Effects of Anion Substitution in (Mo,V)5O14 on Catalytic Performance in Selective Propene Oxidation to Acrolein

Article abstract of DOI:10.1002/cctc.201501076

Molybdenum-based mixed oxides represent well-known model catalysts for the selective oxidation of light alkenes. Here, the anion lattice of (Mo,V)₅O₁₄ was for the first time modified by substituting oxygen ions with nitrogen ions. Investigations by XRD analysis, X-ray absorption spectroscopy, and FTIR spectroscopy revealed that the incorporation of nitrogen in the structure of Mo₅O₁₄ proceeded without changing the average valence of metal centers. Additionally, impedance spectroscopy confirmed the formation of oxygen vacancies. Significant changes in conductivities remained after the removal of nitrogen. Temperature-programmed reduction measurements were performed to investigate oxygen mobility. The enhanced reducibility of oxide nitrides correlated with the increased conductivity. Catalytic performance in selective propene oxidation was determined by online mass spectrometry und gas chromatography at different temperatures. Selectivity towards acrolein increased with increasing conductivity whereas the formation of total oxidation products CO_x decreased.

Full text of DOI:10.1002/cctc.201501076

DOI: 10.1002/cctc.201501076
Full Papers
Effects of Anion Substitution in (Mo,V)5O14 on Catalytic
Performance in Selective Propene Oxidation to Acrolein
S. Kühn, D. Weber, M. Lerch, and T. Ressler*[a]
Molybdenum-based mixed oxides represent well-known model
catalysts for the selective oxidation of light alkenes. Here, the
anion lattice of (Mo,V)5O14 was for the first time modified by
substituting oxygen ions with nitrogen ions. Investigations by
XRD analysis, X-ray absorption spectroscopy, and FTIR spectros-
copy revealed that the incorporation of nitrogen in the struc-
ture of Mo5O14 proceeded without changing the average va-
lence of metal centers. Additionally, impedance spectroscopy
confirmed the formation of oxygen vacancies. Significant
changes in conductivities remained after the removal of nitro-
gen. Temperature-programmed reduction measurements were
performed to investigate oxygen mobility. The enhanced re-
ducibility of oxide nitrides correlated with the increased con-
ductivity. Catalytic performance in selective propene oxidation
was determined by online mass spectrometry und gas chroma-
tography at different temperatures. Selectivity towards acrolein
increased with increasing conductivity whereas the formation
of total oxidation products COx decreased.
Introduction
Molybdenum oxide based catalysts are active for selective oxi-
dation reactions of light alkanes and alkenes, for example, pro-
pene.[1–3] It is assumed that these reactions proceed according
to a redox mechanism (“Mars van Krevelen mechanism”).[4] In
a first step propene is chemisorbed at the surface of the cata-
lyst forming an allylic species. Subsequently, the allylic species
is oxidized by lattice oxygen from the metal oxide catalyst fol-
lowed by reoxidation of the catalyst with gas-phase oxygen.[5]
The reduced chemical and structural complexity of binary
oxides like a-MoO3 makes it the most simple model system for
elucidating structure–activity correlations. Former studies of a-
MoO3 showed that lattice oxygen is involved in selective oxida-
tion of propene to acrolein.[6,7] Further investigations of the
defect structure of a-MoO3 under propene oxidation condi-
tions revealed a slightly decreased average Mo valence, which
might be attributed to the formation of share-structural de-
fects.[8,9] Additionally, after selective oxidation of the reactant
with nucleophilic lattice oxygen the resulting oxygen vacancy
might not be reoxidized immediately at the active site of the
catalyst.[6] Hence, diffusion of oxygen vacancies and mobility of
lattice oxygen is expected to play an important role for catalyt-
ic performance. With respect to suitable model systems for elu-
cidating the effect of oxygen mobility on catalytic performance
it is desirable to preserve the crystallographic structure of the
catalyst. Additionally, the density of vacancies and diffusibility
of oxygen ions has to be varied. Cation substitution is an effec-
tive method for generating oxygen vacancies. Addition of
other transition metals like W, Nb, or V can lead to promoting
effects on activity or selectivity in the selective oxidation reac-
tion of light alkenes.[10–12] However, promoting Mo-oxide based
catalysts with additional metal centers often results in the for-
mation of various crystal structures. Therefore, modification of
the cation lattice to modify oxygen mobility seems to be less
suited.
Recent studies have shown that nitrogen can be incorporat-
ed in the anion lattice of a-MoO3 without changing the crystal-
lographic structure. Resulting oxide nitrides represent suitable
model system for studying correlations between oxygen mobi-
lity and catalytic performance.[13] The corresponding oxide ni-
trides of a-MoO3 exhibited an enhanced electronic conductivi-
ty.[14] In addition, mild ammonolysis resulted in a much im-
proved reducibility, which can be attributed to increased
oxygen mobility. Correlations between catalytic performance
and electrical properties could not be elucidated because of
the low thermal stability of the corresponding oxide nitrides.[13]
However, MoO3 is not the most relevant system for studies on
selective oxidation of propene. Consequently, the range of
oxide nitrides should be extended to more relevant structures
in propene oxidation. Model catalysts with the Mo5O14 struc-
ture constitute a more complex system compared to that of
MoO3. Additionally, structural motives of Mo5O14 were found to
be more relevant for selective oxidation of propene.[15,16] In this
work preparation and structural characterization of (MoV)
oxide nitrides with Mo5O14 structure are presented. Further-
more, the effects of ammonolysis of mixed molybdenum
oxides on oxygen mobility and catalytic performance in selec-
tive oxidation of propene were studied.
