Study of the local structure and oxidation state of iron in complex oxide catalysts for propylene ammoxidation

Article abstract of DOI:10.1039/c4cy00197d

Iron molybdate plays a crucial role in the complex oxide catalysts used for selective oxidation and ammoxidation of hydrocarbons but its structural and electronic properties and their changes in the process of the reaction are poorly understood. A combination of Raman, X-ray absorption, and UV-visible spectroscopy was applied to investigate a commercial catalyst as a function of the reaction time. The results show that an iron-containing compound exists predominantly as ferric molybdate in the fresh catalyst, which is reduced progressively in the process of reaction, forming predominantly ferrous molybdate. The irreversible transformation from Fe₂(MoO ₄)₃ to FeMoO₄ was accompanied by formation of a small amount of Fe₂O₃. These two processes observed in our experiment shed light on the deactivation mechanism of this complex catalyst because they have a negative effect on the selectivity and activity. Specifically, they are responsible for the deterioration of the redox couple, blocking the transmission of lattice oxygen, and irreversibly changing the catalyst structure. Based on the results of the combined techniques, a refined procedure has been proposed to develop a more stable and efficient selective oxidation catalyst.

Full text of DOI:10.1039/c4cy00197d

Catalysis
Science &
Technology
View Article Online
View Journal | View Issue
PAPER
Study of the local structure and oxidation state of
iron in complex oxide catalysts for propylene
ammoxidation
Cite this: Catal. Sci. Technol., 2014,
4, 2512
a
a
b
Li-bin Wu,ab Liang-hua Wu, Wei-min Yang and Anatoly I. Frenkel
*
*
Iron molybdate plays a crucial role in the complex oxide catalysts used for selective oxidation and
ammoxidation of hydrocarbons but its structural and electronic properties and their changes in the
process of the reaction are poorly understood. A combination of Raman, X-ray absorption, and
UV-visible spectroscopy was applied to investigate a commercial catalyst as a function of the reaction
time. The results show that an iron-containing compound exists predominantly as ferric molybdate in
the fresh catalyst, which is reduced progressively in the process of reaction, forming predominantly
ferrous molybdate. The irreversible transformation from Fe2(MoO4)3 to FeMoO4 was accompanied by
formation of a small amount of Fe2O3. These two processes observed in our experiment shed light
on the deactivation mechanism of this complex catalyst because they have a negative effect on the
selectivity and activity. Specifically, they are responsible for the deterioration of the redox couple,
blocking the transmission of lattice oxygen, and irreversibly changing the catalyst structure. Based on the
results of the combined techniques, a refined procedure has been proposed to develop a more stable
and efficient selective oxidation catalyst.
Received 14th February 2014,
Accepted 18th March 2014
DOI: 10.1039/c4cy00197d
www.rsc.org/catalysis
the formation of Bi3FeMo2O12 in another Fe–Mo–Bi type system.
Wolfs and Matsuura concluded that the maximum activity
1. Introduction
Selective catalytic oxidation and ammoxidation of hydrocar-
bons are used in approximately one quarter of the most
important industrial chemicals and basic intermediates pro-
duced by all catalytic processes worldwide.1 A molecular-level
understanding of catalytic behavior in these processes is
important for advancing the general knowledge of catalytic
mechanisms and forming a basis for rational design of new
catalysts and processes. A major breakthrough in the devel-
opment of catalysts for oxidation and ammoxidation of ole-
fins was the discovery of the promoting action of iron in the
bismuth phosphomolybdate catalyst.2,3 Although a number
of other elements such as Ni, Co, Cr, Mn, and K were later
introduced to form the most efficient multicomponent
molybdate (MCM) catalyst,4 iron is still an important pro-
moter in both molybdenum- and antimony-based catalysts.5
Different chemical forms of iron in molybdate catalysts and
its role as a promoter have become subjects of intense
research.6 Early studies by Annenkova et al.7 revealed that
Bi2(MoO4)3, Fe2(MoO4)3, and Bi2Fe4O9 are the main compo-
nents of a Fe–Mo–Bi ternary system. Batist et al.8 reported
and selectivity of the Mg11−xFexMo12BiOn (0 ≤ x ≤ 4) catalyst
is displayed when x = 2.5 but did not explain the significance
of this optimum iron concentration.9 Van Oeffelen arrived at
the conclusion that the role of iron in the same system is to
maintain Bi in the oxidized state by functioning as a redox
couple and to maximize activity and selectivity at x = 2.5.10
Consequently, the equation Bi0+ + 3Fe3+ → Bi3+ + 3Fe2+ was
proposed. The role of iron as a Fe3+/Fe2+ redox couple was
again discussed by Batist.11 Apart from functioning as a redox
couple, iron could also be involved in the formation of other
important compounds (as in the case of bismuth molybdate)
which can display good activity, selectivity and stability by
stabilizing the structure of catalysts.12
The time-dependent changes in the catalyst's composition
offer a possible explanation for the deterioration of its cata-
lytic activity. Deactivation of the MCM catalyst was attributed
to the structural transformation of iron molybdate due to the
loss of MoO3 by volatilization.13 MoO3 is mainly formed by
reduction of Fe2(MoO4)3 in the redox catalytic process according
to the equation: Fe2(MoO4)3 → 2FeMoO4 + MoO3 + [O]lattice
,
evidenced by the presence of a mixture of Fe2(MoO4)3
and Fe2O3 in similar catalysts.14 The iron oxide formed
by oxidation of ferrous molybdate promotes propylene deep
oxidation, deterioration of catalyst activity, target product
degradation and is responsible for the reddish-brown
a Shanghai Research Institute of Petrochemical Technology, 1658 Pudong Beilu,
Shanghai 201208, China. E-mail: yangwm.sshy@sinopec.com
b Physics Department, Yeshiva University, New York, NY 10016, USA.
