyst (Figure 7). Therefore, gold particle size
does not seem to exert a key role in the better catalytic
performance observed for gold catalyst supported on nanocasted
iron oxide compared to the other iron oxide supports.
Fe O -C
2
3
6
AuFe O -HT
2 3
3
AuFe O -HT
2
3
Interestingly, the iron oxide support material in the
Fe O -HT
2
3
6
2 3
AuFe O -HT catalyst appears to have significant silicon oxide
2
0
30
40
50
60
70
content at its surface. As shown in Figures 6(d-e), X-ray energy
dispersive spectra (XEDS) acquired while the electron beam was
scanning the area 1 (the near surface region) and area 2 (the
bulk) of the iron oxide grain respectively. After normalising the two
spectra using the total intensities of Fe K and Cu K peaks, a
significant difference in the Si K peak intensities can be
established. This suggests that the significant additional Si K
signal arising from the near surface region (area 1) as compared
to area 2 is not just from the internal fluorescence in the XEDS
silicon drift detector, which is a common spectral artefact.
2
q (º)
Figure 4. Wide angle XRD patterns of the catalysts obtained by hard templating
(
Fe
2
O
3
-HT, 3AuFe
-SC) and those derived by precipitation from a nitrate salt precursor
-C and 3AuFe -C).
2 3 2 3 2 3
O -HT and 6AuFe O -HT), soft templating (Fe O -SC and
5
AuFe
2
O
3
(Fe
2
O
3
2 3
O
Figure 5 shows representative bright field (BF) TEM
micrographs of the Fe -HT and Fe -C samples. From Figure
(a), it can be observed that the Fe -HT support presents a
2
O
3
2 3
O
O
2 3
5
This article is protected by copyright. All rights reserved.

ChemCatChem
10.1002/cctc.201900210
FULL PAPER
a
b
e
c
d
f
g
h
i
2 3
Figure 6. Representative STEM characterization results from the 6AuFe O -HT
catalyst. (a-c) High angle annular dark field (HAADF) Z-contrast micrographs of
the catalysts showing Au nanoparticles ranging from 2-6 nm in size (indicated
by white arrows). (d) Representative STEM bright field (BF) image of the iron
2 3
oxide support material in the 6AuFe O -HT catalyst and (e) the corresponding
X-ray energy dispersive spectra (XEDS) acquired from the two regions indicated.
The XEDS two spectra in (e) were normalised based on the intensities of the Fe
K peaks and Cu K peaks (the latter of which originates from the TEM grid). (f)
STEM-BF image and the corresponding XEDS maps of (g) Si and (h) Fe. (i)
shows an overlay of the Si (green) and Fe (red) map.
This observation provides clear evidence of a genuine ultra-
thin Si containing layer (most likely silicon dioxide) on the iron
oxide support, which may originate from an incomplete etching
away of the KIT-6 silica template. Qualitative XEDS mapping
shown in Figure 6(f-i) suggested that the surface is likely to be a
mixed phase of Si and Fe oxides.
2 3
Figure 7. Representative STEM-HAADF micrographs from the 5AuFe O -SC
catalyst. (a) Micrograph showing the typical size and morphology of the iron
oxide support grains (b) and (c) small Au nanoparticles about 2 nm in size
(indicated by white arrows) as well as relatively large Au particles (black arrow)
are clearly evident in the 5AuFe -SC catalyst.
2
O
3
The gold oxidation states for the 3AuFe
2 3 2 3
O -HT, 6AuFe O -
Figure 8. X-ray photoelectron spectra of various iron oxide supports and the
HT and 3AuFe -C catalysts were characterized using XPS by
2
O
3
corresponding gold loaded Fe
D) Si 2p spectra.
2 3
O catalysts. (A) Au 4f; (B) Fe 2p; (C) O 1s and
analysing the Au 4f spectra (Figure 8a). All the samples examined
displayed peaks at the binding energies typical of metallic gold.[27]
(
This article is protected by copyright. All rights reserved.

ChemCatChem
10.1002/cctc.201900210
FULL PAPER
The oxidation state of the iron species was also
ºC and 298 ºC, respectively. It is also highly plausible that the low
reducibility of the nanostructured iron oxide might also be
associated with the presence of a mixed phase of Si and Fe
oxides on the nanostructured support, as observed by XEDS and
XPS analysis.
characterized by XPS. The deconvolution of the Fe2p region was
more challenging due to the presence of four peaks,
corresponding to the Fe2p3/2 multiplet (Figure 8b). All the
materials examined presented the four peaks at binding energies
that matched perfectly with those expected for the Fe3+ multiplet
(
i.e., 709.8, 710.7, 711.4 and 712.3 eV), with no peaks at binding
2
+
[28]
energies corresponding to Fe species. The results of these
more surface sensitive analyses are in good agreement with bulk
2 3
phase XRD analyses that revealed α -Fe O as the unique
crystalline phase. Somewhat surprisingly, the O 1s spectra clearly
showed different features when comparing the mesoporous and
nanocrystalline catalysts (i.e., the Fe
2
O
3
2 3
-SC and Fe O -C series)
with the nanostructured ones (i.e., the Fe
2
3
O -HT series). Upon
deconvolution of the O 1s spectra, two surface oxygen species
were detected for all the nanocrystalline samples (Figure 8c). The
binding energy of ca. 529.3 eV, denoted as Fe-O-Fe, is
2
-
characteristic of O lattice oxygen.
However, determining the origin of the species having a
binding energy of ca. 531.2 eV, denoted as Fe-OH, is more
difficult, as this feature could either be the result of hydroxyl
groups or alternatively might be due to the presence of oxygen
vacancies, surface adsorbed oxygen, or carbonate species.[29] In
contrast, the nanostructured catalysts presented the two peaks at
ca. 529.3 and 531.2 eV with lower intensity, but also exhibited two
additional peaks; namely a high-intensity peak centred at ca.
530.0 eV and a low-intensity feature centred at ca. 532.3 eV.
