Chemical Imaging of Mixed Metal Oxide Catalysts for Propylene Oxidation: From Model Binary Systems to Complex Multicomponent Systems

Article abstract of DOI:10.1002/cctc.202100054

Industrially-applied mixed metal oxide catalysts often possess an ensemble of structural components with complementary functions. Characterisation of these hierarchical systems is challenging, particularly moving from binary to quaternary systems. Here a quaternary Bi?Mo?Co?Fe oxide catalyst showing significantly greater activity than binary Bi?Mo oxides for selective propylene oxidation to acrolein was studied with chemical imaging techniques from the microscale to nanoscale. Conventional techniques like XRD and Raman spectroscopy could only distinguish a small number of components. Spatially-resolved characterisation provided a clearer picture of metal oxide phase composition, starting from elemental distribution by SEM-EDX and spatially-resolved mapping of metal oxide components by 2D Raman spectroscopy. This was extended to 3D using multiscale hard X-ray tomography with fluorescence, phase, and diffraction contrast. The identification and co-localisation of phases in 2D and 3D can assist in rationalising catalytic performance during propylene oxidation, based on studies of model, binary, or ternary catalyst systems in literature. This approach is generally applicable and attractive for characterisation of complex mixed metal oxide systems.

Full text of DOI:10.1002/cctc.202100054

Full Papers
doi.org/10.1002/cctc.202100054
ChemCatChem
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
Chemical Imaging of Mixed Metal Oxide Catalysts for
Propylene Oxidation: From Model Binary Systems to
Complex Multicomponent Systems
[a]
[a]
[a, b]
[c]
[d, e, f]
Paul Sprenger, Matthias Stehle, Abhijeet Gaur,
Jana Weiß, Dennis Brueckner,
[d]
[d]
[g, h]
[i]
[i]
Yi Zhang, Jan Garrevoet, Jussi-Petteri Suuronen,
Michael Thomann, Achim Fischer,
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
[a, b]
[a, b]
Jan-Dierk Grunwaldt,*
and Thomas L. Sheppard*
Industrially-applied mixed metal oxide catalysts often possess
an ensemble of structural components with complementary
functions. Characterisation of these hierarchical systems is
challenging, particularly moving from binary to quaternary
metal oxide phase composition, starting from elemental
distribution by SEM-EDX and spatially-resolved mapping of
metal oxide components by 2D Raman spectroscopy. This was
extended to 3D using multiscale hard X-ray tomography with
fluorescence, phase, and diffraction contrast. The identification
and co-localisation of phases in 2D and 3D can assist in
rationalising catalytic performance during propylene oxidation,
based on studies of model, binary, or ternary catalyst systems in
literature. This approach is generally applicable and attractive
for characterisation of complex mixed metal oxide systems.
systems. Here
a
quaternary BiÀ MoÀ CoÀ Fe oxide catalyst
showing significantly greater activity than binary BiÀ Mo oxides
for selective propylene oxidation to acrolein was studied with
chemical imaging techniques from the microscale to nanoscale.
Conventional techniques like XRD and Raman spectroscopy
could only distinguish
a small number of components.
Spatially-resolved characterisation provided a clearer picture of
Introduction
visualised as heterogeneous ensembles of crystalline and
amorphous oxide phases intermixed on at least the micrometre
[6–7]
Mixed metal oxides (MMOs) are involved in many heteroge-
neous catalytic processes for industrial synthesis of specialty
scale.
Due to this high complexity, simplified particle models
and characterisation of model binary systems have historically
[1]
chemicals, and are prominently applied in selective oxidation
been used to illustrate and predict the interaction of individual
[2–4]
[8–10]
of hydrocarbons.
oxide catalysts are especially relevant for propylene oxidation
For example, quaternary BiÀ MoÀ CoÀ Fe
MMO phases.
The present understanding of BiÀ MoÀ CoÀ Fe
oxide interactions is: 1) bismuth molybdates (α-Bi Mo O , β-
2
3
12
[5]
to acrolein and acrylic acid. Acrylic acid alone is a major
commodity chemical with global production on the scale of
several megatons per year. MMO catalysts often show remark-
able enhancements in catalytic performance through coopera-
tion of their different structural components, and can be
Bi Mo O and γ-Bi MoO ) enable selective propylene oxidation,
2 2 9 2 6
2) iron molybdates (FeMoO /Fe Mo O ) improve the availability
4
2
3
12
of lattice oxygen at the reactive centres, and 3) cobalt
molybdates (α-CoMoO and β-CoMoO ) act as a host structure
4
4
[9,11]
or support phase.
However, the precise composition and
[a] Dr. P. Sprenger, M. Stehle, Dr. A. Gaur, Prof. J.-D. Grunwaldt,
Dr. T. L. Sheppard
[f] D. Brueckner
Department Physik
Universität Hamburg
Hamburg 22761 (Germany)
Institute for Chemical Technology and Polymer Chemistry
Karlsruhe Institute of Technology
Karlsruhe 76131 (Germany)
E-mail: grunwaldt@kit.edu
[g] Dr. J.-P. Suuronen
ESRF - The European Synchrotron
Grenoble 38000 (France)
[h] Dr. J.-P. Suuronen
thomas.sheppard@kit.edu
[
b] Dr. A. Gaur, Prof. J.-D. Grunwaldt, Dr. T. L. Sheppard
Institute of Catalysis Research and Technology
Karlsruhe Institute of Technology
Eggenstein-Leopoldshafen 76344 (Germany)
E-mail: grunwaldt@kit.edu
thomas.sheppard@kit.edu
c] Dr. J. Weiß
Current Address:
Xploraytion GmbH
Berlin 10625 (Germany)
[i] Dr. M. Thomann, Dr. A. Fischer
Evonik Operations GmbH
Hanau-Wolfgang 63457 (Germany)
[
Leibniz Institute for Catalysis (LIKAT)
Rostock 18059 (Germany)
Supporting information for this article is available on the WWW under
https://doi.org/10.1002/cctc.202100054
[d] D. Brueckner, Dr. Y. Zhang, Dr. J. Garrevoet
Deutsches Elektronen-Synchrotron DESY
Hamburg 22607 (Germany)
©
2021 The Authors. ChemCatChem published by Wiley-VCH GmbH. This is
an open access article under the terms of the Creative Commons Attribution
Non-Commercial NoDerivs License, which permits use and distribution in
any medium, provided the original work is properly cited, the use is non-
commercial and no modifications or adaptations are made.
[
e] D. Brueckner
Faculty of Chemistry and Biochemistry
Ruhr University Bochum
Bochum 44801 (Germany)
ChemCatChem 2021, 13, 1–12
1
© 2021 The Authors. ChemCatChem published by Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cctc.202100054
ChemCatChem
function of multicomponent catalysts is naturally more complex
than particle models or binary systems may suggest. In
particular, the potential interaction or cooperative function of
specific phases often cannot be accurately determined based
on studies of model or binary systems alone. For example,
catalytic performance may be influenced also by pure metal
oxide phases (e.g., MoO3 or CoO) as well as mixed or
intermediate phases (e.g., Fe Co MoO or Bi FeMo O ) that
the distribution of each elemental component in 3D within the
particle ensemble. Furthermore, the complex crystalline phase
structure was partially deconvoluted within a single catalyst
particle using hard X-ray diffraction microtomography (XRD-CT),
which can non-invasively identify and locate individual phase
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
[24–28]
contributions in 3D space.