E-mail: anatoly.frenkel@yu.edu
2512 | Catal. Sci. Technol., 2014, 4, 2512–2519
This journal is © The Royal Society of Chemistry 2014

View Article Online
Catalysis Science & Technology
Paper
color of the spent MCM catalysts. Additionally, the increasing
loss of the reversible Fe3+/Fe2+ redox couple or Fe2(MoO4)3
structure collapse can also lead to other adverse conse-
quences, such as the increasing pressure drop in the catalytic
bed, degradation of the catalyst mechanical resistance, and
the decreasing residual activity of MCM.15
as Fe–Mo–Bi) from the Shanghai Research Institute of Petro-
chemical Technology (SRIPT) which shows high activity and
acrylonitrile selectivity at 380–450 °C. Xm stands for other ele-
ments such as Cr, Co, Ni, Mg, Mn and K. The catalyst was
synthesized using a co-precipitation method; it was then
spray dried and calcined in a rotating furnace at a tempera-
ture of ~600 °C. Catalytic ammoxidation processes were
studied in a Commercial Fluidized Bed Reactor (CFBR) at the
SRIPT. After 5 to 10 days of use for the ammoxidation reac-
tion in the CFBR, the catalytic activity reached a stable state.
Then the catalyst was studied at the accelerating activity test
facilities that employ a Laboratory-scale Fluidized Bed Reac-
tor (LFBR). Using the LFBR, the samples were studied at
different times of reaction: fresh (unused), 5 days, 8 days,
11 days, 16 days and 48 days from the beginning of the reac-
tion. The corresponding samples were denoted as S0, S5, S8,
S11, S16 and S48, respectively.
It is evident from the prior results that the main challenges
to a better understanding of the role of iron in this important
class of catalysts are the heterogeneity of the chemical states
of iron, the coordination environments around iron atoms
and their changes during the reaction. This complexity pre-
sents significant challenges for their structural and chemical
analyses due to the ensemble-average nature of most char-
acterization techniques. In this work, we report a combined
use of X-ray absorption fine structure (XAFS) spectroscopy,
Raman spectroscopy and diffuse reflectance UV-visible light
(DR-UV-vis) spectroscopy for determining the dominant
chemical states and functional forms of Fe and their changes
during the reaction process in a representative commercial
catalyst. The methods used in our work were found to be useful
for multi-technique studies of complex catalysts.16,17 XAFS is
known for its excellent sensitivity to chemical and coordina-
tion environments of iron complexes,18 and it is used exten-
sively for analysis of the degradation of MCM catalysts. Similar
to other ensemble-average techniques such as XRD, XAFS is
not sensitive to minority species because it is a volume-
average method. Raman spectroscopy, on the other hand, is
capable of detecting such species due to various selection
rules.19 For example, small amounts of MoO3 mixed into a fer-
ric molybdate-rich phase may not be detectable by XAFS or
XRD, but would be detectable by Raman spectroscopy which is
used extensively for studying the structure of molybdates.20 In
addition, the fact that the reduced form of iron molybdate, iron
oxide, has no significant contribution to the Raman spec-
trum due to its weak Raman scattering and is sensitive to
DR-UV-vis spectroscopy should also be taken into consideration.
Another important aspect of catalysis investigations that
this work helps to resolve is the issue of heterogeneity of dif-
ferent chemical forms of the same element in the sample.
Ensemble average techniques such as XAFS, when used
alone, cannot discriminate between the different models: 1)
changes in the chemical states of Fe occur uniformly
throughout the entire sample or 2) the sample has a mixture
of the same two or more states of Fe at all times, and the
volume fraction of each state changes with time.
2.2 Raman spectroscopy
The metal oxide phase present in the MCM catalyst samples
was examined using a Jobin-Yvon LabRam 1B Raman spec-
trometer. Before measurement the spectrometer was cali-
brated using a silicon wafer to a wavelength accuracy of
1 cm−1. The Raman spectra of the fresh and spent catalysts
at different reaction times were then collected under ambient
conditions using a 632.8 nm excitation line of an He–Ne laser
source, equipped with
a confocal Olympus microscope
(BX-30). The laser power was kept below 0.5 mW so as to
minimize any laser-induced alterations of the sample.
2.3 Diffuse reflectance UV-vis spectroscopy
Diffuse reflectance UV-visible spectra of selected samples
were obtained (from 12 500 to 50 000 cm−1) using a Perkin
Elmer 555 double beam spectrophotometer at the SRIPT.
BaSO4 was used as the reference and the slit width was set
to 2.0 nm.
2.4 X-ray absorption spectroscopy
Fe K-edge X-ray absorption spectroscopy (XAS) data were col-
lected in transmission mode at the beamline X-19A of the
National Synchrotron Light Source in Brookhaven National
Laboratory in New York, USA. A double crystal Si (111) mono-
chromator was detuned by 30% to minimize the effect of har-
monics. Gas-filled ionization chamber detectors were used
for measuring incident and transmitted beam intensities. In
addition, a third ionization chamber was used to detect the
beam through a reference Fe foil for energy calibration and
data alignment purposes. The XAS specimens were made by
depositing the catalyst powders onto adhesive tapes and fold-
ing the tape several times for homogeneity. The edge steps of
the X-ray absorption coefficient at the Fe K-edge energy var-
ied between 0.3 and 0.4 for all samples.