These new features also correlated with the fact that these
samples showed a peak at 101.7 eV in the Si 2p spectral area
(
see Figure 8d). Combining both pieces of XPS evidence and the
results found by STEM-XEDS, the presence of a mixed phase of
Si and Fe oxides can be confirmed in these samples, which arises
due to the incomplete removal of the silica template during the
preparation of the Fe
The reducibility of the bare Fe
corresponding gold-loaded catalysts was studied by means of H
2 3
O -HT support.
2
3
O supports and the
2
-
Figure 9. Hydrogen temperature programmed reduction (TPR) profiles of the
temperature programmed reduction (Figure 9). Similarly shaped
profiles were attained for all samples, although reduction occurred
at a range of different temperatures (Table 3). Comparable total
hydrogen consumption values were measured, which were close
2 3
various iron oxide supports and the corresponding gold loaded Fe O catalysts.
After depositing gold nanoparticles onto the nanostructured
support, 3AuFe -HT and 6AuFe -HT catalysts showed a
2
O
3
2 3
O
-1
to the theoretically expected value (ca. 19 mmol g ) for the
reduction of Fe to Fe. A first reduction feature of medium
significant shift in the low-temperature reduction maxima, which
were observed to decrease to 353 ºC and 342 ºC, respectively.
Additionally, the high-temperature maximum decreased from >
600 ºC to 580 ºC in the case of the catalyst with the highest gold
loading. These results suggest that the presence of gold metallic
nanoparticles, partially confined in the internal support
2
O
3
intensity was observed with the maximum at 250-350 ºC, and a
second intense broad feature at 450-650 ºC. These profiles were
related to the following respective transitions Fe
2 3 3 4
O → Fe O and
Fe
3
O
4
→ FeO → Fe, according to literature data[30], [31].
No characteristic reduction features related to gold species were
observed. This result is fully consistent with the observation of
gold in the metallic state from XPS characterisation. Considering
the reducibility of the iron oxide supports alone, it has been
proposed that the formation of nanocrystalline bridges between
the iron oxide nanoparticles, such as those formed in materials
prepared by nanocasting, have an influence on the reducibility of
2 3
mesoporosity, notably promoted the reduction of Fe O , despite
the presence of both the nanocrystalline bridges and the mixed
phase of Si and Fe oxides. In line with this, Mao et al. [ have
recently postulated that the partial confinement of Pt
23]
nanoparticles within the mesopores of microscale CeO
a significant enhancement in the activity of the surface lattice
oxygen of CeO at the interface between Pt nanoparticles and the
CeO support. In the present study, we also observe a
2
, leads to
2
the iron species.[19] Thus, the Fe
O
3
-HT support showed a TPR
2
2
profile which was clearly shifted to a higher temperature, with a
low-temperature peak maximum at 447 ºC, compared to the
Fe O -SC and Fe O -C supports, which displayed peaks at 319
2 3 2 3
modification of the lattice oxygen species from the interaction
between confined metallic Au nanoparticles and the support and
it could be assumed that these lattice oxygen species could be
This article is protected by copyright. All rights reserved.

ChemCatChem
10.1002/cctc.201900210
FULL PAPER
responsible for increasing the catalytic activity for propane total
oxidation.
with a p(3×3) supercell of the (0001) surface, which we denote as
Au10/α-Fe O (0001). The Au10 particle had an initial geometry
2 3
With respect to gold deposition on the more conventional
Fe O -SC and Fe O -C supports, TPR profiles showed that the
2 3 2 3
taken from the f.c.c. bulk lattice structure with a Au(7,3) two-layer
structure. HRTEM has been used to show that Au10 particles
lowest temperature reduction peak was shifted to a lower
temperature, i.e., 304 °C and 273 °C respectively, whereas the
highest temperature peak moved to marginally higher
temperatures. Hence, it is evident that supporting gold
nanoparticles on just the external surface of the iron oxide
structures did not significantly modify the reduction
characteristics of the supports, since both iron oxides
displayed comparable hydrogen reduction profiles with and
without the presence of gold, which is in agreement with data
from other studies.[32] Therefore, it can be determined that gold
present in active Au/FeO
bottom two slab layers (Fe–O
x
catalysts are bilayer in nature.[18] The
–Fe) were fixed during optimization
3
to represent the bulk lattice beneath the surface. After
optimisation of the clean stoichiometric slab and Au10 cluster, the
cluster retained its bi-layer character although movement of the
Au atoms did occur.
To represent the less well-ordered surfaces expected for α-
Fe
Au10/α-Fe
2
O
3
obtained from our nanocasting synthesis, the optimised
(0001) structure was further modified by the
2
O
3
introduction of surface grooves. These grooves were produced by
the removal of stoichiometric sets of atoms from the iron oxide
surface near to the Au10 cluster as shown in Figure 10.
2 3
deposited on the external surface of nanocrystalline α-Fe O
results in a weaker metal-support interaction as the redox
properties of these latter iron oxide supports were not strongly
modified by gold addition.
a)
b)
Table 3. Summary of temperature programmed reduction (TPR) and X-ray
photoelectron spectroscopy (XPS) results for the various iron oxide supports
and corresponding Au/Fe
2 3
O catalysts. The Si 2p peak corresponding to a
SiO pure silica film occurs at a binding energy of 102.6 eV.
2
3
H
2
-TPR
Fe
Si 2p
O 1s
Au
4f
Si/Fe
2p
T
max
BE
(eV)
BE
(eV)
BE
(eV)
BE
(eV)
-
1
2
(ºC)
Fe O
2 3
-HT
447,
710.9
101.6
529.9
1
1
2 3
Figure 10. a) Slab model of the Au10 cluster on a roughened surface of α-Fe O ,
>
600
viewed down the [110] direction. Atoms that are fixed during optimisation are
shown in line drawing mode, relaxed Fe and O atoms as ball and stick models
and Au atoms in 0.7 CPK representations. b) Plan view ([0001] direction) with
the positions used for oxygen defect creation indicated. Line representation is
used for all lower layer atoms for clarity. Atom colours: Fe - blue, O – red and
Au – yellow.