To highlight the power of these
advanced 3D techniques, we compare them to global character-
isation of the MMO catalyst after reaction by PXRD, XAS and
Raman spectroscopy, which could isolate relatively few struc-
tural components. By using such modern spatially-resolved
characterisation tools, we can re-evaluate long-standing MMO
particle models, such as olefin oxidation over BiÀ MoÀ CoÀ Fe
oxide catalysts, which can be refined in future using the
chemical imaging approach shown here.
x
1-x
4
3
2
12
[12–13]
may act as selective, promotor, inert or unselective phases.
Consequently this complicates the collection of meaningful
data describing the presence and interaction of individual
phases in industrially-relevant systems, such as quaternary
MMOs.
In selective propylene oxidation, four-component
BiÀ MoÀ CoÀ Fe oxide systems are known to show enhanced
catalytic activity compared to pure binary bismuth
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
[
9,14–15]
molybdates.
At the same time, for MMO catalysts the Results and Discussion
choice of synthesis method and pre-treatment can strongly
impact catalytic performance.
[16–18]
As catalyst complexity
Catalytic tests
increases together with the number of individual metal oxide
components present (e.g., from binary bismuth molybdates to
quaternary BiÀ MoÀ CoÀ Fe oxide systems), the challenge of
monitoring the distribution and interaction of the individual
MMO components likewise becomes greater. Characterisation
of complex quaternary MMOs is therefore recognised as a
significant challenge, but also necessary in order to rationalise
the often superior catalytic activity and selectivity observed as a
A quaternary BiÀ MoÀ CoÀ Fe oxide catalyst (9/52/29/11 mol%)
was prepared by a hydrothermal synthesis method reported
[18]
previously. It represents a modern complex-structured MMO
catalyst and serves as a model system to those used in industry.
The hydrothermally-prepared BiÀ MoÀ CoÀ Fe oxide catalyst was
tested for its catalytic performance in propylene oxidation. For
comparison, phase pure α-Bi Mo O prepared by hydrothermal
2
3
12
[7,9]
consequence of cooperative phase interaction, for example.
synthesis (HS), and a BiÀ MoÀ CoÀ Fe oxide reference catalyst
prepared by co-precipitation (CP), were also tested. The specific
surface area was measured for each catalyst after synthesis and
The use of spatially-resolved and phase selective analytical
techniques therefore becomes increasingly relevant in order to
[19–21]
2
À 1
accurately address the catalyst structure.
For example,
calcination but prior to catalytic testing, with values of 5 m g
2
À 1
common techniques such as laboratory powder X-ray diffrac-
tion (PXRD) or synchrotron radiation-based X-ray absorption
spectroscopy (XAS) can reveal valuable information on crystal-
line and amorphous phases, including active site properties
such as metal oxidation state and coordination environment.
However, conventional bulk measurements on powders or
packed beds result in averaged data, which may lack the
sensitivity needed to identify dilute or minority species present,
particularly in quaternary MMO catalysts. Ideally, a complete
knowledge of all phases is the ultimate goal to allow full
understanding and rationalisation of both catalytic performance
and the phases contributing to this. No single analytical
technique can solve this problem, rather the careful application
of cooperative techniques is required.
(2-component
BiÀ Mo,
HS),
7 m g
(4-component
(4-component
2
À 1
BiÀ MoÀ CoÀ Fe oxide, CP), and 19 m g
BiÀ MoÀ CoÀ Fe oxide, HS) obtained. The results of catalytic
testing are shown in Figure 1.
À 1 À 1
For feed streams from 27.1 to 81.3 mmolpropylene gcatalyst
h
(weight hourly space velocity (WHSVC3H6) ranging from 0.57 to
À 1
3.42 h ), the hydrothermally-synthesised BiÀ MoÀ CoÀ Fe oxide
catalyst was best performing. For instance, at WHSV
1.14 h , 63% propylene conversion and 76% acrolein selectiv-
=
C3H6
À 1
ity were obtained. Under the same conditions, pure α-Bi Mo O
12
2
3
showed 22% conversion and 89% acrolein selectivity and the
co-precipitated BiÀ MoÀ CoÀ Fe oxide catalyst showed 54%
conversion at 63% selectivity. Furthermore, the presented
multicomponent catalyst showed a higher acrolein selectivity at
similar propylene conversions. At around 39–45% propylene
conversion, the selectivity increased in the order of: CP:
In this work, we applied a range of advanced spatially-
resolved chemical imaging methods for characterisation of
MMO catalysts across multiple length-scales. A highly active
quaternary BiÀ MoÀ CoÀ Fe oxide catalyst for selective propylene
oxidation to acrolein was taken as a case study. The chemical
imaging approach is demonstrated starting with scanning
electron microscopy (SEM) and Raman spectroscopic mapping
for 2D phase characterisation of metal oxide species over
À 1
BiÀ MoÀ CoÀ Fe oxide (64% at 1.71 h ) < HS: α-Bi Mo O (74%
2
3
12
À 1
À 1
at 0.57 h )<HS: BiÀ MoÀ CoÀ Fe oxide (80% at 2.28 h ). In
general, the acrolein selectivity increased with higher WHSVs.
The main by-products observed were CO and CO , whereas the
2
co-precipitated BiÀ MoÀ CoÀ Fe oxide showed a high selectivity
toward acrylic acid. The specific surface area of the 4-
component hydrothermally-synthesised catalyst decreased to
[22–23]
extended regions of powder samples.
Moving beyond 2D
2
À 1
composition towards entire single catalyst particles, we em-
ployed a combination of hard X-ray fluorescence nanotomog-
raphy (XRF-CT) and phase contrast holotomography to observe
7 m g after catalytic testing, closely matching the initial
values for the other catalysts. Overall, catalytic performance
tests illustrate how activity and selectivity are strongly improved
ChemCatChem 2021, 13, 1–12
www.chemcatchem.org
2
© 2021 The Authors. ChemCatChem published by Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cctc.202100054
ChemCatChem
nents. Based on the PXRD pattern acquired after catalytic
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
testing shown in Figure 2, it was not possible to distinguish
between Bi/Mo/Fe or Fe/Mo phases such as Bi FeMo O (the
3
2
12
former indicated by SEM-EDX, Figure 3) due to overlapping
reflections. In general, the PXRD reflections both before (see
ESI) and after testing were highly convoluted and accurate
phase assignment of such complex ensembles was therefore
challenging. Quantitative crystalline phase analysis can in
principle be achieved via Rietveld refinement, but is challenging
on a complex mixture of up to four components with an
unknown total number of phases present. Furthermore, PXRD is
only sensitive to crystalline phases, while MMO catalysts may
also contain amorphous or disordered features, which require
complementary characterisation by XAS or Raman spectro-
scopy, for example.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
In the current study, XAS was performed on pellets of the
quaternary hydrothermally-synthesised catalyst after reaction
(see results in ESI). Analysis indicated the presence of species
including α-Bi Mo O and γ-Bi MoO (based on Mo K edge and
2
3
12
2
6
Bi L edge) in agreement with PXRD, and α-CoMoO4 (based on
3
Co K edge). For conventional PXRD, Raman spectroscopy and
XAS however, the challenge of distinguishing major and minor
phases depends on the focal spot size of the beam relative to
the separation or distribution of microstructural features in the
Figure 1. (a) Illustration of a simple binary particle and a complex quaternary
BiÀ MoÀ CoÀ Fe oxide particle model according to Ref. [8–10]; (b) comparison
of acrolein selectivity over propylene conversion at 380°C; (c) comparison of
combined CO and CO
CP=co-precipitation, HS=hydrothermal synthesis. Data point shading
2
selectivity over propylene conversion at 380°C.
indicates WHSVC3H6 of the measurement.