Our work shows that correlating the results of these tech-
niques is required in order to resolve this challenge, and we
propose a specific model of chemical and structural transfor-
mation in iron species and shed light on the mechanism of
the catalyst deactivation.
2. Experimental
2.1 Catalyst preparation and catalytic tests
Initial data processing was performed by the Athena23
software from the IFEFFIT data analysis package. Several con-
secutive measurements of the same sample were aligned and
The catalyst in this study is a member of the family
Bi0.5–1Fe2–3Mo12–14XmO40–50 (ref. 21 and 22) (denoted hereafter
This journal is © The Royal Society of Chemistry 2014
Catal. Sci. Technol., 2014, 4, 2512–2519 | 2513

View Article Online
Paper
Catalysis Science & Technology
averaged to minimize statistical noise in the data. To directly
compare X-ray Absorption Near Edge Structure (XANES) data
of different samples, the same procedure of pre-edge line
fitting, post-edge curve fitting, and edge-step normalization
was applied to all samples. Quantitative data analysis was
done using the PCA software.24,25
the catalyst because other Fe phases that are not Raman-
active may also be present.
As demonstrated in Fig. 1, Fe2(MoO4)3 shows abundant
load in S0, then its amount decreases monotonically in S5
through S48. For semi-quantitative purposes, the intensity
ratio was calculated between the strongest bands, from
MMoO4 (M = Co, Ni, Mn) and Fe2(MoO4)3 appearing at
~950 cm−1 and 783 cm−1, respectively, assuming that the
majority of MMoO4 phases don't change significantly. The
values of 0.56, 0.39, 0.35, 0.28, 0.20 and 0.20 for S0, S5,
S8, S11, S16 and S48, respectively, demonstrate that the
Fe2(MoO4)3 phase has been decomposing during the reaction
process, from the sample with the highest catalytic activity
(S0) in the very beginning to the most deactivated sample
(S48) with the longest service time. We will discuss the
reason for this decomposition in greater detail below.
As for MoO3 discussed above, it has its characteristic band
at 817 cm−1 which is clearly missing from the data in Fig. 1.31
The absence of the MoO3 phase from all samples (S0 through
S48) is quite reasonable due to their prolonged (5 to 10 days)
treatment in the CFBR and high volatility of MoO3 during the
reaction process.13 As discussed in the Raman spectroscopy
section, the amount of Fe2(MoO4)3 gradually decreased with
reaction time, but what exactly happened to Fe2(MoO4)3
remains unclear. The next section will address this question.
3. Results
3.1. Raman spectroscopy
The spectra corresponding to different reaction times are
shown in Fig. 1. The spectra feature a symmetric stretching
mode, v1, of the MoO4 tetrahedron at 955 cm−1, an asymmet-
ric stretching mode at 890 cm−1 and/or 835 cm−1, bending
modes in-plane and out-of-plane at 430 and 360 cm−1, and
rotation of the entire tetrahedron at 240 cm−1. These observa-
tions are in good agreement with the literature.26–29 Similar
spectra have been obtained for β-CoMoO4 where v1 = 945 cm−1,
Ni/Mo alloys (v1 = 940 cm−1),26 β-HgMoO4 (v1 = 970 cm−1),
α-MnMoO4 (v1 = 940 cm−1), and NiMoO4 (v1 = 960 cm−1).27
Vibrational modes of MoO4 have been reported at 700–900 cm−1
and its bending modes at 300–400 cm−1.28 The spectra shown
in Fig. 1 also reveal the presence of Fe2(MoO4)3 in our cata-
lyst, which has a characteristic high intensity and well iso-
lated band at 783 cm−1. This is in good agreement with the
reported data for Fe2(MoO4)3 that feature major bands at
960, 780, and 350 cm−1.20,29 Quantitatively, the 783 cm−1 peak
intensity can thus be used for estimating the amount of
Fe2(MoO4)3 in our catalysts. Fig. 1 demonstrates that this
compound exhibits a gradual decrease in its quantity with
reaction time. Therefore, our Raman measurements, in agree-
ment with previous reports,30 indicate that the majority
of iron exists in the form of Fe2(MoO4)3 in fresh catalysts
(as well as in short service time catalysts) and its reduction to
FeMoO4 after long-time use. This result is only a partial pic-
ture of the composition and structure of the Fe–Mo phase in
3.2. X-ray absorption near edge structure
Fig. 2 shows Fe K-edge XANES spectra collected from the
Fe–Mo–Bi catalysts. The spectra of all samples, from S0 to
S48, exhibit a gradual shift towards lower energy with the
increase of the reaction time. By comparing our data with
those previously reported,32,33 we firmly believe that this
trend is consistent with the change in the average oxidation
state of Fe from +3 to +2.
Fig. 2 Fe K-edge XANES of Fe–Mo–Bi catalysts at different reaction
times and reference compounds Fe2(MoO4)3, Li2Fe2(MoO4)3, Fe2O3 and
FeO. Arrows show the presence of quasi-isosbestic points indicating
both the one-step transformation within the catalysts from S0 towards
S48 and their agreement with the two standards, (Fe2(MoO4)3, and
Li2Fe2(MoO4)3).