3
6
AuFe
2
O
3
-HT
353,
710.8
101.8
529.9
83.5
>
600
AuFe
2
O
3
-HT
342, 580
319,
710.9
710.5
101.5
-
530.0
529.5
83.8
-
1
-
2 3
Fe O -SC
>
600
5
AuFe
2
3
O -SC
304,
710.4
-
529.5
83.0
-
>
600
We have investigated different positions in the surface layer
Fe
2
O
3
-C
-C
298, 504
273, 545
710.2
710.4
-
-
529.3
529.4
-
-
-
2 3
of the α-Fe O (0001) support for the creation of surface oxygen
3
AuFe
2
O
3
83.3
vacancies. The aim is to examine the effect of the presence of the
metal nanoparticle on the removal of nearby surface oxygen. For
both the Au /α-Fe O (0001) and oxidised cluster Au O /α-
The differences observed confirm the crucial role of internal
1
0
2
3
10
6
porosity in nanostructured supports [33], which facilitates a more
intimate contact between partially confined metal nanoparticles
and the support.
Fe O (0001), calculations were carried out for oxygen positions
2
3
under the nanoparticle, at the interface between the particle and
oxide (nearest), and at oxygen sites next nearest to the Au10
nanoparticle with the whole system optimised in each case. For
the grooved surface, we removed 2 co-ordinate surface oxygen
anions to create the oxygen defects, the sites used are shown in
Figure 10b. In all cases, the single oxygen vacancy formation
energy was calculated as follows:
Finally, although gold supported on nanocasted iron oxide
catalysts has shown an outstanding stability under cyclic
operation, see Figure 2A, the characterization of the used
6
AuFe
was carried out. As expected, no significant differences were
appreciated, either in its structural characteristics (N adsorption
2 3
O -HT sample by different complementary techniques
2
and XRD) or in its chemical surface (XPS). These results are
included as supporting information (Figures S3 to S5).
Evac = E(Au /Fe O ) - E(Au /Fe O108) – 1/2E(O2)
(1)
10
72 107
10
71
where (Au10/Fe72
107
O ) denotes the oxygen defect system,
DFT studies: The role of support surface structure on the
availability of lattice oxygen at the metal-support interface.
(Au /Fe O ) is the stoichiometric system, and (O ) is the gas
1
0
71 108
2
phase oxygen in the triplet ground state. For the oxidised cluster,
should replace Au10 in equation (1). The vacancy creation
Au10
O
6
The Au supported on iron oxide catalysts were modelled by
placing a Au10 nanoparticle onto a 4-layered slab (Fe–O –Fe) built
3
energy defined in equation (1) gives a positive value for an
endothermic process. Table 4 summarises the calculated defect
This article is protected by copyright. All rights reserved.

ChemCatChem
10.1002/cctc.201900210
FULL PAPER
energies for the flat α-Fe
2
O
3
(0001) and grooved grv-α-Fe
2
O
3
for how a silicon containing layer would interface with the
surfaces with and without the inclusion of a Au10 cluster. For the
α-Fe (0001) surface all oxygen atoms are three co-ordinate
predominant Fe
examine the effect of this interface we took a single FeO
octahedral centre from the optimised α-Fe structure and
overlayed it on the olivine (100) surface. The octahedral FeO
structure was overlayed so that 3 oxygens on the surface
matched closely with the O atoms of the FeO fragment and then
the remaining three O atoms where replaced with hydroxyl groups
(Figure S6a). The three O atoms in the octahedral fragment could
be matched within 0.25 Å of the surface O atom positions. These
three fragment O atoms were then deleted to provide a model of
the interface of the silicon-rich olivine surface and an Fe(3+)
centre. The slab model was then re-optimised and the resulting
structure can be seen in Figure S6b). On optimisation the Fe(3+)
2 3
O material in the nanocast catalysts. To
2
O
3
6
and we find a defect energy of 3.02 eV which is lowered to 2.13
eV when the defect is created near to a surface Au cluster. This
is in agreement with our earlier work,[34] where we showed that
this is partly due to the polarisable nature of the Au10 cluster which
stabilises the change in oxidation state of surface Fe cations on
the removal of an oxygen anion and partly due to charge transfer
to the Au10 cluster itself. A similar effect has also been found in
calculations for Au supported on other oxide surfaces, for
2 3
O
6
6
[
35]
example Au/TiO
2
.
For the grooved surface, several different
types of surface O are present. We have concentrated on two co-
ordinate oxygen anions as these give rise to low defect creation
energies. Introducing surface roughness into the model by adding
these groove features has a very strong influence on the defect
creation energy, with values between 1.00 and 1.82 eV, being
calculated for the removal of two co-ordinate surface O anions at
the positions indicated in Figure 10b. This is up to 1.13 eV lower
has moved away from two of the surface O ions and the FeO
6
structure has rotated to bring two of the OH groups into bridging
positions with surface Fe(2+) ions. This results in a loss of the
octahedral symmetry that the Fe(3+) ion would have in α-Fe
and suggests that the interface between an olivine like iron silicate
and α-Fe would be quite strained, which would be expected to
2 3
O
than the Au10 decorated α-Fe
2
O
3
(0001) indicating that the change
2 3
O
in the surface structure of the oxide which could be brought about
by choice of synthesis method will have an important influence on
the availability of lattice oxygen. Interestingly, the addition of a
change the availability of lattice oxygen.
2 3
Comments on the activity of gold-loaded Fe O catalysts.
Au10 cluster to the surface to give the Au/grv-α-Fe
has a relatively minor effect on the defect creation energy, even
for defects quite close to the Au10 cluster (this can be seen by
2
O
3
models now
The explanation for differences in the catalytic activity for propane
total oxidation of the various materials studied here is based on
several parameters. For the bare iron oxide support without gold
nanoparticles, the surface area was not a sufficient descriptor to
explain the total oxidation of propane, as previously reported.[
2 3 2 3
comparing grv-α-Fe O -1 and Au/grv-α-Fe O -1 in Table 4).
19]
As discussed earlier, our STEM-XEDS and XPS
measurements suggest that there is also a significant level of Si
in the surface layer of the nanocast materials. One possibility is
that the silica used in the casting process forms a mixed oxide
Indeed, both the Fe
catalytic activity for propane oxidation than the nanostructured
Fe -HT support, which presents a markedly higher surface area.