[a] S. Kühn, Dr. D. Weber, Prof. M. Lerch, Prof. T. Ressler
Department of Chemistry
Technische Universität Berlin
Strasse des 17. Juni 135, 10623 Berlin (Germany)
E-mail: Thorsten.ressler@tu-berlin.de
ORCID(s) from the author(s) for this article is/are available on the WWW
under http://dx.doi.org/10.1002/cctc.201501076.
ChemCatChem 2016, 8, 758 – 766
758
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Results and Discussion
Ammonolysis of (Mo,V)5O14
(Mo,V)5O14 was ammonolyzed at various temperatures between
498 and 573 K. A surface area ABET of approximately 4 gmÀ2
was determined for all samples by N2 physisorption. The phase
composition of various ammonolysis products was investigat-
ed by XRD technique. The diffraction patterns of the samples
ammonolyzed at 498 K and 523 K showed the same phase
composition as that of the starting material (Mo,V)5O14
(Figure 1). At higher temperatures (Mo,V)5O14 was partially re-
Figure 2. Rietveld refinements of (Mo,V)5O14 crystal structure to XRD powder
patterns of (Mo,V)5O14 and its corresponding oxides nitrides with different ni-
trogen contents. Difference curves are shown below each refinement. The
resulting fit parameters are given in Table 1.
Figure 1. XRD powder patterns of the products obtained by ammonolysis of
(Mo,V)5O14 at different temperatures. Identified phases are marked by sym-
&
^
*
bols ( Mo5O14, Mo9O26, Mo3N2).
Table 1. Analysis of lattice parameter a and c, unit cell volume V from
Rietveld refinement (goodness of fit GOF and Chi2 are listed) and William-
son–Hall analysis (particle size D, lattice strain e) of (Mo,V)5O14 and oxide
nitrides with different nitrogen contents.
duced to Mo9O25 by H2 formed during decomposition of gas-
eous NH3.[20] The ammonolysis product at 573 K revealed the
structure of Mo3N2.[21] Elemental analysis provided information
about the nitrogen content in the prepared samples. The nitro-
gen content could be adjusted by varying the ammonia flow
rate and the reaction time. Hence, oxide nitrides with Mo5O14
structure and 1, 1.3, and 1.5 wt% nitrogen were prepared for
detailed investigations. The Mo5O14-type oxide nitrides with
1 and 1.3 wt% nitrogen were prepared at 498 K with 5 or
10 LhÀ1 ammonia. The temperature was increased to 523 K for
preparing oxide nitride with 1.5 wt% nitrogen adjusting an
ammonia gas flow of 5 LhÀ1. Metal contents were not affected
by ammonolysis.
Oxide
Oxide nitride
1 wt% N
Oxide nitride
1.3 wt% N
Oxide nitride
1.5 wt% N
a [Š]
c [Š]
V [Š3]
e
22.8666(4)
4.0027(9)
2092.9
0.061
22.8969(8)
3.9939(2)
2093.2
0.095
22.9302(8)
3.9950(2)
2100.6
0.148
22.9491(8)
3.9936(2)
2103.3
0.171
D [nm]
GOF
Chi2
(57Æ2)
1.3
1.8
(56Æ2)
1.3
1.6
(54Æ2)
2.1
1.5
(57Æ2)
1.9
1.4
XRD powder pattern of a mixture of 50 wt% of the sample
and 50 wt% of a-Al2O3 as described in the literature.[24] The
a and b parameters of the oxide nitrides were slightly in-
creased compared to those of the oxide whereas no significant
change in c parameter was observed. Consequently, the cell
volume was also increased after ammonolysis. The increase of
cell parameters might be explained by incorporation of nitro-
gen in the anion lattice of (Mo,V)5O14.[25] Logvinovich et al. re-
ported a similar behavior of strontium–molybdenum oxide ni-
trides owing to the larger effective ionic radius of N3À (1.32 Š)
than that of O2À (1.26 Š).[26]
Structural effects of nitrogen incorporation
A Rietveld refinement of the Mo5O14 crystal structure to XRD
patterns of (Mo,V)5O14 and its corresponding oxide nitrides
with 1, 1.3, and 1.5 wt% nitrogen was performed. A pseudo-
Voigt function convoluted with an axial divergence asymmetry
function was used for Rietveld refinement (Figure 2, ICSD
27202).[22,23] The asymmetry function was necessary to account
for peak asymmetry at low diffraction angels resulting from in-
strumental effects. Results obtained from XRD refinement are
listed in Table 1. The good agreement of the experimental and
theoretical data showed that other crystalline phases were not
detected. The formation of significant amounts of amorphous
phases during ammonolysis was excluded by analyzing the
A line broadening of the X-ray diffraction peaks after ammo-
nolysis was observed. The full width at half maximum (FWHM)
of the most intensive line at 2q=22.48 2q increased from 0.148
2q to 0.238 2q after ammonolysis. The line broadening can be
ChemCatChem 2016, 8, 758 – 766
www.chemcatchem.org
759
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
attributed to a decrease in particle size or an increase in lattice
strain. For a detailed analysis of the line broadening a size–
strain analysis by the Williamson–Hall method was per-
formed.[27] The Williamson–Hall analysis assumes that XRD line
broadening is represented by the sum of the contributions of
particle size and lattice strain. Particle size D and lattice strain
e could be determined by plotting (bhkl·cosq)/l versus sinq ac-
cording to the Williamson–Hall equation:[28]
was refined to experimental data in two steps. One overall E0
shift and respective Debye–Waller factors for all distances were
refined in the first step with average distances kept invariant.