Fig.
1 Raman spectra of Fe–Mo–Bi catalysts corresponding to
different times after the beginning of the reaction (0, 5, 8, 11, 16,
48 days). The arrow indicates a peak corresponding to the Fe2(MoO4)3
complex.
2514 | Catal. Sci. Technol., 2014, 4, 2512–2519
This journal is © The Royal Society of Chemistry 2014

View Article Online
Catalysis Science & Technology
Paper
Fig. 2 and its inset show dramatic changes in the Fe
K-edge XAS data for the different stage catalysts and refer-
ence compounds, Fe2(MoO4)3, Li2Fe2(MoO4)3, Fe2O3 and FeO.
The catalyst data and the Fe2(MoO4)3, Li2Fe2(MoO4)3, and
Fe2O3 references were obtained in two different experiments
(the reference compounds were measured by Shirakawa et al.32).
In order to directly compare them, all data were first aligned
in absolute energy using the reference data of a Fe foil
(in the case of the catalyst samples) and Fe2O3 (in the case of
Fe2(MoO4)3 and Li2Fe2(MoO4)3) measured simultaneously
with all of the samples. By measuring the XAFS data for the
Fe foil and Fe2O3 in the same transmission experiment we
obtained the relative shifts needed for all data sets to share
the same X-ray energy origin.
A nearly perfect agreement between the starting sample,
S0, and Fe2(MoO4)3 demonstrates that: 1) the charge state of
Fe ions was +3 in the fresh catalyst and 2) its Fe phase was
predominantly Fe2(MoO4)3. The absorption edge positions
are defined as the main absorption peak maxima throughout
this article. For the Fe2(MoO4)3 and S0 data, the Fe K-edge
positions are at 7131.8 eV and 7131.5 eV, respectively. Upon
catalytic reaction for some period of time, the spectra shifted
to lower energies towards the reduced form of iron molybdate,
FeMoO4, with the charge state of Fe equal to +2. Instead of
measuring FeMoO4, we compared our data with another Fe+2
compound of Li2Fe2(MoO4)3 (ref. 34) which has the same
local structure around Fe as in FeMoO4.35 As shown in Table 1,
the distributions of the first shell Fe–O distances and the
second shell Fe–Mo distances around Fe as well as the coordi-
nation environment in these two compounds are very similar.
Hence, the use of Li2Fe2(MoO4)3 for comparing with the
experimental data for the Fe–Mo–Bi catalysts is justified.
In the spectra of the last sample, S48, its Fe K-edge energy
(7129 eV) is higher than that of Li2Fe2(MoO4)3 (7127 eV), as
shown in Fig. 2. Hence, the average charge state of Fe in the
last sample is not equal to but approaching the value of +2.
As evidenced by the presence of quasi-isosbestic points in
Fig. 2, Fe ions in all samples undergo transformation from
ferric molybdate to ferrous molybdate other than iron oxide,
FeO, since its spectroscopic features are different from both
S0 and S48.
it does not offer a model-independent determination of the
number of independent species. Principal Component Analy-
sis (PCA) is a superior method for that purpose because it
allows to 1) find the number of independent species mixed
together in the sample at all reaction times and 2) obtain
their unique identities.25,36
Fig. 3(a) shows the “scree test”, demonstrating that the
number of principal components required to reproduce all 6
experimental spectra is equal to 2, as evidenced by the negli-
gible eigenvalue (0.005) there. The standard compounds of
Fe2(MoO4)3 and Li2Fe2(MoO4)3 were well reproduced by the
combination of the two principal components and the target
transform was performed from the basis of the abstract com-
ponents to the basis corresponding to the two standards. The
mixing fraction of the Fe+3 and Fe+2 states was obtained
using a linear combination fit using the PCA software
package.37 Data reproduction of the experimental spectra and
the two standards is shown in Fig. 3(b) as an example.
We have also tested a three-species model against our
experimental data, using Fe2O3 as a possible standard, in
One possible method of quantitative analysis of these
XANES spectra is by linear combination analysis. In this
method, data are represented as a linear combination of two
or more standard compounds and their mixing fractions cor-
respond to the volume fractions of the corresponding iron
species in the sample. The problem with this method is that
Table
1 Relevant structure parameters of FeMoO4 (ref. 35) and
Li2Fe2(MoO4)3 (ref. 34)
Fe environment
FeMoO4 Li2Fe2(MoO4)3 Difference
Fig. 3 (a) The “scree test” shows the change trend of the eigenvalues
obtained by principal component analysis. The change in slope after
the two leading eigenvalues indicates that the number of independent
species in the sample is equal to 2. (b) The representative linear
combination fit of sample S8 using the standard trends of Fe2(MoO4)3
and Li2Fe2(MoO4)3. Their mixing fraction was the only fitting variable.
Charge valence
+2
+2
0
Fe–O coordination number
Fe–Mo coordination number
Average Fe–O distance (Å)
Average Fe–Mo distance (Å)
6
6
0
6
6
0
2.14
3.61
2.12
3.67
<1%
<2%
This journal is © The Royal Society of Chemistry 2014
Catal. Sci. Technol., 2014, 4, 2512–2519 | 2515

View Article Online
Paper
Catalysis Science & Technology
addition to the previous two-species standards as described
above. The best fit values of the mixing fractions, correspond-
ing to species after the target transform procedure, had large
negative values, which is non-physical. Hence, our XANES
data gave no evidence for the formation of Fe2O3 within the
accuracy of the PCA method used.
specifically Fe2(MoO4)3
ammoxidation process. This transformation will be discussed
in greater detail below.