As observed by H -TPR, the nanostructured Fe -HT support
showed a lower reducibility than both nanocrystalline Fe -C
and mesoporous Fe -SC supports. This behaviour might be
2 3 2 3
O -SC and Fe O -C materials show higher
2
O
3
phase with Fe
also considered the olivine structure, Fe
unit cell was optimised using a similar DFT approach to that
described for α-Fe . A cell expansion using a cell with
stoichiometry Fe Si 16 showed an energy minimum just 1.3 %
higher in volume than the experimental reference (a =10.607, b =
2
O
3
. As a model of such a mixed phase we have
2
2 3
O
.[36], [37] The olivine
2
SiO
4
2 3
O
2
O
3
2
O
3
related to the presence of crystalline bridges between adjacent
iron oxide nanoparticles, which themselves were partially covered
by a mixed phase of Si and Fe oxides, stabilizing the catalyst
surface and leading to the formation of iron species that were
more difficult to reduce. The specific nature of the active sites
responsible for the total oxidation of propane in metal oxides is
not completely understood; however, the catalytic activity
displayed during the deep oxidation of light alkanes, such as
propane, is closely related to the ease of reduction and re-
oxidation of the active sites in the catalyst. Hence, many studies
have demonstrated that alkane oxidation on metal oxides takes
8
4
O
6.164, c = 4.870 cf a =10.460, b = 6.082, c = 4.815 (CSD:
9007046)). The most stable surface facet is indexed (100) in this
-2
setting for which we obtain a surface energy of 0.86 J m , which
in reasonable agreement with earlier DFT studies of this
material.[38]
2 3 2 3
Table 4. Calculated defect formation energies for Fe O and Au/Fe O DFT
models.
place via a Mars-van Krevelen mechanism involving lattice
oxygen through a redox cycle.[39]-[42] Accordingly, for the total
oxidation of propane we have previously established a clear
correlation between reducibility (quantified as the temperature of
the maximum of the first reduction feature) and catalytic activity
System
Defect
Energy (eV)
3.02
System
Defect
Energy (eV)
2.13
α-Fe
2
3
O (0001)
2 3
Au/α-Fe O (0001)
grv-α-Fe
grv-α-Fe
grv-α-Fe
O
2 3
O
2 3
O
2 3
– 1
– 2
– 3
1.00
Au/grv-α-Fe
2
O
3
- 1
– 2
– 3
1.21
1.50
Au/grv-α-Fe
2
O
3
1.21
(
2 3
normalized per unit surface area) for different Fe O catalysts.
1.82
Au/grv-α-Fe
2
O
3
1.57
This suggests that the rate limiting step for propane oxidation over
iron oxides is the reduction step and that the oxidation proceeds
Olivine only contains Fe in the 2+ oxidation state. This
means that it is unlikely that oxygen can easily be removed from
the lattice as the accompanying reduction of cations cannot take
place. However, the surface structure provides us with a model
utilising bulk lattice oxygen through a Mars-van Krevelen
mechanism, as we have also observed in our current study.
However, this same trend is not fully observed when iron oxide
catalysts loaded with gold nanoparticles are considered, as
This article is protected by copyright. All rights reserved.

ChemCatChem
10.1002/cctc.201900210
FULL PAPER
shown in Table 1. Whilst a higher specific activity is observed for
the gold-loaded catalysts on the nanostructured Fe O -HT
2 3
support as compared to the bare support, which is in accordance
with its improved reducibility, the opposite effect is true when gold-
nature of the support is also thought to have an important
influence on the number of gold-support perimeter sites. This
parameter could also be increased by the presence of Au
2 3
nanoparticles being in contact with more than one Fe O particle
nanoparticles are supported on Fe
2
O
3
-SC and Fe
2
O
3
-C supports,
simultaneously – i.e., creating two or more planar interfaces (and
thus more periphery line length) per Au particle. Superior metal-
support contact area could be attained for the mesoporous
nanostructured support compared to the gold nanoparticles
despite the fact that their reducibility was slightly increased.
Therefore, it can be postulated that the improved redox properties
of gold-loaded iron oxide nanostructures is not the sole parameter
controlling its catalytic performance for VOC removal.
The reaction mechanism for gold nanoparticle catalysts
deposited on reducible supports is more complex, as gold plays a
key role promoting both adsorption of VOCs and oxygen
activation. It is generally accepted that the rate of reaction is
supported on the external surface of nanocrystalline Fe
in the inter-particle mesoporosity of Fe -SC. Hence, the
enhanced activity achieved with the gold-based Fe -HT
2 3
O -C, or
2 3
O
2
O
3
catalyst could also be associated with an increased amount of
perimeter sites at the metal-support interface, controlled by
support nanostructure, where oxygen can be more easily
activated due to the presence of a roughened (microporosity),
2
dictated by the dissociation of O to yield atomic oxygen at the
oxygen vacancies near the metal–support interface. This step is
relatively slow in gold-based systems[43],[44] due to the high
activation barrier of ~2.2 eV.[43] However, for the gold sitting on a
porous iron oxide support, we have calculated that the presence
of surface roughness on the oxide support can significantly
decrease this value to ~1.2 eV when a Au10 cluster is located near
the surface corrugation. This theoretical scenario might be
increasingly important when gold nanoparticles are deposited on
nanostructured supports prepared by nanocasting. As shown
[46]
leading to an overall enhancement of activity.
2 3
Finally, it has been observed that the surface of the Fe O -
HT support is covered by a mixed phase of Si and Fe oxides,
where gold nanoparticles are deposited. This is not the situation
for the other iron oxide supports studied, as they do not contain
the silicon. As a comparable model of the mixed Si-Fe oxide
phase, DFT simulation studies have suggested that the interface
2 3
between an olivine -like iron silicate and α-Fe O is quite strained,
previously, micropores intersecting the Fe
2
O
3
-HT surface seem to
most likely improving the availability of surface lattice oxygen.