Afterwards, free-ranged averaged distances were added to the
refinement procedure. The final results of all refinements are
shown in Table 2. The significance of the fitted parameters was
Table 2. Type, number (N), and XAFS disorder parameter (s2) of atoms at
a distance R from the Mo atoms in (Mo,V)5O14 and oxide nitrides with dif-
ferent nitrogen contents. Experimental distances and disorder parameters
were obtained from refinement of a model of averaged distances (k
range from 3 to 14 ŠÀ1, R range from 0.9 to 3.96 Š, Nind =23, Nfree =5–11).
Subscript c indicates parameters that were correlated in the refinement
and subscript f indicates fixed parameters.
b
k
D
4e
l
hkl cosq ¼
þ
sinq
ð1Þ
l
Integral breadth of LaB6 660a was used as XRD standard to
correct the measured integral breadth for instrumental effects.
The corrected integral breadth bhkl was estimated by the fol-
lowing expression.[29]
Type
N
Oxide
Oxide nitride
1 wt% N
R [Š] s2 [Š2]
Oxide nitride
1.5 wt% N
R [Š] s2 [Š2]
R [Š] s2 [Š2]
2
2
ð1=2Þ
instrumental
MoÀO “short”
MoÀO “middle”
MoÀO “long”
MoÀMo “short”
1.6f 1.74 0.0012c 1.74 0.0013c 1.74 0.0016c
3.8f 2.00 0.014 2.00 0.014 2.00 0.014
0.7f 2.36 0.0012c 2.36 0.0013c 2.37 0.0016c
1.3f 3.32 0.0058 3.33 0.0055 3.33 0.0057
ð2Þ
bhkl ¼ ððbhkl
Þ
measuredÀðbhkl
Þ
Þ
The Williamson–Hall plots for all samples resulted in
a straight line over a wide angle range (Figure 3). Integral
MoÀMo “middle” 2.3f 3.80 0.0036c 3.80 0.0033c 3.81 0.0034c
MoÀMo “long”
1f
4.09 0.0036c 4.06 0.0033c 4.06 0.0034c
R
10.5
8.5
9.2
E0 [eV]
À1.78
9.08
7.88
determined by calculating confidence limits and statistical F
parameters. The X-ray absorption fine structure (XAFS) spectra
of (Mo,V)5O14 and its corresponding oxide nitrides could be
well described with three Mo–O and three Mo–Mo distances.
The disorder parameter, s2, of the shortest Mo–O distance in-
creased with increasing nitrogen content whereas s2 of the
second Mo–O distance was not affected by the ammonolysis
(Figure 4). The formation of structural defects such as oxygen
vacancies or incorporated ions might cause structural disorder.
The analysis of the local structure around absorbing Mo atoms
revealed a modification of the anion lattice. Ammonolysis of
(Mo,V)5O14 did not significantly affect the cation lattice. The
XRD technique represents a complementary method to XAS
technique by providing information about the long-range
structure. Combining these two methods, a detailed characteri-
Figure 3. Size and strain analysis of all samples by the Williamson–Hall
method using a set of reflections of all crystallographic directions.
breadth of selected reflections with varying hkl was deter-
mined by single line fitting. A linear evolution of Williamson–
Hall plots indicated a homogenous distribution of particle size
and lattice strain in the samples. The average particle size
could be calculated from the intercept with the axis and the
lattice strain was obtained from the slope. The intercept with
the axis revealed the average particle size of about 56 nm for
all samples (Table 1). Conversely, the lattice strain changed sig-
nificantly from 0.061 to 0.171 after ammonolysis of (Mo,V)5O14.
In addition to XRD investigations X-ray absorption spectros-
copy was performed to analyze the influence of nitrogen incor-
poration on the short-range structure of the MoV oxide nitrides
with 1 and 1.5 wt% N. The Mo5O14 structure contains six metal
sites. Thus, the refinement strategy was adapted to simulate
an averaged spectrum of the Mo5O14 structure. Averaging dis-
tances can be used for analyzing complex structures such as
(Mo,V)5O14.[30,31] The metal–oxygen and metal–metal distances
based on the XRD Rietveld refinements were averaged and
grouped into three types of distances. This structure model
Figure 4. Evolution of the disorder parameter s2 of two Mo–O distances ob-
tained by EXAFS refinements as a function of nitrogen content.
ChemCatChem 2016, 8, 758 – 766
www.chemcatchem.org
760
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
zation of the structural influences of ammonolysis of
(Mo,V)5O14 could be performed. The increasing lattice strain ob-
tained by XRD confirmed the increasing structural disorder in
the local structure of the anion lattice.
oxygen mobility and altered redox properties for the selective
oxidation of propene.[35,36]
With respect to the IR spectrum of (Mo,V)5O14, an additional
band at 1055 cmÀ1 appeared in the spectrum of all oxide ni-
trides. This band could be assigned to the stretching mode of
MoÀN bonds.[37] Detecting a MoÀN stretching mode indicated
the incorporation of nitrogen atoms in the anion lattice of
(Mo,V)5O14. Additionally, oxide nitrides with 1.3 and 1.5 wt% ni-
trogen showed a NÀH bending mode at 1405 cmÀ1.[14] NÀH
stretching modes at higher wavenumbers were not detectable
in the used setup. Hence, intensified treatment of (Mo,V)5O14
with ammonia resulted in the formation of incorporated
(NH)2À.
FTIR spectra were recorded to compare characteristic vibra-
tional modes before and after ammonolysis (Figure 5). Refining
Compensation of additional negative charge
Substitution of oxygen ions O2À by nitrogen ions N3À in the
anion lattice of (Mo,V)5O14 leads to an additional negative
charge, which must be compensated. Charge compensation
processes include: (i) oxidation of one type of metal center
(Mo or V), (ii) additional incorporation of positive charges like
H+, for example, in the form of (NH)2À, or (iii) generation of
anion vacancies, for example:[38]
Figure 5. FTIR spectra of (Mo,V)5O14 and its corresponding oxide nitrides
with different nitrogen contents.