→
FeMoO4 during the selective
3.3. UV-vis diffuse reflectance spectroscopy
Guided by the PCA results, we treated the changes of the
Fe oxidation state in the catalyst during the reaction as a
one-step transformation from Fe+3 to Fe+2 and summarized
the quantitative information about the rate of this transfor-
mation in Fig. 4. The best fit results for samples 0, 5, 8, 11,
16 and 48 show a gradual and monotonic decrease of the
fraction of Fe2(MoO4)3 from 99% to 37%. According to these
results, the average Fe charge states in all samples vary from
+3 for S0 to +2.37 for S48. This result demonstrates that the
average Fe oxidation state in all spent catalysts, even for the
longest reaction time, is still much higher than +2 which is
the oxidation state of Fe in FeMoO4.
Fig. 5 shows the DR-UV-vis spectra of selected samples of
Fe–Mo–Bi catalysts at different reaction times. All of the spec-
tra exhibit a characteristic absorption band of Fe2(MoO4)3
at ~460 nm.12,38 The broad absorption in the UV region is
ascribed to the presence of both tetrahedral and octahedral
oxomolybdate groups,39 whereas the one in the visual region
is attributed to Fe2O3.40 This iron oxide becomes increasingly
abundant, accompanied by the catalyst sample showing a red-
brown color characteristic of fresh Fe2O3,41 with longer reac-
tion times according to either or both of these mechanisms:
Due to the ensemble-averaging nature of XAFS, there are
two models that can be used to interpret the PCA results: 1)
“homogeneous transformation”, in which every Fe atom
changes its charge state from +3 to +2.37 (sample S0 to sam-
ple S48) and 2) “heterogeneous transformation”, where Fe
atoms are divided into two groups at all times, with charge
states of +3 and +2, with volume fractions x and 1 − x, respec-
tively. The volume fraction of Fe3+ changes from 99% to 37%
for samples S0 to S48, respectively. Both models gave identi-
cal XANES trends and identical results based on linear com-
bination analysis. Hence, within the results of just one
technique, XANES, it is impossible to differentiate between
the two models. However, Raman spectra show that Fe3+, in
particular Fe2(MoO4)3, remains in all of the samples through
different reaction stages, not only in the fresh catalyst, S0,
but also in the last one, S48. That observation is consistent
with model 2) and inconsistent with model 1). Thus, the
combination of XAFS and Raman measurements is required
to validate a hypothesis of partial Fe2(MoO4)3 transformation,
3FeMoO4 + 3/4O2 → Fe2(MoO4)3 + 1/2Fe2O3
2FeMoO4 + 1/2O2 → Fe2O3 + 2MoO3
(1)
(2)
Taking into account the different crystal structures of α-
and β-FeMoO4 phases (Mo6+ coordination is octahedral in α
and tetrahedral in β), it is reasonable to conclude that mech-
anism (1) occurs preferentially for the β phase and (2) for the
α phase, as Mo6+ coordination is tetrahedral in Fe2(MoO4)3
but octahedral in MoO3. Given that no MoO3 was detected in
the Raman spectra from S0 through S48, one can further con-
clude that eqn (1) is the main reoxidation path during the
entire catalytic process. As a result, eqn (1) will lead to more
stable catalytic activities than eqn (2) due to the regeneration
of Fe2(MoO4)3. As for the increasing accumulation of Fe2O3,
as evidenced by the UV-Vis spectra (Fig. 5), its absence in the
Raman spectra and in the Fe K-edge XANES spectra indicates
that it is present in the sample as a minority Fe species
whose volume fraction does not exceed ca. 5%, which is the
Fig. 4 Fractions of Fe2(MoO4)3 in samples S0 through S48 obtained
using two different characterization methods, Raman and XANES,
exhibit similar trends. The two measurements indicate that the amount
of Fe2(MoO4)3 has decreased with reaction time.
Fig. 5 Diffuse reflectance UV-visible spectra of selected Fe–Mo–Bi
catalyst samples. The data were normalized by the absorption maxi-
mum. The increasing intensity in the region between 400 and 700 nm
shows the increasing amount of Fe2O3 with reaction time.
2516 | Catal. Sci. Technol., 2014, 4, 2512–2519
This journal is © The Royal Society of Chemistry 2014

View Article Online
Catalysis Science & Technology
Paper
uncertainty in the Fe speciation by principal component anal-
ysis of XANES spectra.
process, and the balance should be fully reversible. In that
case, the catalytic activity of the spent catalyst, e.g., conver-
sion of propylene, will be same as that of the fresh one. How-
ever, due to the loss of MoO3 via volatilization and the much
easier sublimation of MoO2(OH)2 after combination with
resulting H2O (eqn (3)),50 eqn (4) loses its reversibility. There-
fore, ferrous molybdate can only partially get oxidized and
restored back to Fe2(MoO4)3 incompletely, resulting in an
increased amount of the resultant FeMoO4. We propose that
eqn (1) is the likely oxidation route because no MoO3 is
detected in the Raman spectra (Fig. 1) and it is accompanied
by a small amount of Fe2O3 which was detected by DR-UV-Vis
spectroscopy (Fig. 5). The presence of a minute amount of
iron oxide is not in contradiction with our XAS results
because the uncertainties in the LCA-derived error bars on
mixing fractions allow for 5% of another iron species in the
sample (vide supra), and UV-vis is highly sensitive to Fe2O3.