Therefore, the deposition of gold nanoparticles over this mixed Si-
Fe oxide phase, where oxygen can be more easily activated at
the gold-support interphase, might be another key parameter in
the better performance observed for the nanostructured catalysts.
This promoting effect is quite surprising and further studies are
needed to confirm the key role of this type of mixed surface phase
over surface oxygen activation, since silica-type materials are not
usually considered as an adequate matrix for gold-based
catalysts. Silica supports tend to enlarge gold particles deposited
over its surface due to the presence of weak Au–support
act as anchoring points for small gold nanoparticles at the inner
mesoporosity of the nanostructured support, which could facilitate
oxygen activation at the metal-support interface according to DFT
calculations. In addition, it is worth commenting that although the
Fe O -SC support is also a mesoporous support, there is neither
2 3
any intra-particle mesoporosity, nor surface porosity, which could
be equated to surface roughness. Therefore, gold is only
deposited on the inter-particle external surfaces in Fe O -SC-type
2 3
supports, limiting the extent to which oxygen activation can occur
at the metal-support interface. Indeed, gold deposition on the
Fe
since the specific activity of the 5Au Fe
decreased compared to the bare Fe -SC material, despite
improved redox properties. The same is shown to be true for the
gold catalyst supported on the nanocrystalline iron oxide denoted
2 3
O -SC support seems to be blocking the iron oxide active sites, interactions. Additionally, gold on silica cannot readily form
2
O
3
-SC catalyst was
activated oxygen species and consequently the C-H bond
cleavage of propane demands more severe conditions.
Summarising, it can be tentatively proposed that propane
2
O
3
total oxidation on gold nanoparticles confined into
a
Fe
2
O
3
-C.
As inferred from N
deposition of gold nanoparticles on the Fe
obtained by the hard-templating route, leads to
nanostructured Fe catalyst followed an Au-assisted Mars-van
2 3
O
2
adsorption and TEM measurements, the
-HT support,
partial
Krevelen mechanism, in which propane is adsorbed onto the
catalyst surface and reacts with active oxygen species created in
the vicinity of the metal-support interfacial region. The presence
2
O
3
a
confinement of small gold nanoparticles into the intra-particle
mesoporosity of the support. According to the work of Behm and
2 3
of a porous nanostructure in the Fe O -HT support has several
positive effects: i) a stronger metal-support interaction facilitating
oxygen activation at the interface due to the presence of confined
small gold nanoparticles, which are stabilised due to the
nanostructured nature of the support (microporosity and
mesoporosity); ii) a higher contact surface area between the
partially confined metal nanoparticles and the support; iii) a more
reducible support due to the formation of more active surface
lattice oxygen, and iv) a higher defect concentration and
consequently a better availability of lattice oxygen due to both the
surface microporosity of the support and the presence of a mixed
Si-Fe oxide phase.
co-workers, who studied Au/TiO
2
catalysts,[45] the quantity of
removable oxygen was essentially correlated with the number of
perimeter sites between gold nanoparticles and the support,
showing a linear relationship. Smaller gold nanoparticles have a
greater quantity of perimeter sites when normalised for gold
content, and thus gold particle size effects can lead to increased
redox activity. In spite of the absence of a relevant amount of large
2 3
gold particles in Au/Fe O -HT catalysts, which certainly appears
in the other gold-based iron oxide catalysts, gold particle size is
not believed to be the only effect responsible for the enhanced
activity, since both gold-based iron oxide catalyst shows
comparable gold particle size distribution. The nanostructured
This article is protected by copyright. All rights reserved.

ChemCatChem
10.1002/cctc.201900210
FULL PAPER
precipitation process, with target gold loadings of 3 and 6 wt %, giving
catalysts denoted as 3AuFe O -HT and 6AuFe O -HT respectively.
2 3 2 3
Table 5. Total oxidation of propane on catalysts containing iron
Catalysts
[C
3
H
8
]
GHSV
T10
T50
T90
/h-1
For comparative purposes, two iron oxides with different
characteristics were also employed as supports for gold. A mesoporous
support formed by aggregation of iron oxide nanocrystals was prepared
using aqueous iron nitrate with oxalic acid added as a swelling agent
/
ppm
[
[
19]
Fe
Fe
Fe
2
O
O
O
3
8000
8000
20000
20000
[b]
375
265
>425
315
hard template[19]
370
2
2
3
3
16]
15000
15000
~ 450
~ 320
(molar ratio = 1:5). The solution was heated at 80 °C until most of the water
MnOx-FeOx[16]
[b]
had evaporated. The solid was dried for 16 h at 120 °C and then calcined
in static air in two steps; 2 h at 300 °C and then 2 h at 500 °C.[18] The iron
Clay honeycomb
1
0000
2300
359
420
473
[48]
2 3
O
oxide obtained with this method was named Fe
gold was deposited by deposition-precipitation, with a target gold loading
of 5 wt %, yielding a catalyst denoted as 5Au Fe -SC.
2 3
O -SC and subsequently,
monolith 11%Fe
LaCaFeOx perovskite[49]
10000
2000
2000
5000
20000
13000
13000
15000
325
362
217
210
375
384
228
325
400
400
238
370
2 3
O
[50]
2 4
O
ZnFe
Pt/ZnFe
PdCeFe[51]
Au/FeOx[52]
nanocasting[a]
nanocasting[a]
[50]
O
4
2
2 3
Finally, a nanocrystalline Fe O was also synthesized by dissolving
iron (II) nitrate (Fluka, purity > 98%) in deionised water. This solution was
evaporated to dryness, and then dried for 16 h at 120 ºC, and finally
calcined in static air at 500 ºC for 4 h.[18] The iron oxide obtained with this
7
1000
3600
300
Fe O
2 3
8000
8000
100000
100000
317
268
370
334
418
370
method was named Fe
deposition-precipitation, with a target gold loading of 3 wt %, giving a
catalyst denoted as 3AuFe -C.
2 3
O -C and subsequently, gold was deposited by
Au/Fe O
2 3
2 3
O
[
[
a] This work.
b] 3 m2 of exposed area of catalyst and 300 ml/min
Characterization of materials.