2NH3 þ 3OXO ¼ 2NO0 þ VÁOÁ þ 3H2O
ð3Þ
of Gaussian functions to the spectra resulted in maximum
peak positions which are summarized in Table 3. The spectra
showed the symmetric stretching vibrations of the Mo=O
group in a wavenumber range from 900 to 1000 cmÀ1. The
bands at lower wavenumbers are assigned to asymmetric vi-
brations of MoÀOÀMe (Me=Mo, V) bridging bonds.[32–34] After
ammonolysis the same bands could be identified. Additionally,
a slight shift of MoÀOÀMe and MoÀO bands to lower wave-
numbers was revealed by analysis of band positions in the
FTIR spectra of Mo5O14-type oxide nitrides. Gallego reported
a similar behavior for the characterization of LaFeO3ÀxNx and
attributed this shift to an incorporation of nitrogen in iron–
oxygen octahedral.[25] The band shift to lower wavenumbers
corresponds to a weakening of bond strength. Hence, decreas-
ing metal–oxygen bond strength may lead to increased
X-ray absorption near edge spectroscopy (XANES) was used
to determine the oxidation state of the Mo and V metal cen-
ters. Analyzing characteristic maxima in the Mo XANES spectra
revealed the average valence of molybdenum metal centers.[9]
Various molybdenum oxides were used as references for a cor-
relation between average valence and energy of the Mo K-
edge (Figure 6). Analysis of Mo XANES spectra revealed an
average molybdenum valence of 5.66 for the oxide and 5.61
for the oxide nitride with 1.0 and 1.5 wt% nitrogen.
A refinement of three Gaussian functions and an arctangent
function to simulate the absorption edge jump resulted in
a good agreement with V XANES spectra.[39] The centroid of
the arctangent function represented the position of the ab-
sorption edge. Fitting XANES spectra of a set of vana-
dium oxide references (V2O3, V6O13, and V7O13) yielded
a linear correlation between adsorption edge energy
and average vanadium valance. Applying this fitting
Table 3. Positions and assigned vibrational modes from FTIR refinements of
(Mo,V)5O14 and oxide nitrides with different nitrogen contents.[14,32–34,37]
method resulted in an average vanadium valence of
4.50 for the oxide and 4.40 for the oxide nitrides.
In summary, oxidation of Mo or V metal centers for
charge compensation could be excluded by XAS
measurements. An additional incorporation of posi-
tive charge by H+ was only detectable for samples
with 1.3 and 1.5 wt% nitrogen. Elemental analysis of
oxide nitrides confirmed the absence of hydrogen at
1.0 wt% nitrogen. It might be possible that formation
of (NH)2À is a parallel process by intensified treatment
of (Mo,V)5O14. Apparently, at low nitrogen contents
the additional negative charge was compensated by
the formation of oxygen vacancies.
Oxide
Oxide nitride
1 wt% N
Oxide nitride
1.3 wt% N
Oxide nitride
1.5 wt% N
Vibrational
mode
569
592
628
731
814
865
909
565
589
632
728
810
861
904
564
589
632
732
813
865
907
559
585
631
731
813
861
895
MoÀOÀM
VÀOÀMo
MoÀOÀM
MoÀOÀM
MoÀOÀM
MoÀOÀM
MoÀO
symmetric stretching
MoÀO
980
976
976
955
symmetric stretching
MoÀN stretching
NÀH bending
–
–
1055
–
1057
1405
1057
1409
ChemCatChem 2016, 8, 758 – 766
www.chemcatchem.org
761
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Figure 6. Determination of average valences of molybdenum and vanadium.
The calibration was performed by a set of references of each element
(squares).
Figure 8. Evolution of conductivity during thermal treatment (330–650 K) of
(Mo,V)5O14 and its corresponding oxide nitrides with different nitrogen con-
tents in air and ion current m/Z 28 (nitrogen, green line).
Electronic properties
(Figure 8). All samples exhibited a semiconducting behavior
corresponding to an increasing conductivity with temperature.
In contrast to (Mo,V)5O14 the conductivities of all oxide nitrides
decreased slightly in a temperature range of 550 to 620 K. Ana-
lyzing the evolved gas phase by mass spectrometer revealed
a release of nitrogen (m/e 28) in the same temperature range.
After release of nitrogen the conductivities increased again. El-
emental analysis of treated samples revealed complete remov-
al of nitrogen. Comparable studies on MoO3-type oxide nitrides
showed that conductivities after nitrogen removal were the
same as that for a-MoO3.[13] Conversely, conductivities of
Mo5O14-type oxide nitrides exhibited a dependence on nitro-
gen content above 620 K. Apparently, structural defects caused
by anion substitution in (Mo,V)5O14 were preserved after nitro-
gen removal in the temperature range of selective propene ox-
idation.
Substitution of anions associated with the formation of oxygen
vacancies should strongly influence the electronic properties
of the oxide nitrides. Electrical properties were measured by
impedance spectroscopy. Nyquist plots (negative imaginary
part of impedance Z’’ vs. real part of impedance Z’) are shown
in Figure 7. The spectrum of each sample consisted of slightly
From an Arrhenius-type presentation of conductivities of all
samples up to 500 K activation energies for extrinsic conduc-
tion were calculated (Figure 9). Owing to nitrogen removal
from 550 to 620 K and phase transformation above 673 K in air
no reliable activation energies could be obtained at higher
temperatures. The activation energies for extrinsic conduction
decreased with higher degree of substitution from 0.44 to
0.35 eV. The decreasing activation energies corroborated the
Figure 7. Nyquist presentation of impedance measurements of (Mo,V)5O14
and its corresponding oxide nitrides with different nitrogen contents at
298 K.
depressed semicircles. The ohmic resistances were determined
by refining a calculated equivalent circuit to the experimental
spectra. A parallel connection of an ohmic resistor R and a con-
stant phase element (CPE) was used as equivalent circuit.[40]
Using a CPE described the deformed semicircles in the experi-
mental data and represents a widely used empirical model.