Figure 2. PXRD pattern of the hydrothermally-synthesised catalyst after
catalytic tests. Assigned phases: * α-Bi Mo O (ICSD: 2650), γ-Bi MoO
2
3
12
2
6
(
ICSD: 47139),
◇
β-CoMoO
4
(JCPDS no. 21–868), Co0.7Fe0.3MoO
4
(ICSD:
for the quaternary BiÀ Mo-FeÀ Co oxide systems, which can be
explained in principle by the cooperation of interacting phases.
This includes phases which may not be present in simple binary
or model systems. Furthermore, the performance clearly also
depends on the synthetic history of the catalyst material. Due
to the excellent performance of the hydrothermally-synthesised
quaternary system, the specific phase composition and possible
interactions were investigated further by both global and
spatially-resolved techniques.
280035) and Bi FeMo O (ICSD: 45).
3
2
12
Global characterisation
Laboratory PXRD was performed to access the overall composi-
tion of crystalline phases, revealing in particular β-CoMoO4,
Co Fe MoO , α-Bi Mo O and γ-Bi MoO as major compo-
Figure 3. SEM-EDX images of hydrothermally-synthesised BiÀ MoÀ CoÀ Fe
oxide catalyst after testing with colour coding for individual metals, and the
overlap between: (a) Fe, Co, Mo; (b) Fe, Bi, Mo.
0.7
0.3
4
2
3
12
2
6
ChemCatChem 2021, 13, 1–12
www.chemcatchem.org
3
© 2021 The Authors. ChemCatChem published by Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cctc.202100054
ChemCatChem
material. Due to the averaging nature of the measurements in
this case, they are not sufficiently sensitive to minor phases. A
spatially-resolved approach in 2D or 3D is necessary to avoid
convoluted or superimposed spectra of different metal oxide
components which may be difficult to isolate.
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
SEM-EDX mapping
To illustrate the heterogeneity of the hydrothermally-synthes-
ised catalyst, SEM-EDX mapping was performed after catalytic
testing. As shown in Figure 3, larger irregular shaped particles
were found intermixed with sub-micrometre-sized and well-
defined crystals. Area mapping with SEM-EDX gave access to
the elemental distribution, revealing agglomerates of mainly
binary BiÀ Mo oxide as expected, but additionally clear
indications of mixed binary CoÀ Mo oxide and ternary BiÀ Mo-Fe
oxide species. Notably, the larger homogeneous areas coin-
cided with CoÀ Mo oxide agglomerates, with several smaller
areas associated with BiÀ Mo-Fe oxides. However, while SEM-
EDX mapping is useful in determining elemental distribution, it
is not sensitive to the phase composition of the sample (e.g.,
bismuth molybdate morphology present in BiÀ Mo oxide areas).
In principle, the phase composition can be studied with
selected area electron diffraction (SAED), but this is only
suitable for analysing local composition in a small area with
potentially invasive sample preparation. For the same reason,
transmission electron microscopy (TEM) also has limited
practical use in this case due to observed heterogeneity on the
micrometre scale and the highly localised nature of the images
which can be acquired. For more complete phase identification
of MMO catalysts in their native form as microscale particles,
alternative spatially-resolved techniques such as 2D Raman
spectroscopy are required.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
Spatially-resolved Raman spectroscopy and 2D mapping
By employing a focused laser beam, Raman spectroscopy is
especially suited for spatially-resolved 2D sample mapping and
is additionally sensitive to both crystalline and amorphous
phases. Here Raman mapping was performed on the hydro-
thermally-synthesised quaternary BiÀ MoÀ CoÀ Fe oxide catalyst
after testing by raster-scanning a focused beam over the surface
of the catalyst sample (532 nm laser, laser spot size: 0.38 μm,
2
area 56×56 μm , 0.2 μm raster step size; Figure 4a). For each of
the 78,400 single spectra recorded, a direct classical least
squares (DCLS) component analysis was performed using
reference spectra including pure α-Bi Mo O , β-Bi Mo O , γ-
2
3
12
2
2
9
Bi MoO ,
α-CoMoO4,
β-CoMoO4,
β-FeMoO4,
Fe Mo O ,
2 3 12
2
6
Figure 4. (a) 2D Raman spectroscopic phase distribution obtained from
DCLS-based component analysis and plotted over the optical microscopic
image of HS: BiÀ MoÀ CoÀ Fe oxide after testing. (b) Averaged Raman
spectrum for the whole area, and spectra extracted from regions of interest,
which were colour-coded assigned to certain phases based on DCLS
Bi FeMo O and MoO3 (Figure 4c-h, not all references are
3
2
12
shown).
Notably, the phase distribution observed was quite hetero-
geneous. Numerous overlapping regions of α-Bi Mo O (dark
component analysis: (c) α-Bi
Bi MoO (red), (f) α/β-CoMoO
(yellow).
2
Mo
3
O
12 (blue), (d) β-Bi
2
Mo
2
O
9
(magenta), (e) γ-
FeMo
2
3
12
2
6
4
(cyan), (g) FeMoO
4
(green), (h) Bi
3
2 12
O
blue spectrum) and γ-Bi MoO (red spectrum) were observed as
2
6
larger well-distributed agglomerates (Figure 4c,e). This can also
be observed in the spectrum extracted from ROI 3 for example,
ChemCatChem 2021, 13, 1–12
www.chemcatchem.org
4
© 2021 The Authors. ChemCatChem published by Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cctc.202100054
ChemCatChem
where in addition to main features from the γ-Bi MoO
scopy mapping may be sufficient to address the structure, and
the much more complex quaternary BiÀ MoÀ CoÀ Fe oxide
systems. In addition, the limitations of 2D mapping over a
region of the catalyst surface (according to the penetration
depth), compared to examination of larger bulk particles should
also be considered. For effective chemical imaging at higher
spatial resolution on larger samples, it is therefore necessary to
consider hard X-rays as the more suitable probe.
2
6
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
À 1
reference, additional bands at 901 cm (α-Bi Mo O ) and
2
3
12
À 1
9
35 cm (β-CoMoO ) are clearly visible. This indicates an
4
intermixing of the above phases. On the other hand, β-Bi Mo O
9
2
2
(pink spectrum) appeared mainly in isolated domains, indicated
by the good match of the relevant reference with the ROI 2
spectrum (Figure 4d). However, the β-Bi Mo O content was low
2
2
9
and it was neither detected by PXRD, nor in averaged Raman
spectra of the scanned area, due to overshadowing or
convolution with more intense neighbouring bands. The main
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
visible component was α/β-CoMoO (light blue spectrum), as
X-ray holotomography and XRF-CT
4
evident from the average spectrum (Figure 4b) and the large
proportion of matching coloured areas in Figure 4a. It is
uncertain whether this phase was particularly highly concen-
trated, or whether it could essentially not be resolved from
neighbouring phases due to the spatial resolution of the
The highly brilliant hard X-rays produced by modern synchro-
tron light sources offer excellent opportunities for catalyst
characterisation by means of chemical imaging in 2D and 3D. In
particular, hard X-ray tomography offers high penetration depth
for measuring samples with larger diameter, compared to
optical or electron microscopy. The acquisition of 3D spatially-
resolved data using tomography is therefore feasible on large
samples, which allows for deconvolution of internal structural
features. A range of contrast modes are also available based on
the interactions of hard X-rays with matter. Here, a single
particle of the hydrothermally-synthesised quaternary
BiÀ MoÀ CoÀ Fe oxide catalyst after catalytic testing was studied
at beamline ID16B of the European Synchrotron Radiation
Facility (ESRF), first using full-field phase contrast holotomog-
raphy, followed by nanofocused XRF-CT, with beam attenuation
measured simultaneously by tomographic scanning transmis-
sion X-ray microscopy (STXM-CT).