The KIT-6 template, Fe
characterized by N adsorption at -196 ºC, using a Micromeritics ASAP
020 apparatus. Samples were degassed at 150 ºC prior to analysis. From
these data, the following textural parameters were calculated. Multi-point
Brunauer–Emmet–Teller (BET) surface area (SBET) was estimated over
the relative pressure range from 0.05 to 0.25. The total pore volume (VT)
was calculated using the adsorbed volume at a relative pressure of 0.95.
2 3
O supports and catalyst samples were
Finally, the activity of the most active catalyst shown in this
work was benchmarked versus other iron-based catalysts
reported in the literature in Table 5. 6AuFe -HT displays a
remarkable high reactivity, although those containing Pd or Pt
show better performance.
2
2
2
O
3
0
The pore size distribution and mean pore size (d ) of the mesoporous
materials were analysed using the Barrett–Joyner–Halenda (BJH) and the
NL-DFT methods applied to the adsorption branch of the isotherm and
assuming cylindrical pore geometry.
Conclusions
The addition of gold by a deposition-precipitation method to iron
oxide prepared using a hard template has led to a substantial
enhancement in the total oxidation of propane. This positive effect
of gold does not take place if other iron oxides are used as
supports. The enhanced behaviour of gold catalysts supported on
Powder X-ray diffraction (XRD) was used to identify the crystalline
phases present in the samples. A Bruker D8 Advance diffractometer with
a monochromatic Cu Kα X-ray source operated at 40 kV and 40 mA was
employed. The experimental patterns were calibrated against a silicon
standard and the crystalline phases were identified by matching the
experimental patterns to the JCPDS powder diffraction file database.
nanocast Fe
the support.
2
O
3
is mainly related to the porous nanostructure of
In these catalysts a stronger metal-support
interaction takes place, facilitating oxygen activation at the
interface. DFT studies have concluded that formation of defects
are more favourable and there is better availability of lattice
X-ray photoelectron spectroscopy (XPS) measurements were made
on a Kratos Axis ultra DLD photoelectron spectrometer using a non-
monochromatized Mg K X-ray source (hν = 1253.6 eV). An analyser pass
α
energy of 50 eV was used for survey scans and 20 eV for detailed scans.
Binding energies were referenced to the C1s peak from adventitious
carbonaceous species, assumed to have a binding energy of 284.8 eV.
2 3
oxygen in the Au/Fe O -HT catalysts which is related to both the
surface roughness of the support and the formation of a Si-Fe
mixed oxide phase. The presence of small gold nanoparticles,
which are stabilised due to the nanostructured nature of the
support (microporosity and mesoporosity), also plays an
important role in the enhanced reactivity.
XPS data were analysed using CasaXPS software.
A Gaussian–
Lorentzian shape function was used to peak fit the corrected spectra.
Iterations were performed using the Marquardt method. Relative standard
deviations were always lower than 1.5%.
Morphological and structural characterization of the sample was
performed by transmission electron microscopy (TEM) by using an FEI
Tecnai G F20 microscope equipped with a FEG source and operated at
Experimental Section
2
Synthesis of materials.
200 kV. The samples were prepared by sonication in absolute ethanol for
a few minutes, and a drop of the resulting suspension was deposited onto
a holey-carbon film supported on a copper grid, which was subsequently
dried. Higher resolution imaging experiments were carried out on dry
dispersed samples in an aberration corrected JEOL ARM 200CF scanning
transmission electron microscope (STEM) operated at 200kV. This
An ordered mesoporous Fe
vessel following a hard templating nanocasting route using a silica KIT-
template. The Fe obtained by this method was denoted Fe -HT,
and subsequently gold was deposited onto this material by a deposition-
2 3
O catalyst was prepared in an open
[
47]
6
O
2 3
2 3
O
This article is protected by copyright. All rights reserved.

ChemCatChem
10.1002/cctc.201900210
FULL PAPER
instrument was equipped with a Centurio silicon drift detector for X-ray
Energy Dispersive Spectroscopy (XEDS) analysis.
ARCHER, the UK’s national high-performance computing service,
which is funded by the Office of Science and Technology through
EPSRC’s High End Computing Programme. CJK gratefully
acknowledges funding from the National Science Foundation
Major Research Instrumentation program (GR# MRI/DMR-
1040229). Authors from UV thank the University of Valencia (UV-
INV-AE16-484416 project) and MINECO (MAT2017-84118-C2-1-
R project) for funding.
Temperature programmed reduction (TPR) analyses were carried
out in a Micromeritics Autochem 2920 instrument equipped with a thermal
conductivity detector, operated under a 50 mL min-1 10% H
temperatures between -50 and 800 °C, with a heating rate of 10 °C min .
/Ar flow at
2
-
1
DFT calculations were performed using the density functional theory
plane wave basis set code VASP, (Vienna ab-initio simulation
package).[53]-[56] A plane wave cut-off of 500 eV was employed. The iron
atom 3d, 4s, oxygen 2s, 2p, and gold 5d, 6s electrons were treated
explicitly as valence electrons and the projector augmented wave (PAW)
method was used to represent the remaining core states of all atoms.[57],[58]
2 3
Keywords: nanocasting • mesoporous α-Fe O • catalytic total
oxidation • propane • gold
The Perdew-Burke-Ernzerhof (PBE) functional was used throughout this
work.[
59],[60]
[1] A. Corma, H. Garcia, Chem. Soc. Rev. 2008, 37, 2096-2126.
2] S. Scirè, L.F. Liotta, Appl. Catal. B. 2012, 125, 222-246.
[3] R.D. Waters, J.J. Weimer, J.E. Smith, Catal. Letters 1995, 30, 181-188.
[4] F. Ying, S. Wang, C-T. Au, S-Y Lai, Gold Bull. 2010, 43, 241-251.
An on-site Coulomb interaction correction (PBE+U) following
Dudarev’s approach (Ueff=U – J) was applied to describe the strongly
correlated d-states of iron.[61] We have applied a Ueff of 4.0 eV. The Fe3+
ions in the weak field of the oxide lattice have a formal d5 configuration.