The refinement resulted in an increased conductivity with
higher nitrogen contents. Bulk conductivities exhibited an ex-
ponential dependence on nitrogen content. The increase of
electrical conductivity with higher substitution degree might
be due to the formation of oxygen vacancies.[41] Formation of
oxygen vacancies can also influence the mobility of lattice
oxygen.[42,43]
Figure 9. Arrhenius type presentation of conductivities (330–500 K) of
(Mo,V)5O14 and its corresponding oxides nitrides. Activation energies of ex-
trinsic conduction process are indicated.
Conductivities of (Mo,V)5O14 and oxide nitrides with different
nitrogen contents as a function of temperature are shown in
ChemCatChem 2016, 8, 758 – 766
www.chemcatchem.org
762
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
formation of structural defects. Apparently, the formation of
oxygen vacancies led to changes in band structure. Conse-
quently, electron transition from valence band to conduction
band was facilitated.
tion a=0.5 (phase ratio Mo5O14/MoO2 =1) and conductivity.
This relation corroborated that the increasing conductivity was
associated with increasing oxygen availability.
Catalytic behavior in selective propene oxidation
Reaction rates for propene conversion and acrolein formation
at different temperatures are shown in Figure 11. Only the
Reducibility
According to the redox mechanism for the selective oxidation
of propene, the availability of lattice oxygen plays an impor-
tant role. Temperature-programmed reduction experiments
(TPR) are commonly used to elucidate the availability of lattice
oxygen of catalysts in selective oxidation reactions.[44,45] Here,
reducibility of (Mo,V)5O14 and the corresponding oxide nitrides
was investigated by using in situ XAS during TPR in 5% pro-
pene in helium up to 763 K. The reduction of (Mo,V)5O14 pro-
ceeded in a one-step reaction to MoO2 without any detectable
intermediate phases.[10] The measured XANES spectra were
fitted with a linear combination (LC) of (Mo,V)5O14 and MoO2
reference spectra to determine the phase composition during
TPR (Figure 10). Results of LC XANES fits revealed a strong de-
pendence of the onset of reduction on nitrogen content. The
onset temperature was lowered from 700 to 650 K by incorpo-
ration of nitrogen. With regard to the similar long-range struc-
ture and crystallinity, the improved reducibility of the oxide ni-
trides can be attributed to an increased availability of lattice
oxygen. This increased oxygen availability correlated with an
increased density of oxygen vacancies. A linear correlation
could be established between temperature of extent of reduc-
Figure 11. Reaction rates for propene conversion (bottom) and acrolein for-
mation (top) based on the catalyst’s initial weight at 648, 673, and 698 K in
5% propene and 5% oxygen.
rates of the sample with the highest nitrogen content were
slightly increased at 698 K. Propene conversion and acrolein
formation exhibited a similar evolution after nitrogen substitu-
tion. Selectivities towards acrolein were determined for each
sample at different temperatures and similar propene conver-
sions as well as similar time on stream (Figure 12). The selectiv-
ities towards acrolein increased with higher nitrogen contents
of the samples at every temperature while the fraction of total
oxidation products COx decreased. The increase of selectivity
for the oxide nitride with 1.0 wt% nitrogen was independent
of temperature. Selectivities of oxide nitrides with 1.3 and
1.5 wt% nitrogen were dependent on temperature. The
change of selectivity with increasing nitrogen content was
more distinct at low temperatures. At 698 K the increase of se-
lectivity could be described by a linear regression over the
Figure 10. Top: Evolution of phase composition during TPR in 5% propene
up to 763 K of (Mo,V)5O14 and its oxide nitrides with 1.0 and 1.5 wt% N.
Bottom: Conductivities of impedance measurements as a function of tem-
peratures at a conversion degree a=0.5 during TPR indicated a correlation
between conductivity and oxygen availability.
Figure 12. Dependency of selectivity towards acrolein on nitrogen contents
at 648 K, 673 K and 698 K in 5% propene and 5% oxygen.
ChemCatChem 2016, 8, 758 – 766
www.chemcatchem.org
763
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
whole range of nitrogen contents. FTIR measurements indicat-
ed the formation of NÀH bonds in the samples with 1.3 and
1.5 wt% nitrogen. Hence, the formation of NÀH groups might
cause an additional effect on selectivity. An elemental analysis
of samples treated at 698 K yielded the remaining nitrogen
content of the samples (0.25 wt%). The dependence of selec-
tivities of oxide nitrides with 1.3 and 1.5 wt% nitrogen on tem-
perature might result from the partial removal of nitrogen.
XRD analysis of these samples showed that the Mo5O14 struc-
ture was preserved during catalytic treatment. This was corro-
borated by in situ XRD and XAS experiments.
Several examples from the literature show that supported
and highly dispersed transition metal oxides species are also
active in selective propene oxidation.[48,49] Hence, migration of
anion vacancy in the bulk may not be mandatory for catalytic
activity. However, the reoxidation of supported catalysts is also
associated with a transfer of electrons to adsorbed oxygen
species on MÀOÀM units. Isolated monomeric species on sup-
port material showed very low catalytic activity owing to the
missing transfer of electrons.[50] However, the role of electron
transport is difficult to access because conductivity represents
a bulk characteristic. Consequently, the influence of conductivi-
ty can best be investigated by using bulk model systems. Here,
the correlation between oxygen mobility and selectivity to-
wards acrolein could be demonstrated for the first time with-
out changing the crystal structure of the catalyst or addition of
promoting cations.