Phase contrast holotomography produced a macroscopic
view of the entire catalyst particle with around 25 nm voxel size
and an assumed resolution of >100 nm (Figure 5a), allowing
analysis of structurally diverse regions based on their relative
electron density. Three representative slices of the catalyst
particle are shown, from which needle or plate-like structures
were observed (Figure 5b). Comparing the holotomogram to
the sample attenuation measured at the same position by
STXM-CT (Figure 5c) supported the presence of highly absorb-
ing components attributed to metal oxides, together with
regions of low attenuation (see ESI, possibly from carbon
deposits or glue which was not fully segmented during visual-
isation). A non-discriminating imaging method such as holoto-
mography or STXM-CT is therefore highly useful for visualising
global sample structure regardless of composition, although
these methods cannot directly identify features contributing to
differing contrast regions.
measurement (see ESI). FeMoO (green spectrum) was occasion-
4
ally present in smaller particles and highly dispersed, compared
to the Co and Bi molybdate phases mentioned previously. A
Bi FeMo O mixed phase (yellow spectrum) was also found in
3
2
12
the form of randomly dispersed micrometre-sized particles,
indicated by the good match between the reference spectrum
and ROI 6 (Figure 4h). This is in agreement with the presence of
Bi/Mo/Fe domains found via SEM-EDX. However, no Raman
bands corresponding to Fe Mo O and MoO were found.
2
3
12
3
The apparent low amount of binary Fe Mo O phases by
2
3
12
both Raman mapping and SEM-EDX on the microscale is
significant, as it implies incorporation of iron into a ternary
BiÀ Mo-Fe oxide form, which has been linked to improved
[11,29–30]
acrolein selectivity as observed here.
While limited
amounts of MoO3 were suggested to be beneficial for the
catalyst in terms of replenishing reduced oxide species, excess
MoO has been linked with decreased activity for propylene
3
oxidation. This was particularly observed for catalysts with
[9,31]
relatively higher Mo content.
The absence of any visible
isolated MoO3 species here is also in agreement with the
generally high catalytic performance.
For proper data interpretation, the limitations of Raman
spectroscopy must be considered. Since inelastic Raman
scattering is generally a weak effect, quantification is not easily
achievable especially for minority phases or those with weak
Raman bands. The absence of some phases in the measured
Raman spectrum might originate from low scattering intensity,
even though spatially resolved acquisition was performed. In
addition, the resolution of >0.38 μm obtained here (see ESI) is
mainly restricted by the diffraction limit of visible light (532 nm,
1
00x objective, NA=0.85), and might not be high enough to
Consequently, XRF-CT was applied to isolate individual
signals of Bi, Mo, Co and Fe within the catalyst particle
(Figure 5d). The needle or plate-shaped structures observed by
holotomography were thus composed mainly of Mo, in addition
resolve neighbouring phases completely. Furthermore, phase
pure references are needed for DCLS analysis, which is
challenging and time consuming in the absence of standard
databases for Raman spectroscopy, such as those which exist
for XRD data. While the resolution can be increased using
[
34]
to contributions of Bi. This is consistent with previous studies.
However, while Mo was distributed rather homogeneously
throughout the entire sample, Co and Fe were often observed
as a pair and were mostly segregated from the other metals
(see overlay images in ESI). This indicates that phases composed
of at least two components (Bi and Mo) formed larger or more
extended structures, while Co and Fe may have formed in a
[32–33]
methods like tip enhanced Raman spectroscopy (TERS),
this
may not be suitable for representative imaging of large samples
used in heterogeneous catalysis.
A clear distinction should be made between investigation of
simple binary BiÀ Mo oxide systems, for which Raman spectro-
ChemCatChem 2021, 13, 1–12
www.chemcatchem.org
5
© 2021 The Authors. ChemCatChem published by Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cctc.202100054
ChemCatChem
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
Figure 5. (a) Holotomography visualisation of the entire HS: BiÀ MoÀ CoÀ Fe oxide particle following catalytic tests, with the sample glued to the top of a
tomography pin; identical orthogonal slices extracted from the tomographic volume showing individual signals obtained from (b) holotomography; (c) STXM-
CT; and (d) normalised XRF-CT showing absolute intensity (signal including 0–100% of histogram values) of each metal component.
more particulate or agglomerated manner around these
structures, in agreement with observations in the literature.
after testing. Crystalline phases within the entire particle were
measured with a beam spot size of around 650 nm. This
allowed us to deconvolute the diffraction patterns and identify
local phases within the catalyst particle, which is not possible
using conventional PXRD either in the laboratory or at the
synchrotron. Figure 6 shows diffractograms of selected regions
of interest (ROI 1–3) which revealed a heterogeneous phase
distribution across different regions of the catalyst particle,
compared to the average diffractogram of one complete slice
of the catalyst particle (Sum ROI – representative of a ‘typical’
PXRD pattern as shown in Figure 2). At the same time, the
relatively non-linear background shown in the average diffrac-
tion pattern further suggests some contribution of amorphous
species, although these cannot be identified with XRD. While
many crystalline features were still convoluted, based on
available reference patterns it was possible to partly assign and
isolate several individual phases (Figure 6), including: β-
[11]
Notably, Co and Fe species often overlapped, occasionally also
with Bi (see ESI). Due to the relatively high global composition
of Mo observed, this probably also indicates the presence of
some CoÀ Fe-Mo oxide species. This may partly explain the
relatively high selectivity towards acrolein, since the presence
of mixed CoÀ Fe molybdates was shown to be beneficial in this
[11]
regard compared to individual CoMoO or FeMoO phases.
4
4
Compared to SEM-EDX (Figure 3), XRF-CT has the benefit of
non-invasively probing the metal speciation within entire
catalyst particles. However, it should be noted with caution that
Co and Fe have relatively low fluorescence signal due to the
high incident X-ray energy applied (29.6 keV). In addition, there
is strong potential for self-absorption artefacts from the
relatively low energy of Co-Kα and Fe-Kα emission lines
combined with high concentration of these elements in a
relatively thick sample. In addition, none of the above methods
are sensitive to specific crystalline phase, therefore the exact
same particle was further investigated by XRD-CT.