The magnetic moment of ±5 in the sequence + − − + + − − + for the iron
layers in the direction perpendicular to the surface plane were used
throughout this work. Tests with alternative ordering patterns confirmed
that this gives an optimised structure with the lowest energy magnetic
moment arrangement. These parameters were set based on our previous
work in which we have shown good agreement with experimental data for
[
[
[
[
[
5] L. Delannoy, K. Fajerwerg, P. Lakshmanan, C. Potvin, C. Methivier, C.
Louis, Appl. Catal. B. 2010, 94, 117-124.
6] P. Lakshmanan, L. Delannoy, V. Richard, C Methivier, C Potvin, C. Louis,
Appl. Catal. B. 2010, 96, 117-125.
7] D. Andreeva, P. Petrova, J.W. Sobczak, L. Ilieva, M. Abrashev, Appl.
Catal. B. 2006, 67, 237-245.
8] G.C. Bond, D.T. Thompson, Gold Bull. 2000, 33, 41-50.
the lattice parameters, bulk moduli, the density of states, band gap, and
system.[
34],[62]
[9] B. Liu, C, Li, Y. Zhang, Y. Liu, W. Hu, Q. Wang, L. Han, J. Zhang, Appl.
Catal. B. 2012, 111–112, 467-475.
Fe site magnetic moment for the α-Fe
O
2 3
To accommodate
the gold nanoparticle, a vacuum gap of 25 Å perpendicular to the surface
of the α-Fe (0001) slab was introduced. The large surface area of the
[
10] J. Zhang, Y. Jin, C. Li, Y. Shen, L. Han, Z. Hu, X. Di, Z. Liu, Appl. Catal.
B. 2009, 91, 11-20.
2 3
O
slab allows structural optimisations to be carried out at the Γ-point. A dipole
correction along the z-direction of the slab was applied in all calculations.
[
11] T. García, B. Solsona, S.H. Taylor, in Handbook of advanced methods
and processes in oxidation catalysis, Imperial College Press: UK. 2014,
pp. 51-90.
Catalytic testing.
[
[
[
[
[
[
[
[
12] B.E. Solsona, T. García, C.D. Jones, S.H. Taylor, A.F. Carley, G.J.
Hutchings, Appl. Catal. A. 2006, 312, 67-76.
The propane oxidation activity of the catalysts was measured using
a fixed-bed laboratory microreactor. For each experiment, 100 mg of
powdered catalyst was placed in a 1/2inch o.d. quartz reactor tube. The
reactor feed contained 8000 ppm propane in air, with a flow rate to achieve
13] T.V. Choudhary, S. Banerjee, V.R. Choudhary, Appl. Catal. A. 2002, 234,
1-23.
14] S. Minicò, S. Scirè, C. Crisafulli, R. Maggiore, S. Galvagno, Appl. Catal.
B. 2000, 28, 245-251.
-
1
a gas hourly space velocity (GHSV) of 100,000 h . Reactants and
products were analysed by an on-line gas chromatograph fitted with a
thermal conductivity and a flame ionization detector. Two chromatographic
15] S. Scirè, S. Minicò, C. Crisafulli, S. Galvagno, Catal. Comm. 2001, 2,
229-232.
16] M. Baldi, V. Sánchez-Escribano, J.M. Gallardo Amores, F. Milella, G.
Busca, Appl. Catal. B 1998, 17, 175-182.
columns were employed: (i) Porapak Q (for CO
ii) Molecular Sieve 5A (to separate CO, O
2
and hydrocarbons) and
and N ). The difference
(
2
2
17] F.G. Duran, B.P. Barbero, L.E. Cadus, C. Rojas, M.A. Centeno, J.A.
Odriozola, Appl. Catal. B 2009, 92, 194-201.
between the inlet and outlet concentrations was used to calculate
conversion data. In order to corroborate these data, the chromatographic
area of CO was used as the comparative reference. Measured carbon
2
balances were always 100 ± 2%. Blank experiments were conducted in an
empty reactor, which showed negligible activity over the temperature
range used in this study.
18] Y. Xia, H. Dai, H. Jiang, L. Zhang, J. Deng, Y. Liu, J. Hazard. Mater. 2011,
186, 84-91.
19] B. Solsona, T. García, R. Sanchis, M.D. Soriano, M. Moreno, E.
Rodríguez-Castellón, S. Agouram, A. Dejoz, J.M. López Nieto, Chem.
Eng. J. 2016, 290, 273-281.
[
[
20] Y. Zhang, J. Deng, L. Zhang, H. Dai, Chin. Sci. Bull. 2014, 59, 3993-4002.
21] S. Carabineiro, N. Bogdanchikova, P.B. Tavares, J.L. Figueiredo, RSC
Adv. 2012, 2, 2957-2965.
Acknowledgements
[
[
22] Z. Zhong, J. Ho, J. Teo, S. Shen, A. Gedanken, Chem. Mater. 2007, 19,
4
776-4782.
23] M. Mao, H. Lv, Y. Li, Y. Yang, M. Zeng, N. Li, X. Zhao, ACS Catal. 2016,
, 418-427.
T. García would like to acknowledge the Spanish Ministry of
Education, Culture and Sports through its Mobility program for
stays of senior researchers and professors in foreign countries
6
[24] F. Kleitz, S.H. Choi, R. Ryoo, Chem. Comm. 2003, 39, 2136-2137.
[25] Y. Ren, Z. Ma, R.E. Morris, Z. Liu, F. Jiao, S. Dai, P.G. Bruce, Nat. Comm.
(
MECD, grant number PRX15/00144) for funding. Computational
2
013, 4, 2015.
26] K. Zhao, H. Tang, B. Qiao, L. Li, J. Wang, ACS Catal. 2015, 5, 3528-
539.
27] S. Minicò, S. Scirè, C. Crisafulli, S. Galvagno, Appl. Catal. B 2001, 34,
77-285.
facilities supplied by Wales’ national supercomputing service
provider, High-Performance Computing (HPC) Wales, are
gratefully acknowledged. Via our membership of the UK’s HPC
Materials Chemistry Consortium, which is funded by EPSRC
[
[
3
2
(
EP/L000202), this work made use of the facilities of HECToR and
This article is protected by copyright. All rights reserved.