Hence, the role of the UV-vis measurement is very important
here because, without it, the presence of Fe2O3 could not
have been detected.
4. Discussion
As well documented elsewhere,42,43 propylene ammoxidation
obeys the Mars–van Krevelen mechanism and is a six-
electron redox process.44 Iron molybdate, both ferric and fer-
rous forms, is a good example of an efficient redox couple. It
can promote air–dioxygen dissociation on Fe2+ into lattice
oxygen, [O]L, and its transfer to the active site using Fe3+.45
The latter, in turn, will reoxidize the resulting reduced Bi and
Mo,10,46 further turning C3H6 into the target product, C3H3N:
Fe2+ + 1/2O2 → Fe3+ + [O] + e
Bi0+(Mo4+) + 3Fe3+ → Bi3+(Mo6+) + 3Fe2+
C3H6 + NH3 + 3[O] → C3H3N + 3H2O
(3-1)
(3-2)
(3-3)
It is evident from these examples that iron molybdate is
an active species throughout the entire catalytic reaction and
plays crucial roles in the ammoxidation process. To achieve
high performance on the Fe–Mo–Bi catalyst, the balance
between Fe3+ and Fe2+ should be kept close to that in the ini-
tial state, which is known to have the highest catalytic activity
and selectivity.47 Our experiment demonstrated (Fig. 1)
that the initial state of iron molybdate, Fe2(MoO4)3, has
decomposed significantly for all catalyst samples from S5
through S48. The rate of decomposition was fastest during
the first 16 days of the catalytic process.
XAS results not only have revealed the time-dependent
decomposition of iron molybdate during the reaction, but
also have revealed that the iron-containing phase in the fresh
sample, S0, is predominantly Fe2(MoO4)3. In addition, the
combined use of Raman spectroscopy and XAS helped us to
propose the deactivation mechanism for Fe2(MoO4)3, which
follows the Fe2(MoO4)3 → FeMoO4 step. Another important
result emerging from this work is that Fe3+ directly trans-
forms to Fe2+ with no intermediate phase. During this trans-
formation, the volume fraction of Fe3+ changes from 100 to
37% for samples S0 to S48, respectively, and the volume frac-
tion of Fe2+ increases accordingly. While speciation of the
chemical states of iron has been made possible by XAS in ear-
lier studies,48,49 it is due to the complementarity of XAS and
Raman experiments that this conclusion can be made in the
present case. Each technique, taken alone, will only show an
incomplete picture.
In conclusion, based on the observations described above,
we find that continuously increasing accumulation of
FeMoO4 and depletion of Fe2(MoO4)3 is inevitable with the
increased reaction time, provided that no additional MoO3 is
added to the reaction mixture to reverse eqn (4) towards the
Fe2(MoO4)3 direction. As a result, the catalytic performance
of the Fe–Mo–Bi catalyst degrades with reaction time.
Two recommendations for rationally designing better
ammoxidation catalysts emerged from our findings. First, it
is important to maintain a sufficient number of Fe3+ sites
in an overall reducing atmosphere (propylene ammoxidation
to acrylonitrile); it is necessary to stabilize the Fe3+ state
structurally and/or functionally. One candidate for such sta-
bilizer is the Cr3+/Cr2+ redox couple. It is generally more
effective at higher temperature than the iron couple;43 in
addition, it acts as a structural diluent to iron and a booster
of the Fe3+ state in the Fe–Mo–Bi matrix. Relevant tests are
presently under way in our group and will be reported
elsewhere. Second, it is important to replenish MoO3 in
the reacting mixture to compensate for its loss through
volatilization or sublimation to slow down the collapse of
ferric molybdate. MoO3 may come from a separate compo-
nent compound or some other compound as suggested
elsewhere.5,6
Based on the discussion stated above, we conclude that
ideal Fe–Mo–Bi catalysts for propylene ammoxidation should
be both functionally and structurally stable. Functional sta-
bility requires that the functional form of iron molybdate
should be kept the same as in the fresh catalyst. Based on
our results and the work of others, a revised feasible model
for a highly active and long-term stable Fe–Mo–Bi catalyst
may be proposed, as illustrated in Fig. 6. In this catalyst
the Fe3+/Fe2+ redox couple is crucial, both to composing
reversible redox processes and stabilizing the main structure
of catalysts by suppressing its deformation, due to active
component decomposition.
We now turn our attention to the possible origins of
decomposition of Fe2(MoO4)3. The following equation is a
plausible description of the process:
Fe2(MoO4)3 ⇋ 2FeMoO4 + MoO3 + 1/2O2
(4)
As an ideal catalyst, iron molybdate could strike a balance
between its ferric and ferrous forms during the catalysis
This journal is © The Royal Society of Chemistry 2014
Catal. Sci. Technol., 2014, 4, 2512–2519 | 2517

View Article Online
Paper
Catalysis Science & Technology
Drs. Yuanyuan Li and Nebojsa Marinkovic for their help with
synchrotron measurements.
Notes and references
1 R. Grasselli, J. Burrington, D. Buttrey, P. DeSanto Jr.,
C. Lugmair, A. Volpe Jr. and T. Weingand, Top. Catal.,
2003, 23, 5–22.