Co Fe MoO /β-CoMoO
(2θ=10.06/10.07°),
α-Bi Mo O
0
.7
0.3
4
4
2 3 12
(11.04°) and Bi FeMo O (7.04°). Note that the features at 2θ=
3
2
12
10.06/10.07° cannot be distinguished from each other, and are
therefore given an ambiguous label to indicate the presence of
either phase, or both phases. It was furthermore possible to
present the distribution of these isolated phases within the
particle, revealing a relatively homogeneous distribution of β-
Co Fe MoO /β-CoMoO , also indicated by the strong feature
XRD-CT
XRD-CT was performed on a single particle of the hydro-
0
.7
0.3
4
4
thermally-synthesised quaternary BiÀ MoÀ CoÀ Fe oxide catalyst
at 10.06° in the Sum ROI (Figure 6a). Several closely overlapping
ChemCatChem 2021, 13, 1–12
www.chemcatchem.org
6
© 2021 The Authors. ChemCatChem published by Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cctc.202100054
ChemCatChem
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
Figure 6. XRD tomography of a single catalyst particle after reaction: (a) spatially-resolved XRD patterns of selected areas (ROI 1–3) and summed ROI of all
pixels in an entire slice (Sum ROI) – the particle boundary is illustrated by a transparent rendering; (b) integral of selected reflections (highlights at 10.07°,
1
2
1.04°, 7.04°) from the XRD patterns acquired in ROI 1–3 – the slices on the left indicate the normalised individual pixel intensity of the selected reflection in
D space, the volumes on the right show the pixels with top 15% measured intensity across all slices in 3D space; (c) illustration of the overlap between
selected reflections presented as a single orthographic slice through the 3D rendered volume (left) and the entire 3D volume (right). The selected reflections
shown contain 2θ values characteristic of β-CoMoO /Co0.7Fe0.3MoO (10.07°), α-Bi Mo O (11.04°) and Bi FeMo O (7.04°) as well as the overlap of all three
4
4
2
3
12
3
2
12
phases (λ=0.05904 nm).
but heterogeneously dispersed regions of α-Bi Mo O (11.04°)
was present together with binary bismuth molybdate phases.
The strong overlap of binary BiÀ Mo oxide and ternary BiÀ Mo-Fe
oxide phases observed at the particle scale is consistent with
the concept of phase cooperation in selective oxidation
2
3
12
and Bi FeMo O (7.04°), indicated by overlap of the intense
3
2
12
regions of these features in Figure 6b. This strongly supports
the results observed by 2D Raman spectroscopic mapping of
catalyst powder surfaces (Figure 4) and XRF nanotomography
imaging of the entire catalyst particle (Figure 5), whereby Fe
and Co species were found to agglomerate, and Bi FeMo O
[35]
catalysis described by Grasselli.
In combination with the
overlapping binary CoÀ Mo oxide or ternary CoÀ Fe-Mo oxide
observed, these results may partly explain the superior catalytic
3
2
12
ChemCatChem 2021, 13, 1–12
www.chemcatchem.org
7
© 2021 The Authors. ChemCatChem published by Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cctc.202100054
ChemCatChem
performance observed in propylene oxidation for the quater-
overview over the heterogeneous metal distribution and
clustering of certain elements. With 2D Raman spectroscopic
mapping, phase identification and distribution of both crystal-
line and amorphous phases on a sub-micrometre resolution
was monitored, allowing identification of by-phases that were
masked during previous analysis with global characterisation
methods. Deconvolution of Raman data by DCLS-based analysis
allowed a deeper insight into the catalyst composition. This
makes Raman spectroscopy especially suitable for complex
MMOs, but notably depending on whether appropriate refer-
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
[36]
nary compared to the binary system. Due to overlap between
neighbouring features, it was not possible to conclusively
isolate more individual phase contributions from the XRD-CT
data obtained in this instance. We further emphasise that
without use of spatially-resolved techniques such as XRD-CT, it
would be challenging or even impossible to conclusively
identify close mixtures of phases with similar diffraction
patterns (see references in ESI). At the same time, due to the
apparent high heterogeneity across single particles of
BiÀ MoÀ CoÀ Fe oxide shown here, localised probes such as TEM
or SAED risk being unrepresentative of catalyst composition at
the particle scale. This demonstrates the importance of
analysing the complex phase structure of such catalysts by
spatially-resolved methods capable of probing at the particle
scale. Although hindered in this case by the high complexity of
the sample, XRD-CT still demonstrates the capability to define
the spatial distribution of crystalline phases in 3D-space. Thus, it
is possible to produce a physical visualisation of the simple
particle model, even considering more complex by-phases.
Nevertheless, XRD-CT is still limited to observation of crystalline
structure, therefore in future it may be advantageous to
consider tomographic XAS, total scattering or pair distribution
function analysis, which offer the necessary sensitivity regard-
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
[38]
ence materials are available for DCLS. In general, for MMOs
that are intermixed on the micrometre-scale, Raman mapping
offers a convenient resolution. Large scanning areas can be
sampled with modern Raman spectrometers on a reasonable
timescale (ca. 2 d). Combined with the latest operando cells,
[
39–40]
MMO mapping is even possible during catalytic reactions.
Nevertheless, Raman spectroscopy has a limited material
sensitivity due to the weak scattering signal, and additionally a
limited information-depth due to high attenuation of optical
wavelengths including near-UV, visible, and near-IR regions.
While Raman spectroscopic mapping can be applied with 3D
[41]
spatial resolution (e.g., confocal microscopy),
this has
typically been limited to micron sized samples at most, and it is
difficult to imagine this applied to larger catalyst particles on
the order of >50 μm as studied here. In comparison, hard X-ray
tomography allows non-invasive analysis of large particles with
[37]
less of the presence of crystalline or amorphous components.
In the case of MMOs for selective propylene oxidation most
phases are probably crystalline, but this does not exclude
formation of non-crystalline features at surfaces or particle
interfaces. The results show the excellent potential of tomo-
graphic data compared to global analysis such as conventional
PXRD. It is of considerable interest in the future to further
address the structure of complex heterogeneous catalytic
systems such as MMOs using tomography.
a
range of contrast modes, including absorption, X-ray
fluorescence, diffraction and phase as demonstrated in this case
study. For the presented catalyst, crystalline phases and their
agglomeration and, thus, cooperation, was visualised for a
spent catalyst particle in 3D-space via XRD-CT. The crystalline
phases were identified by a relatively simple approach of
assigning specific individual reflections from known references,
but XRD analysis and phase mapping can be extended by
[27,42]
Rietveld refinement as demonstrated elsewhere.
Notably,
Structural insights from spatially-resolved analysis
this approach is most successful when powder diffraction
conditions can be maintained during tomography acquisition,
and single crystal artefacts and preferred orientation of larger
crystallites are minimised. Despite the high complexity of four-
component catalysts, the information derived in this study
strongly complements the presently oversimplified binary MMO
systems (α-, β-, γ-bismuth molybdates). Nevertheless, this
increased complexity and the presence of additional metal
oxide phases is directly responsible for the superior activity of
quaternary systems. Only through a careful characterisation
approach can the necessary information be extracted to further
develop and improve on well-known particle models which are
still in use today. In particular, while the current study presents
only ex situ characterisation, it is also crucial to consider
spectroscopic and tomographic measurements under in situ
conditions, which have the potential to track changes in the
catalyst structure and activity as a function of chemical
A
multiphase system of quaternary BiÀ MoÀ CoÀ Fe oxide
prepared by hydrothermal synthesis was studied as highly
active and selective catalyst for propylene oxidation to acrolein.
The performance of the quaternary catalyst was shown to
exceed that of a binary BiÀ Mo oxide system. The structural
complexity of this quaternary MMO catalyst is difficult to
analyse by global characterisation methods, therefore making it
an excellent representative case study for demonstrating the
importance and application of emerging spatially-resolved
analytical techniques in catalysis.