Correlation of conductivity and selectivity
Characterization of (Mo,V)5O14 and the corresponding oxide ni-
trides revealed a significant correlation between conductivity,
selectivity, and the nitrogen content in the samples. The plot
in Figure 13 shows the selectivity towards acrolein and COx as
Conclusions
Oxygen ions of (Mo,V)5O14 were substituted by nitrogen ions
through ammonolysis to prepare mixed molybdenum oxide ni-
tride. Compared to previous studies on MoO3-type oxide ni-
trides this approach could be successfully applied to a more
relevant system for studies in the selective oxidation of pro-
pene. In addition, the stability of oxide nitrides was enhanced.
Investigations on the ammonolysis of (Mo,V)5O14 resulted in
oxide nitrides with a preserved crystal structure and an invari-
ant cation composition. However, modification of the anion
lattice led to a significantly varied conductivity and reducibility.
The formation of oxygen vacancies compensated the addi-
tional negative charge of nitrogen ions compared to that of
oxygen ions. Samples with higher nitrogen content showed an
additional formation of (NH)2À. Ammonolysis led to a weaken-
ing of metal–oxygen bond strength as well as increased bulk
conductivity. Moreover, the resulting enhanced conductivity
correlated with an increased selectivity towards acrolein in se-
lective oxidation of propene. Apparently, Mo5O14 oxide nitrides
represent a suitable model system for studying correlations be-
tween conductivity, oxygen availability, and catalytic per-
formance.
Figure 13. Selectivity towards acrolein and COx as a function of conductivi-
ties indicating the oxygen availability at 648 K.
a function of conductivity at 648 K. Selectivity towards acrolein
increased linearly with increasing conductivity. As discussed
above conductivity represented a measure of mobility of lat-
tice oxygen. Consequently, increasing oxygen mobility led to
an increased selectivity towards acrolein and a decreased for-
mation of COx. According to the redox mechanism nucleophilic
lattice oxygen of the bulk is inserted in the adsorbed allylic
species to form acrolein. Conversely, electrophilic species are
more likely to form total oxidation products.[46] According to
Grasselli the resulting anion vacancy migrates to another site
and is replaced by adjacent lattice oxygen.[6] Gaseous oxygen
is adsorbed at another site on the surface and replenishes the
anion vacancies:[47]
Experimental Section
Catalyst preparation
The (Mo,V)5O14 precursor was prepared as described elsewhere.[10]
An aqueous solution of vanadyl oxalate was obtained by adding
oxalic acid (180 mg) to a suspension of V2O5 (109.2 mg) in bidestil-
led water (50 mL) at 353 K. The resulting blue solution was mixed
with a solution of ammonia heptamolybdate (AHM, 3 g) in bidestil-
led water (20 mL). After stirring the solution for 1 h at 353 K the
MoV oxide precursor was crystallized at 338 K. The MoV oxide pre-
cursor was treated in helium at 773 K for 4 h followed by dissolu-
tion of impurities in 1m ammonia solution. Finally, the (Mo,V)5O14
model catalyst was obtained by an additional treatment in helium
at 773 K for 4 h. The metal content in the (Mo,V)5O14 phase was
93 mol% Mo and 7 mol% V.
O2;ads þ eÀ Ð OÀ2 þ eÀ Ð O22À Ð 2OÀ þ 2eÀ Ð 2O2À
ð4Þ
The required electrons are transported in the bulk sample.
Investigations of conductivities of the samples revealed an in-
creased electron transport in the bulk sample and electrons
might be readily available to form nucleophilic lattice oxygen
O2À.
ChemCatChem 2016, 8, 758 – 766
www.chemcatchem.org
764
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
According to previous work, a furnace with a silica tube and
a direct gas supply was used to produce the corresponding oxide
nitrides.[13] Preparation conditions (temperature and ammonia flow
rate) were varied to preserve the Mo5O14 structure and optimize
the incorporation of nitrogen. Gas flows were adjusted by mass
flow controllers (Bronkhorst). Ammonolysis was performed in
a temperature range from 498 K to 573 K and the ammonia gas
Catalytic activity
Catalytic activities were measured in a conventional fixed-bed reac-
tor connected to an online gas chromatography system (CP-3800,
Varian). The fixed-bed reactor consisted of a SiO2 tube (length
30 cm, inner diameter 9 mm) placed vertically in a tube furnace.
The sample was placed on a P3 frit in the center of the isothermal
zone. The catalyst bed in the reactor was approximately 2 cm in
height. To achieve a constant volume in the reactor and to quench
thermal effects, catalyst samples (about 50 mg) were diluted with
boron nitride (hexagonal, Alfa Aesar, 99.5%) to result in an overall
sample mass of 250 mg. To ensure differential reaction conditions,
the reactor was operated at propene conversion levels below 10%.
flow was varied between 5 and 20 LhÀ1
.