2 R. K. Grasselli, Proceedings of the DGMK Conference, 2001,
Hamburg, Germany, p. 147.
3 T. S. R. P. Rao and K. R. Krishnamurthy, J. Catal., 1985, 95,
209–219.
4 R. K. Grasselli and H. F. Hardman, U. S. Patent 3642930, 1972.
5 R. K. Grasselli, D. D. Suresh and H. F. Hardman, U. S. Patent
4001317, 1977; R. K. Grasselli, D. D. Suresh and A. F. Miller,
U. S. Patent 4167494, 1979.
6 G. W. Keulks, L. D. Krenzke and T. M. Notermann, in
Advances in Catalysis, ed. H. P. D. D. Eley and B. W. Paul,
Academic Press, 1979, vol. 27, pp. 183–225.
7 I. B. Annenkova, T. G. Alkazov and M. S. Belenku, Kinet.
Catal., 1969, 10, 1305–1311.
8 P. A. Batist, C. G. M. van de Moesdijk, I. Matsuura and
G. C. A. Schuit, J. Catal., 1971, 20, 40–57.
Fig. 6 A schematic model of the Fe–Mo–Bi catalyst showing the role
of the Fe3+/Fe2+ redox couple during the propylene catalytic
ammoxidation process. Propylene, C3H6, is converted into the target
compound, C3H3N, on bismuth molybdate through H-abstraction and
lattice O-incorporation, while iron molybdate (including ferric and
ferrous) facilitates dioxygen dissociation and lattice oxygen transfer.
Two highlighted strategies proposed here are the introduction of the
Cr3+/Cr2+ redox couple in the catalyst preparation and replenishment
of MoO3 via a separate component or compound during the reaction
process.
9 M. W. J. Wolfs, Ph.D. thesis, Selective Oxidation of Olefins
over Multicomponent Molybdate Catalysts. Technische Hoge
School, Eindhoven 1974; I. Matsuura and M. W. J. Wolfs,
J. Catal., 1975, 37, 174–178.
5. Summary and conclusions
In our Fe–Mo–Bi catalysts, an iron molybdate phase identi-
fied as Fe2(MoO4)3 has undergone dramatic partial (more
than 60%) decomposition with increasing reaction time. This
decomposition is a one-step transformation from Fe3+ into
another form of iron molybdate, namely FeMoO4, with an
iron charge state of +2. These results were obtained by a com-
bination of Raman spectroscopy and XANES spectroscopy
studies of the fresh and spent catalysts at different reaction
times. DR-UV-vis measurements revealed a small fraction
(less than 5 volume%) of Fe2O3 during the reaction process.
The combination of these measurements allowed us to pro-
pose a mechanism of transformation of iron molybdate and
its role in the propylene ammoxidation process. This work
highlights the critical role of Fe3+ species for the stable and
efficient conversion of propylene to acrylonitrile in selective
ammoxidation. A modification of the Fe–Mo–Bi catalyst that
will help stabilize Fe3+ ions is proposed.
10 P. A. G. Van Oeffelen, Ph.D. thesis, Selective Oxidation of
Olefins on Molybdate Catalysts, Technische Hoge School,
Eindhoven, 1978.
11 P. A. Batist, Surf. Technol., 1979, 9, 443–446.
12 M. W. J. Wolfs and P. H. A. Batist, J. Catal., 1974, 32, 25–36.
13 A. P. V. Soares, M. F. Portela, A. Kiennemann and L. Hilaire,
Chem. Eng. Sci., 2003, 58, 1315–1322.
14 N. Pernicone, Catal. Today, 1991, 11, 85–91.
15 N. Burriesci, F. Garbassi, M. Petrera, G. Petrini and
N. Pernicone, in Studies in Surface Science and Catalysis,
ed. B. Delmon and G. F. Froment, Elsevier, 1980, vol. 6,
pp. 115–126.
16 S. Bordiga, E. Groppo, G. Agostini, J. A. van Bokhoven and
C. Lamberti, Chem. Rev., 2013, 113, 1736–1850; I. E. Wachs
and C. A. Roberts, Chem. Soc. Rev., 2010, 39, 5002–5017.
17 R. A. Schoonheydt, Chem. Soc. Rev., 2010, 39, 5051–5066;
F. Cesano, S. Bertarione, A. Piovano, G. Agostini, M. M. Rahman,
E. Groppo, F. Bonino, D. Scarano, C. Lamberti, S. Bordiga,
L. Montanari, L. Bonoldi, R. Millini and A. Zecchina, Catal.
Sci. Technol., 2011, 1, 123–136.
Acknowledgements
AIF acknowledges support from the Chemical Sciences,
Geosciences, and Biosciences Division, Office of Basic Energy
Sciences, Office of Science, U. S. Department of Energy (grant
no. DE-FG02-03ER15476). Use of the NSLS is supported by
the U.S. Department of Energy, Office of Science, Office of
Basic Energy Sciences under contract no. DE-AC02-98CH10886.
Beamline X19A at the NSLS is supported in part by the
Synchrotron Catalysis Consortium, U.S. Department of
Energy (grant no. DE-FG02-05ER15688). We are grateful to
18 G. Waychunas, M. Apted and G. Brown Jr., Phys. Chem.
Miner., 1983, 10, 1–9.
19 A. Patlolla, P. Baumann, W. Xu, S. D. Senanayake,
J. A. Rodriguez and A. I. Frenkel, Top. Catal., 2013, 56,
896–904.
20 Q. Xu, G. Jia, J. Zhang, Z. Feng and C. Li, J. Phys. Chem. C,
2008, 112, 9387–9393.
21 R. K. Grasselli, Catal. Today, 1999, 49, 141; R. K. Grasselli,
Appl. Catal., 1985, 15, 127–139.
2518 | Catal. Sci. Technol., 2014, 4, 2512–2519
This journal is © The Royal Society of Chemistry 2014

View Article Online
Catalysis Science & Technology
Paper
22 D. D. Suresh, M. S. Friedrich and M. J. Seely, U. S. Patent
5212137, 1993; O. V. Udalova, D. P. Shashkin, M. D. Shibanova
and O. V. Krylov, Kinet. Catal., 2005, 46, 535–544.
23 B. Ravel and M. Newville, J. Synchrotron Radiat., 2005, 12,
537–541.
24 S. R. Wasserman, J. Phys. IV, 1997, 7, C2-203–205.
25 A. I. Frenkel, O. Kleifeld, S. R. Wasserman and I. Sagi,
J. Chem. Phys., 2002, 116, 9449–9456.
37 S. R. Wasserman, P. G. Allen, D. K. Shuh, J. J. Bucher and
N. M. Edelstein, J. Synchrotron Radiat., 1999, 6, 284–286.
38 P. Forzatti, P. L. Villa, N. Ferlazzo and D. Jones, J. Catal.,
1982, 76, 188–207.
39 W. D. Kovats and C. G. Hill, Appl. Spectrosc., 1986, 40,
1215–1223.
40 A. C. Scheinost, A. Chavernas, V. Barron and J. Torrent,
Clays Clay Miner., 1998, 46, 528–536.
26 P. Delichere, A. Hugot-Le Goff and S. Joiret, Surf. Interface
Anal., 1988, 12, 419–423.
41 Z. Zhang, M. F. Hossain and T. Takahashi, Appl. Catal., B,
2010, 95, 423–429.
27 I. Kanesaka, H. Hashiba and I. Matsuura, J. Raman
Spectrosc., 1988, 19, 213–218.
28 H. Tian, I. E. Wachs and L. E. Briand, J. Phys. Chem. B,
2005, 109, 23491–23499; I. Matsuura, R. Schut and K. Hirakawa,
J. Catal., 1980, 63, 152–166.
29 G. Hill Jr and J. H. Wilson III, J. Mol. Catal., 1990, 63, 65–94.
30 X. Ge, J. Shen and H. Zhang, Sci. China, Ser. B: Chem.,
1996, 39, 53–63.
31 M. Dieterle, G. Weinberg and G. Mestl, Phys. Chem. Chem.
Phys., 2002, 4, 812–821.
32 J. Shirakawa, M. Nakayama, M. Wakihara and Y. Uchimoto,
J. Phys. Chem. B, 2007, 111, 1424–1430.
33 A. M. Beale, S. D. M. Jacques, E. Sacaliuc-Parvalescu,
M. G. O'Brien, P. Barnes and B. M. Weckhuysen, Appl.
Catal., A, 2009, 363, 143–152.
34 C. C. Torardi and E. Prince, Mater. Res. Bull., 1986, 21,
719–726.
35 A. W. Sleight, B. L. Chamberland and J. F. Weiher, Inorg.
Chem., 1968, 7, 1093–1098.
42 R. K. Grasselli, J. Chem. Educ., 1986, 63, 216–221.
43 P. Mars and D. W. van Krevelen, Chem. Eng. Sci., 1954, 3,
41–59.
44 R. K. Grasselli, Ammoxidation, in Handbook of Heterogeneous
Catalysis, ed. G. Ertl, J. Weitkamp, and H. Knoezinger,
John Wiley & Son Inc., New York, 1997, vol. 5, p. 2302.
45 R. K. Grasselli, Catal. Today, 1999, 49, 141–153.
46 R. Schlögl, A. Knop-Gericke, M. Hävecker, U. Wild, D. Frickel,
T. Ressler, R. E. Jentoft, J. Wienold, G. Mestl, A. Blume,
O. Timpe and Y. Uchida, Top. Catal., 2001, 15, 219–228.
47 A. A. Firsova, Y. V. Maksimov, V. Y. Bychkov, O. V. Isaev,
I. P. Suzdalev and V. N. Korchak, Kinet. Catal., 2000, 41,
116–121.
48 E. Borfecchia, L. Mino, D. Gianolio, C. Groppo, N. Malaspina,
G. Martinez-Criado, J. A. Sans, S. Poli, D. Castelli and
C. Lamberti, J. Anal. At. Spectrom., 2012, 27, 1725–1733.
49 G. Berlier, G. Spoto, S. Bordiga, G. Ricchiardi, P. Fisicaro,
A. Zecchina, I. Rossetti, E. Selli, L. Forni, E. Giamello and
C. Lamberti, J. Catal., 2002, 208, 64–82.
36 J. Y. Kim, J. A. Rodriguez, J. C. Hanson, A. I. Frenkel and
P. L. Lee, J. Am. Chem. Soc., 2003, 125, 10684–10692.
50 L. Zhang, D. Liu, B. Yang and J. Zhao, Appl. Catal., A,
1994, 117, 163–171.
This journal is © The Royal Society of Chemistry 2014
Catal. Sci. Technol., 2014, 4, 2512–2519 | 2519