This work demonstrates that all the necessary tools are in
place for systematic investigation of complex quaternary MMO
catalysts. Indeed, a combined characterisation approach may be
necessary to conclusively identify interactions and synergy
between specific phases. Global methods including Raman
spectroscopy, PXRD and XAS are unable to address the catalyst
complexity in a sufficient manner. Therefore, a systematic
assessment was performed using 2D spatially-resolved SEM-EDX
and Raman spectroscopy mapping, and 3D spatially-resolved X-
ray micro- and nanotomography. SEM-EDX mapping gave an
[43]
environment.
ChemCatChem 2021, 13, 1–12
www.chemcatchem.org
8
© 2021 The Authors. ChemCatChem published by Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cctc.202100054
ChemCatChem
Conclusion
Catalytic performance tests
The setup for evaluation of the catalytic performance is described
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
By exploiting the synergy between spatially-resolved techni-
ques, it was possible to advance from a simplified particle
model of interacting phases, to empirical 2D and 3D images of
the catalyst particle with identification of several main- and by-
phases. A clear distinction can be made between model or
binary systems which are less structurally diverse, and the
complex quaternary systems investigated here. The latter case
requires the use of advanced spatially-resolved characterisation
tools to aid in making meaningful observations on the catalyst
structure. The strong activity of the catalyst for selective
oxidation of propylene can be rationalised in part by: (i) the
presence of mixed ternary metal oxide phases including BiÀ Mo-
Fe oxide and CoÀ Fe-Mo oxide in addition to binary BiÀ Mo
[44]
in detail in the literature.
The hydrothermally-prepared and
subsequently calcined catalyst HS: BiÀ MoÀ CoÀ Fe oxide was ground,
pressed and sieved to give a fraction of 300–450 μm. 800 mg of the
catalyst were placed in a quartz tubular reactor (6 mm inner
diameter). For preconditioning in the setup, all catalysts were
À 1
heated in synthetic air (N /O =80/20, 100 NmLmin ) to 180°C
2
2
À 1
(
T
5°C min ). Afterwards, sequential ramp steps to T
=345,
1,oven
À 1
=380, T
=380, T
4,oven
=400°C (2°Cmin ) under reaction
2
,catalyst
3,oven
conditions (N /O /C H /H O=70/14/8/8) with variation of total flow
2
2
3
6
2
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
À 1
(100, 150, 200, 300 NmLmin ) at each temperature were per-
formed. Due to the exothermicity of the reaction, the oven
temperature was regulated during testing to keep the catalyst bed
temperature at 380°C. The mass specific catalytic activity (MSCA)
was calculated according to Equation (1). The weight hourly space
_
À 1
velocity (WHSV ) was calculated according to Equation (2) with V:
C3H6
3
À 1
-3
oxide; (ii) the absence of isolated MoO , which in excess is
volume flow rate [m h ], 1: density [gm ], M: molar mass [gmol ]
3
and m: mass [g]. Calculation of propylene conversion and acrolein
known to hinder catalytic activity. These observations are
challenging without the use of complementary characterisation
tools sensitive to different elements, crystalline and amorphous
phases, and capable of deconvolution and colocation in 2D and
[44]
selectivity can be found elsewhere.
_
V
� 1
propylene
propylene
MSCA ¼
(1)
(2)
Mpropylene � mcatalyst
3
D space. Future studies should aim to provide a complete
characterisation of the MMO structure by means of better
spatial- or energy-resolution during XRD-CT or alternative
methods, such as XANES-CT for individual mapping of metal
oxidation states and coordination environment, or PDF-CT and
total scattering for high quality analysis of nanostructured
phases. A modern spatially-resolved characterisation approach
must demonstrably consider multiple length scales, detection
methods and sample sizes, in order to unlock new information
about complex MMO systems. Strategies for handling and
evaluating the large quantities of data produced by such
detailed characterisation tools are an essential concern which
must be strongly in focus moving forward.
_
V
� 1propylene
propylene
WHSVC3H6
¼
m
catalyst
Catalyst characterisation
The hydrothermally-prepared four-component catalyst (HS:
BiÀ MoÀ CoÀ Fe oxide) was characterised by N2 physisorption,
powder X-ray diffraction (PXRD), Raman spectroscopy, combined
scanning electron microscopy and energy dispersive X-ray spectro-
scopy (SEM-EDX), inductively coupled plasma optical emission
spectrometry (ICP-OES) and X-ray absorption spectroscopy (XAS).
Characterisation results and additional information for the HS: α-
[44]
Bi Mo O catalyst can be found in the literature.
Where
2
3
12
characterisation was performed after catalytic tests, this indicates
the catalyst was removed from the reactor and characterised
directly without any intermediate steps.
Experimental Section
Catalyst preparation
The specific surface area of the samples was measured by N2
physisorption at À 196°C using a BELSORP mini-II instrument
The four-component catalyst system (HS: BiÀ MoÀ CoÀ Fe oxide) was
(
MicrotracBEL). Prior to the measurement, all samples were
prepared by hydrothermal synthesis according to a recently
[
18]
degassed in vacuum at 300°C for 2 h. PXRD patterns were recorded
established procedure. The calculated precursor ratio is listed in
Table S1 (see ESI). The precursor (NH ) Mo O ·4H O was solubilised
in 20 mL distilled water and stirred for 15 min. In parallel, a solution
consisting of Bi(NO ) ·5H O, Co(NO ) ·6H O and Fe(NO ) ·9H O in
20 mL HNO3 (2 M) was prepared and stirred for 15 min. Both
solutions were combined in a Teflon® inlay and the pH was
adjusted (pH=7) by adding dropwise an aqueous NH3 solution
on a Bruker D8 Advance diffractometer (Cu K (λ=0.154 nm), Ni-K
α
β
4
6
7
24
2
filter, Vantec PSD) in the 2θ range of 8 to 80° with a step size of
0
.0165° and a dwell time of 2 s for each step. Phase assignment
3
3
2
3 2
2
3 3
2
was performed using Inorganic Crystal Structure Database (ICSD)
and Joint Committee of Powder Diffraction Standards (JCPDS)
references. Raman spectroscopy was performed on a Renishaw
inVia Reflex Raman Spectrometer using a frequency doubled Nd:
YAG laser (532 nm, 100 mW at source) and a Leica optical micro-
(25%) under continuous stirring. After additional stirring for 15 min,
the inlay was transferred to a stainless-steel autoclave. The
autoclave was heated to 180°C for 24 h and afterwards allowed to
cool down to RT (24 h). The obtained solid was filtered off (G4 glass
2
scope (100× objective). For 2D mapping an area of 56×56 μm was
measured with a raster step size of 0.2 μm and a laser spot size of
À 1
0
.38 μm. Spectra were recorded in the range of 60–1320 cm using
frit) and washed three times with 10 mL H O and three times with
2
À 1
a 2400 linesmm grating with an acquisition time of 2 s. The laser
intensity was set to 1%. SEM-EDX measurements were conducted
using a JSM 7600F (JEOL) field-emission SEM at 20 kV, and an AZtec
EDX system (Oxford Instruments). Samples were first embedded in
epoxy glue and microtomed. The elemental metal composition was
determined by ICP-OES using an Agilent 720/725-ES spectrometer.
The sample was dissolved in 6 mL hydrochloric acid, 2 mL nitric
acid and 1 mL hydrogen peroxide by using a microwave operated
1
0 mL acetone. Finally, the catalyst was dried for 48 h at room
temperature and subsequently calcined for 5 h at 320°C. The
preparation of the two-component catalyst system HS: α-Bi Mo O
2
3
12
[44]
is described elsewhere.
ChemCatChem 2021, 13, 1–12
www.chemcatchem.org
9
© 2021 The Authors. ChemCatChem published by Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cctc.202100054
ChemCatChem
at 600 W for 45 min. XPS measurements were performed using an
ESCALAB 250 xi (ThermoFisher Scientific) equipped with an Al Kα
We acknowledge KIT and DFG for financing the inVia Raman
spectrometer system (INST 121384/73-1). We thank Federico Benzi
for contributing during tomography experiments, Hermann Köhler
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
source at 1.487 keV, with a beam spot size of 900 μm (results shown
in ESI). X-ray absorption spectroscopy was performed at the ROCK
beamline of the Soleil synchrotron (St. Aubin, France). Spectra of
the catalyst sample after catalytic testing and of reference
substances were obtained on pellets of the corresponding materi-
(IKFT, KIT) for ICP-OES analysis, and acknowledge Evonik Industries
AG for scientific discussion. Open access funding enabled and
organized by Projekt DEAL.
[
45]
als. For XAS data analysis the software package IFEFFIT was used.
For further information on XAS data treatment and results see ESI.
Conflict of Interest
X-ray tomography
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
The authors declare no conflict of interest.
The spent catalyst HS: BiÀ MoÀ CoÀ Fe oxide was investigated at the
hard X-ray nanoimaging beamline ID16B of the European Synchro-
tron Radiation Facility (ESRF, Grenoble, France). Experimental details
regarding: (i) full-field X-ray holotomography; (ii) scanning X-ray
fluorescence nanotomography (XRF-CT); and (iii) scanning trans-
mission X-ray tomography (STXM-CT) can be found in previous
Keywords: electron microscopy
· Raman spectroscopy ·
spatially-resolved spectroscopy · X-ray microscopy · X-ray
tomography
[46]
work.
The exact same catalyst particle was further investigated at the
hard X-ray microprobe endstation of beamline P06 at PETRA III
[
[
1] M. B. Gawande, R. K. Pandey, R. V. Jayaram, Catal. Sci. Technol. 2012, 2,
1113–1125.
2] J. Haber, in Handbook of Heterogeneous Catalysis (Eds.: G. Ertl, H.
Knözinger, F. Schüth, J. Weitkamp), Wiley-VCH Verlag GmbH & Co.
KGaA, Weinheim, Germany, 2008, pp. 3359–3384.
(
2
DESY, Hamburg, Germany). X-rays with an incident energy of
1 keV were focused to 650×500 nm (h×v) beam size using KB
mirrors. Diffraction patterns were recorded at 50 ms exposure time
using an Eiger X 4 M detector (Dectris; 2070×2167 pixels, pixel size
of 75 μm) while raster scanning the sample with a horizontal step
size of 0.4 μm. The sample was scanned in a translate-rotate
[3] A. Trunschke, in Nanostructured Catalysts: Selective Oxidations (Eds.: C.
Hess, R. Schlögl), Royal Society of Chemistry, London, UK, 2011, pp. 56–
95.
[
4] A. Trunschke, J. Noack, S. Trojanov, F. Girgsdies, T. Lunkenbein, V.
measurement scheme over 138 horizontal points (field of view
2
Pfeifer, M. Hävecker, P. Kube, C. Sprung, F. Rosowski, ACS Catal. 2017, 7,
5
5.2×55.2 μm ) and 180° rotation range (approx. 0.5° step size).
3
061–3071.
The sample was then translated 0.5 μm vertically and the acquis-
ition procedure repeated to record a 3D volume consisting of
approximately 50 slices. The calibration of the Debye-Scherrer rings
[
5] K. Weissermel, H.-J. Arpe, in Industrial Organic Chemistry, Wiley-VCH
Verlag GmbH & Co. KGaA, Weinheim, Germany, 2008, pp. 285–294.
[6] P. Sprenger, T. Sheppard, J.-P. Suuronen, A. Gaur, F. Benzi, J.-D.
Grunwaldt, Catalysts 2018, 8, 356–376.
[7] I.E. Wachs, K. Routray, ACS Catal. 2012, 2, 1235–1246.
[8] M. Wolfs, P. Batist, J. Catal. 1974, 32, 25–36.
[9] Y. Moro-Oka, W. Ueda, Adv. Catal. 1994, 40, 233–273.
[10] P. Sprenger, W. Kleist, J.-D. Grunwaldt, ACS Catal. 2017, 7, 5628–5642.
was performed using LaB as a standard. Azimuthal integration of
6
each sample point in the measured tomograms was performed
[47]
using the pyFAI package
resulting in 660 individual 2θ bins.
Tomographic reconstruction was performed using maximum like-
[
48]
lihood expectation maximisation (MLEM) algorithm. Slice-by-slice
analysis and extraction of regions of interest from the tomography
volumes based on unique reflections was performed using
[
11] J. Millet, H. Ponceblanc, G. Coudurier, J. Herrmann, J. Vedrine, J. Catal.
993, 142, 381–391.
1
[12] H. Ponceblanc, J. Millet, G. Coudurier, O. Legendre, J. Vedrine, J. Phys.
Chem. 1992, 96, 9462–9465.
[13] J. F. Brazdil, Catal. Sci. Technol. 2015, 5, 3452–3458.
[14] W. Ueda, Y. Moro-Oka, T. Ikawa, I. Matsuura, Chem. Lett. 1982, 11, 1365–
1368.
[49]
DAWN. Image normalisation, labelling, colour-mapping and 2D
[
50]
visualisation was performed using Fiji. The Avizo 9.3 (Thermo
Fisher Scientific) software package was used for 3D data visual-
isation. For further information on XRD data treatment and
reference materials see ESI.
[
[
15] Y. Moro-Oka, W. Ueda, K.-H. Lee, J. Mol. Catal. A 2003, 199, 139–148.
16] P. Beato, A. Blume, F. Girgsdies, R. E. Jentoft, R. Schlögl, O. Timpe, A.
Trunschke, G. Weinberg, Q. Basher, F. A. Hamid, Appl. Catal. A 2006, 307,
137–147.
[
17] K. V. Raun, L. F. Lundegaard, P. Beato, C. C. Appel, K. Nielsen, M.
Thorhauge, M. Schumann, A. D. Jensen, J.-D. Grunwaldt, M. Høj, Catal.
Lett. 2019, 1–11.
Acknowledgements
[18] P. Sprenger, J.-D. Grunwaldt, W. Kleist, A. Fischer, Vol. EP3219386 A1
E. P. Office), EU, 2017.
We acknowledge DESY (Hamburg, Germany), a member of the
Helmholtz Association HGF, for the provision of experimental
facilities. Parts of this research (XRD-CT experiments) were carried
out at beamline P06 (PETRA III). We acknowledge the European
Synchrotron Radiation Facility (ESRF, Grenoble, France) for
provision of synchrotron radiation facilities and we would like to
thank Remi Tucoulou for assistance in using beamline ID16B for
holotomography, STXM-CT and XRF-CT measurements. We ac-
knowledge SOLEIL for provision of synchrotron radiation facilities
and we would like to thank Stéphanie Belin and Valérie Briois for
assistance in using beamline ROCK (proposal number 20151321).
The work at ROCK was supported by a public grant overseen by
the French National Research Agency (ANR) as part of the
(
[19] D. Linke, D. Wolf, M. Baerns, O. Timpe, R. Schlögl, S. Zeyβ, U.
Dingerdissen, J. Catal. 2002, 205, 16–31.
[20] K. V. Raun, L. F. Lundegaard, J. Chevallier, P. Beato, C. C. Appel, K.
Nielsen, M. Thorhauge, A. D. Jensen, M. Høj, Catal. Sci. Technol. 2018, 8,
4626–4637.
[
21] A. Gaur, M. Schumann, K. V. Raun, M. Stehle, P. Beato, A. D. Jensen, J. D.
Grunwaldt, M. Høj, ChemCatChem 2019, 11, 4871–4883.
[
22] J. Strunk, M. A. Bañares, I.E. Wachs, Top. Catal. 2017, 60, 1577–1617.
[23] H. Tian, I.E. Wachs, L. E. Briand, J. Phys. Chem. B 2005, 109, 23491–23499.
[
24] D. Matras, S. D. Jacques, S. Poulston, N. Grosjean, C. Estruch Bosch, B.
Rollins, J. Wright, M. Di Michiel, A. Vamvakeros, R. J. Cernik, J. Phys.
Chem. C 2019, 123, 1751–1760.
[25] A. Vamvakeros, S. D. Jacques, M. Di Michiel, P. Senecal, V. Middelkoop,
R. J. Cernik, A. M. Beale, J. Appl. Crystallogr. 2016, 49, 485–496.
[26] S. D. Jacques, M. Di Michiel, A. M. Beale, T. Sochi, M. G. O’Brien, L.
Espinosa-Alonso, B. M. Weckhuysen, P. Barnes, Angew. Chem. Int. Ed.
2011, 50, 10148–10152; Angew. Chem. 2011, 123, 10330–10334.
“Investissements d’Avenir” program (reference: ANR10-EQPX45).
ChemCatChem 2021, 13, 1–12
www.chemcatchem.org
10
© 2021 The Authors. ChemCatChem published by Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cctc.202100054
ChemCatChem
[
27] A. Vamvakeros, S. Jacques, M. Di Michiel, D. Matras, V. Middelkoop, I.
[41] O. Attila, H. E. King, F. Meirer, B. M. Weckhuysen, Chem. Eur. J. 2019, 25,
7158–7167.
[42] A. M. Beale, S. D. Jacques, E. K. Gibson, M. Di Michiel, Coord. Chem. Rev.
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
Ismagilov, E. Matus, V. Kuznetsov, J. Drnec, P. Senecal, Nat. Commun.
2018, 9, 1–11.
[
28] T. Li, T. M. Heenan, M. F. Rabuni, B. Wang, N. M. Farandos, G. H. Kelsall,
D. Matras, C. Tan, X. Lu, S. D. Jacques, Nat. Commun. 2019, 10, 1–11.
29] J. Brazdil, M. Mehicic, L. Glaeser, M. Hazle, R. Grasselli, in Catalyst
Characterization Science (Eds.: M. L. Deviey, J. L. Gland), ACS Publica-
tions, 1985, pp. 26–36.
2014, 277, 208–223.
[43] J. Becher, D. F. Sanchez, D. E. Doronkin, D. Zengel, D. M. Meira, S.
Pascarelli, J.-D. Grunwaldt, T. L. Sheppard, Nat. Catal. 2021, 4, 46–53.
[44] P. Sprenger, M. Stehle, A. Gaur, A. M. Gänzler, D. Gashnikova, W. Kleist,
J.-D. Grunwaldt, ACS Catal. 2018, 8, 6462–6475.
[45] B. Ravel, M. Newville, J. Synchrotron Radiat. 2005, 12, 537–541.
[46] P. Sprenger, T. Sheppard, J.-P. Suuronen, A. Gaur, F. Benzi, J.-D.
Grunwaldt, Catalysts 2018, 8, 356.
[47] J. Kieffer, J. Wright, Powder Diffr. 2013, 28, S339-S350.
[48] P. P. Bruyant, J. Nucl. Med. 2002, 43, 1343–1358.
[49] M. Basham, J. Filik, M. T. Wharmby, P. C. Chang, B. El Kassaby, M.
Gerring, J. Aishima, K. Levik, B. C. Pulford, I. Sikharulidze, J. Synchrotron
Radiat. 2015, 22, 853–858.
[50] J. Schindelin, I. Arganda-Carreras, E. Frise, V. Kaynig, M. Longair, T.
Pietzsch, S. Preibisch, C. Rueden, S. Saalfeld, B. Schmid, Nat. Methods
2012, 9, 676–682.
[
[
[
30] B. Grzybowska-Świerkosz, Top. Catal. 2002, 21, 35–46.
31] O. Udalova, D. Shashkin, M. Shibanova, O. Krylov, Kinet. Catal. 2005, 46,
535–544.
[
[
[
[
32] N. Kumar, B. Stephanidis, R. Zenobi, A. J. Wain, D. Roy, Nanoscale 2015,
7, 7133–7137.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
33] T. Hartman, C. S. Wondergem, N. Kumar, A. van den Berg, B. M.
Weckhuysen, J. Phys. Chem. Lett. 2016, 7, 1570–1584.
34] K. Schuh, W. Kleist, M. Høj, V. Trouillet, P. Beato, A. D. Jensen, G. R.
Patzke, J.-D. Grunwaldt, Appl. Catal. A 2014, 482, 145–156.
35] R. K. Grasselli, Catal. Today 2014, 238, 10–27.
[36] H. Ponceblanc, J. Millet, G. Coudurier, J. Védrine, in Catalytic Selective
Oxidation (Eds.: S. T. Oyama, J. W. Hightower), ACS Publications,
Washington, DC, 1993, pp. 262–272.
[
[
[
[
37] S. D. Jacques, M. Di Michiel, S. A. Kimber, X. Yang, R. J. Cernik, A. M.
Beale, S. J. Billinge, Nat. Commun. 2013, 4, 1–7.
38] S. Böcklein, G. Mestl, S. V. Auras, J. Wintterlin, Top. Catal. 2017, 60,
1682–1697.
Manuscript received: January 11, 2021
Revised manuscript received: March 1, 2021
Accepted manuscript online: March 1, 2021
Version of record online: ■■■, ■■■■
39] J. Engeldinger, J. Radnik, C. Kreyenschulte, F. Devred, E. M. Gaigneaux,
A. Fischer, H. W. Zanthoff, U. Bentrup, ChemCatChem 2016, 8, 976–983.
40] M. A. Bañares, Catal. Today 2005, 100, 71–77.
ChemCatChem 2021, 13, 1–12
www.chemcatchem.org
11
© 2021 The Authors. ChemCatChem published by Wiley-VCH GmbH
��
These are not the final page numbers!

1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
FULL PAPERS
See the bigger picture: Mixed metal
oxide catalysts often have a highly
complex structure. Spatially-resolved
characterisation tools applied in 2D
and 3D can help to define the
presence and interaction of different
metal oxide phases within a catalyst
particle. In turn this can reveal
important information on the
Dr. P. Sprenger, M. Stehle, Dr. A. Gaur,
Dr. J. Weiß, D. Brueckner, Dr. Y. Zhang,
Dr. J. Garrevoet, Dr. J.-P. Suuronen,
Dr. M. Thomann, Dr. A. Fischer, Prof. J.-
D. Grunwaldt*, Dr. T. L. Sheppard*
1
– 12
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
Chemical Imaging of Mixed Metal
Oxide Catalysts for Propylene
Oxidation: From Model Binary
Systems to Complex Multicompo-
nent Systems
structure and function of mixed
metal oxide catalysts.