ChemCatChem
10.1002/cctc.201900210
FULL PAPER
[
[
[
[
[
[
28] A.P. Grosvenor, B.A. Kobe, M.C. Biesinger, N.S. McIntyr, Surf. Interface
Anal. 2004, 36, 1564-1574.
29] A.M. Venezia, G. Pantaleo, A. Longo, G. Di Carlo, M.P. Casaletto, F.L.
Liotta, G. Deganello, J. Phys. Chem. B 2005, 109, 2821-2827.
30] K. Park, J.H. Jung, H. Seo, O. Kwon, Microp. Mesop.Mat. 2009, 121,
219-225.
31] F.G.E. Nogueira, J.H. Lopes, A.C. Silva, R.M. Lago, J.D. Fabris, L.C.A.
Oliveira, Appl. Clay Sci. 2011, 51, 385-389.
32] M.A. Soria, P. Pérez, S.A.C Carabineiro, F.J. Maldonado-Hódar, A.
Mendes, L.M. Madeira, Appl. Catal. A 2014, 470, 45-55.
33] H. Tang, Y. Zeng, D. Liu, D. Qu, J. Luo, K. Binnemans, D.E. De Vos, J.
Fransaer, D. Qu, S-G. Sun, Nano Energy 2016, 26, 131-138.
34] K.L. Howard, D.J. Willock, Faraday Disc. 2011, 152, 135-151.
35] P. Schlexer, D. Widmann, R. J. Behm, G. Pacchioni, ACS Catal. 2018,
[
[
8(7), 6513-6525.
[
[
[
36] G. Nover, G. Will, Z. Kristallogr. 1981, 155, 27-45.
37] O. Ballet, H. Fuess, T. Friezche, Phys. Chem. Minerals 1987, 15, 54-58.
38] C.A. Downing, B Ahmady, C.R.A. Catlow, N.H. de Leeuw, Phil Trans R
Soc A 2013, 371, 20110592.
[
39] C. Li, Y. Shen, M. Jia, S. Sheng, M.O. Adebajo, H Zhu, Catal. Commun.
2008, 9, 355-361.
[
[
40] N. Bahlawaue, Appl. Catal. B 2006, 67, 168-176.
41] B. Solsona, T.E. Davies, T. García, I. Vázquez, A. Dejoz, H. Taylor, Appl.
Catal. B 2008, 84, 176-184.
[
[
42] B. Solsona, M. Pérez-Cabero, I. Vázquez, A. Dejoz, T. García, J.
Álvarez-Rodríguez, J. El-Haskouri, D. Beltrán, P. Amorós, Chem. Eng. J.
2
012, 187, 391-400.
43] X.Y. Deng, B.K. Min, A. Guloy, C.M. Friend, J. Am. Chem. Soc. 2005,
27, 9267-9270.
1
[
[
44] Z.P. Liu, P. Hu, A. Alavi, J. Am. Chem. Soc. 2002, 124, 14770-14779.
45] M. Kotobuki, R. Leppelt, D.A. Hansgen, D.R.J. Widmann, J. Behm, Catal.
2009, 264, 67-76.
[
[
46] T.F. Zhang, S.M. Driver, S.J. Pratt, D.A. King, Surf. Sci. 2013, 615, 1-5.
47] F. Jiao, A. Harrison, J.C. Jumas, A.V. Chadwick, W. Kockelmann, P.G.
Bruce, J. Am. Chem. Soc. 2006, 128, 5468-5474.
[
[
48] M. Assebban, A. El Kasmi, S. Harti, T. Chafil, J. Hazard. Mat. 2015, 300,
5
90-597.
49] B. P. Barbero, J. A. Gamboa, L. E. Cadús, Appl. Cat. B, 2006, 65 (1-2),
1-30.
2
[
[
50] M. Zawadzki, J. Okal. Catal. Today, 2015, 257, 136-143.
51] J.-Y. Luo, M. Meng, J.-S. Yao, X.-G. Li, Y.-Q. Zha, X. Wang, T.-Y. Zhang,
Appl. Catal. B., 2009, 87 (12), 92-103.
[
52] M. Gasior, B. Grzybowska, K. Samson, M. Ruszel, J. Haber, Catal.
Today, 2004, 91–92, 131-135.
[
[
[
[
[
[
[
53] G. Kresse, J. Hafner, Phys. Rev. B 1993, 47, 558-561.
54] G. Kresse, J. Hafner, Phys. Rev. B 1994, 49, 14251-14269.
55] G. Kresse, J. Furthmüller, Comp. Mater. Sci. 1996, 6, 15-50.
56] G. Kresse, J. Furthmüller, Phys. Rev. B 1996, 54, 11169-11186.
57] P.E. Blochl, Phys. Rev. B 1994, 50, 17953-17979.
58] G. Kresse, D. Joubert, Phys. Rev. B 1999, 59, 1758-1775.
59] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 1996, 77, 3865-
3868.
[
[
60] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 1997, 78, 1396.
61] S.L. Dudarev, G.A. Botton, S.Y. Savrasov, C.J. Humphreys, A.P. Sutton,
Phys. Rev. B. 1998, 57, 1505-1509.
[
62] S.W. Hoh, L. Thomas, G. Jones, D.J. Willock, Research on Chemical
Intermediates 2015, 41, 9587-9601.
This article is protected by copyright. All rights reserved.

ChemCatChem
10.1002/cctc.201900210
FULL PAPER
Entry for the Table of Contents (Please choose one layout)
Layout 1:
FULL PAPER
Confined gold particles on the
Tomás García,* José M. López,
Benjamín Solsona,* Rut Sanchis, David
J. Willock, Thomas E. Davies, Li Lu, Qian
He, Christopher J. Kiely, and Stuart H.
Taylor*
nanocasted Fe
2
O
3
facilitates the
production of highly reactive oxygen
vacancies at the metal-support
interface.
Page No. – Page No.
The key role of nanocasting in gold-
based Fe
2
O
3
catalysts for oxygen
activation at the metal-support
interface
This article is protected by copyright. All rights reserved.