Characterization methods
XRD patterns were recorded on an X’Pert PRO MPD diffractometer
(Panalytical) in q–q geometry using CuKa radiation and a solid-state
multichannel PIXcel detector. Measurements were performed in re-
flection mode in a range of 5 to 808 2q in steps of 0.0138 2q with
a silicon sample holder. Rietveld refinements were performed by
using the FullPROF program.[17]
Hydrocarbons and oxygenated reaction products (acetic aldehyde,
propionic aldehyde, acetone, acrolein, isopropyl alcohol, n-propa-
nol, allyl alcohol, acetic acid, propanoic acid, acrylic acid, acryloni-
trile, acetonitrile, propanenitrile) were by analyzed using a Carbo-
wax 52CB capillary column, connected to an Al2O3/MAPD capillary
column, and a fused silica restriction (25 m·0.32 mm). Each column
was connected to a flame ionization detector. Permanent gases
(O2, N2, CO2, CO) were separated and analyzed using a Varian CP-
3800 permanent gas analyzer connected to a thermal conductivity
detector. Reactant gas flow rates of oxygen, propene, and helium
were adjusted through separate mass flow controllers (Bronkhorst)
to a total flow of 40 mLminÀ1. A mixture of 5% propene and 5%
oxygen in helium was used for catalytic tests in the temperature
range of 648–698 K. Additionally, a mass spectrometer (Omnistar,
Pfeiffer) was connected to continuously monitor reactant and
product gas composition. The experimental error of the given reac-
tion rates was estimated from the relative errors of reactant gas
flow rates and catalyst sample masses.
A Magna System 750 (Nicolet) was used to measure IR spectra of
the samples in a wavenumber range of 400–4000 cmÀ1. Samples
were diluted with CsI (1:300) pressed into pellets of 13 mm in di-
ameter.
Elemental contents of C, H, and N were determined by using an
analyzer (FlashEA 1112 NC, ThermoFinnigan/ThermoElectron) with
CHNS-O configuration. Measurements were performed to deter-
mine nitrogen contents after ammonolysis of (Mo,V)5O14.
Impedance spectra of mixed molybdenum oxides and oxide ni-
trides were obtained by measuring the magnitude jZj and the
phase f of an alternating current as a response of an applied alter-
nating potential (impedance analyzer N4L: IAI+PSM1735). The real
part Z’ and the imaginary part Z’’ of the impedance were calculat-
ed from these results. The impedance was measured as a function
of frequency (1 Hz–10 MHz) and temperature. Oxides and oxide ni-
tride samples were pressed to pellets with a diameter of 5 mm (ini-
tial weight 50 mg, 750 MPa pressure) and placed between two Au
disc electrodes for impedance measurements.
Acknowledgements
The Hamburg Synchrotron Radiation Laboratory, HASYLAB, is ac-
knowledged for providing beamtime for this work. We are grate-
ful to Dr. J. Scholz, A. Müller, Dr. R. Zubrzycki, and G. Koch for
contributing to the characterization of the materials. We also ac-
knowledge financial support from the Deutsche Forschungsge-
meinschaft (DFG).
Transmission XAS experiments were conducted at Mo K-edge
(20.0 keV) and V K-edge (5.465 keV) at beamline X1 and C at the
Hamburg Synchrotron Radiation Laboratory, HASYLAB, respectively.
For ex situ measurements samples were diluted with wax (Hoechst
wax C micropowder, Merck) and pressed into self-supporting pel-
lets with a diameter of 13 mm. Boron nitride (hexagonal, Alfa
Aesar, 99.5%) was used as diluent for in situ measurements using
self-supporting pellets with a diameter of 5 mm. Sample masses
were calculated to result in an edge jump around Dm(d)=1.5 at
the Mo K edge and Dm(d)=0.3 at the V K edge.
Keywords: conducting materials · molybdenum · nitrides ·
oxidation · structure–activity relationships
´
[1] B. Grzybowska-Swierkosz, Top. Catal. 2000, 11/12, 23–42.
[2] J. Haber, E. Lalik, Catal. Today 1997, 33, 119–137.
[3] R. K. Grasselli, Catal. Today 1999, 49, 141–153.
[4] P. Mars, D. W. van Krevelen, Chem. Eng. Sci. 1954, 3, 41–59.
[5] G. W. Keulks, L. D. Krenzke, T. M. Notermann, Adv. Catal. 1979, 27, 183–
225.
[6] R. K. Grasselli, Top. Catal. 2002, 21, 79–88.
[7] A. Guerrero-Ruiz, I. Rodriguez-Ramos, P. Ferreira-Aparicio, M. Abon, J. C.
Volta, Catal. Today 1996, 32, 223–227.
In situ XAS experiments were performed in transmission mode in
a flow reactor at atmospheric pressure with 5% propene in helium
(~30 mLminÀ1).[18] TPR experiments were performed in a tempera-
ture range from 298 to 763 K. The gas-phase composition at the
cell outlet was continuously monitored by using a mass spectrom-
eter (Omnistar, Pfeifer).
[8] T. Ressler, J. Wienold, R. E. Jentoft, F. Girgsdies, Eur. J. Inorg. Chem. 2003,
301–312.
[9] T. Ressler, J. Catal. 2002, 210, 67–83.
[10] E. Rçdel, O. Timpe, A. Trunschke, G. Zenkovets, G. Kryukova, R. Schlçgl,
T. Ressler, Catal. Today 2007, 126, 112–118.
[11] M. O. Guerrero-PØrez, M. A. BaÇares, Catal. Today 2009, 142, 245–251.
[12] X. Li, J. Zhao, W. Ji, Z. Zhang, Y. Chen, C. Au, S. Han, H. Hibst, J. Catal.
2006, 237, 58–66.
Data analysis was performed by using the software package
WinXAS v3.2.[19] Background subtraction and normalization were
performed by fitting a linear polynomial and a third-degree poly-
nomial to the pre-edge and post-edge region of an absorption
spectrum, respectively. The EXAFS c(k) was extracted by using
cubic splines to obtain a smooth atomic background m0(k). The
FT(c(k)·k3), often referred to a pseudo radial distribution function,
was calculated by Fourier transforming the k3-weighted experimen-
tal c(k) function, multiplied by a Bessel window, in the R space.
[13] S. Kühn, P. Schmidt-Zhang, A. H. P. Hahn, M. Huber, M. Lerch, T. Ressler,
Chem. Cent. J. 2011, 5, 42.
ChemCatChem 2016, 8, 758 – 766
www.chemcatchem.org
765
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
[14] X.-J. Wang, R. Nesper, C. Villevieille, P. Novµk, Adv. Energy Mater. 2013, 3,
606–614.
[15] S. Knobl, J. Catal. 2003, 215, 177–187.
[32] F. Cariati, J. Bart, A. Sgamellotti, Inorg. Chim. Acta 1981, 48, 97–103.
[33] S. Ishikawa, T. Murayama, S. Ohmura, M. Sadakane, W. Ueda, Chem.
Mater. 2013, 25, 2211–2219.
[16] R. Schlçgl, A. Knop-Gericke, M. Hävecker, U. Wild, D. Frickel, T. Ressler,
R. E. Jentoft, J. Wienold, G. Mestl, A. Blume, O. Timpe, Y. Uchida, Top.
Catal. 2001, 15, 219–228.
[17] J. Rodriguez-Carvajal, Abstracts of the Satellite Meeting on Powder Dif-
fraction of the XV Congress of the IUCr 1990, 127.
[18] T. Ressler, Anal. Bioanal. Chem. 2003, 376, 584–593.
[19] T. Ressler, J. Synchrotron Radiat. 1998, 5, 118–122.
[20] F. Portemer, M. Sundberg, L. Kihlborg, M. Figlarz, J. Solid State Chem.
1993, 103, 403–414.
[34] P. Botella, J. Catal. 2002, 209, 445–455.
[35] D. Martin, D. Duprez, J. Phys. Chem. 1996, 100, 9429–9438.
[36] R. B. Watson, U. S. Ozkan, J. Mol. Catal. A 2003, 194, 115–135.
[37] X. Wang, L. Andrews, R. Lindh, V. Veryazov, B. O. Roos, J. Phys. Chem. A
2008, 112, 8030–8037.
[38] S. G. Ebbinghaus, H.-P. Abicht, R. Dronskowski, T. Müller, A. Reller, A.
Weidenkaff, Prog. Solid State Chem. 2009, 37, 173–205.
[39] D. Haskel, Z. Islam, J. Lang, C. Kmety, G. Srajer, K. Pokhodnya, A. Epstein,
J. Miller, Phys. Rev. B 2004, 70, 054422.
[21] N. V. Troitskaya, Z. G. Pinsker, Kristallografiya 1959, 4, 33.
[22] L. W. Finger, D. E. Cox, A. P. Jephcoat, J. Appl. Crystallogr. 1994, 27, 892–
900.
[40] M. Shoar Abouzari, F. Berkemeier, G. Schmitz, D. Wilmer, Solid State
Ionics 2009, 180, 922–927.
[41] A. Pantazidis, A. Auroux, J.-M. Herrmann, C. Mirodatos, Catal. Today
[23] M. A. G. Aranda, E. R. Losilla, A. Cabeza, S. Bruque, J. Appl. Crystallogr.
1998, 31, 16–21.
1996, 32, 81–88.
´
[42] B. Grzybowska-Swierkosz, Top. Catal. 2002, 21, 35–46.
[24] A. G. de La Torre, S. Bruque, M. A. G. Aranda, J. Appl. Crystallogr. 2001,
34, 196–202.
[25] G. Sierra Gallego, N. Marín Alzate, O. Arnache, J. Alloys Compd. 2013,
549, 163–169.
[26] D. Logvinovich, R. Aguiar, R. Robert, M. Trottmann, S. Ebbinghaus, A.
Reller, A. Weidenkaff, J. Solid State Chem. 2007, 180, 2649–2654.
[27] G. Williamson, W. Hall, Acta Metall. 1953, 1, 22–31.
[28] N. GonÅalves, J. Carvalho, Z. Lima, J. Sasaki, Mater. Lett. 2012, 72, 36–
38.
[43] L. Y. Margolis, A. A. Firsova, Int. Rev. Phys. Chem. 1989, 8, 1–20.
[44] T. Franzke, F. Rosowski, M. Muhler, Chem. Ing. Tech. 2011, 83, 1705–
1710.
[45] L. F. Liotta, M. Ousmane, G. Di Carlo, G. Pantaleo, G. Deganello, G. Marcì,
L. Retailleau, A. Giroir-Fendler, Appl. Catal. A 2008, 347, 81–88.
[46] J. Haber, W. Turek, J. Catal. 2000, 190, 320–326.
[47] E. Bordes, C. R. Acad. Sci. Ser. IIc 2000, 3, 725–733.
[48] R. Zubrzycki, J. D. Epping, T. Ressler, ChemCatChem 2015, 7, 1112–1121.
[49] A. N. Desikan, W. M. Zhang, S. T. Oyama, J. Catal. 1995, 157, 740–748.
[50] J. Scholz, A. Walter, T. Ressler, J. Catal. 2014, 309, 105–114.
[29] K. Venkateswarlu, A. Chandra Bose, N. Rameshbabu, Phys. B 2010, 405,
4256–4261.
[30] T. Y. Kardash, D. I. Kochubei, L. M. Plyasova, V. M. Bondareva, J. Struct.
Chem. 2008, 49, 116–123.
Received: October 1, 2015
[31] E. Rçdel, Dissertation, TU Berlin, Berlin, 2006.
Published online on January 15, 2016
ChemCatChem 2016, 8, 758 – 766
www.chemcatchem.org
766
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim