How important is the (001) plane of M1 for selective oxidation of propane to acrylic acid?

Article abstract of DOI:10.1016/j.jcat.2008.05.028

The role of the (001) crystallographic plane of the M1 phase of MoVTeNb mixed-oxide catalysts in selective oxidation of propane to acrylic acid was addressed by investigating a phase-pure M1 material preferentially exposing this surface. A model catalyst was prepared by complete silylation of M1, followed by breakage of the SiO₂-covered needles. Using this approach, the reactivity of the M1 (001) surface was investigated by combining a microreactor study of propane oxidation with high-sensitivity low-energy ion scattering (HS-LEIS). Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) were used to study the shape and microstructure of the model system and verify the surface exposure of the model catalyst. The specific rate of formation of acrylic acid on the model catalyst was found to be similar to that on the phase-pure M1 reference material, indicating that the (001) plane of the M1 crystal structure did not have better catalytic properties compared with the lateral surface of M1 needles in propane oxidation.

Full text of DOI:10.1016/j.jcat.2008.05.028

Journal of Catalysis 258 (2008) 35–43
Contents lists available at ScienceDirect
Journal of Catalysis
www.elsevier.com/locate/jcat
How important is the (001) plane of M1 for selective oxidation of propane to
acrylic acid?
A. Celaya Sanfiz a, T.W. Hansen a, A. Sakthivel a, A. Trunschke a,∗, R. Schlögl a, A. Knoester b, H.H. Brongersma b,
M.H. Looi c, S.B.A. Hamid c
a
Fritz Haber Institute of the Max Planck Society, Department of Inorganic Chemistry, Faradayweg 4-6, D-14195 Berlin, Germany
Calipso B.V., P.O. Box 513, 5600 MB Eindhoven, The Netherlands
COMBICAT, University Malaya, 50603 Kuala Lumpur, Malaysia
b
c
a r t i c l e i n f o
a b s t r a c t
Article history:
The role of the (001) crystallographic plane of the M1 phase of MoVTeNb mixed-oxide catalysts in
selective oxidation of propane to acrylic acid was addressed by investigating a phase-pure M1 material
preferentially exposing this surface. A model catalyst was prepared by complete silylation of M1, followed
by breakage of the SiO2-covered needles. Using this approach, the reactivity of the M1 (001) surface was
investigated by combining a microreactor study of propane oxidation with high-sensitivity low-energy ion
scattering (HS-LEIS). Scanning electron microscopy (SEM) and transmission electron microscopy (TEM)
were used to study the shape and microstructure of the model system and verify the surface exposure
of the model catalyst. The specific rate of formation of acrylic acid on the model catalyst was found to
be similar to that on the phase-pure M1 reference material, indicating that the (001) plane of the M1
crystal structure did not have better catalytic properties compared with the lateral surface of M1 needles
in propane oxidation.
Received 7 March 2008
Revised 22 May 2008
Accepted 23 May 2008
Available online 3 July 2008
Keywords:
MoVTeNb oxide catalyst
M1
Propane oxidation
Propane ammoxidation
Low-energy ion scattering
LEIS
© 2008 Elsevier Inc. All rights reserved.
1. Introduction
by Nb in M1 with higher Nb content [10]. A bronze-like channel
structure is established by stacking layers of the polyhedrons in
the [001] direction, resulting in a needle-like crystal morphology
in which the {001} planes are arranged perpendicular to the long
axis of the needle. It has been suggested that terminating (001)
planes of the M1 phase contain the active and selective surface
sites for partial oxidation reactions [11–14]. This claim has stimu-
lated research on the structure and properties of the M1 surface
by low-energy ion scattering (LEIS) [15,16]. One unique feature of
LEIS analysis is that it gives the atomic composition of the outer-
most atoms of a surface. These outer atoms are precisely the atoms
that largely control the catalytic properties of the solid. It has been
shown previously [17] that in cases where conventional surface an-
alytic techniques, such as XPS, do not show correlation with the
catalytic activity, the extreme surface sensitivity of LEIS gives a
direct relationship between composition and catalysis. A recent re-
view described the underlying principles behind and quantification
of LEIS [18]. The general absence of matrix effects allows the use of
simple reference samples for quantifying the surface composition.
The present study addresses the origin of the positive effect
of grinding on the catalytic activity of MoVTeNb mixed oxides
in the selective oxidation of propane to acrylic acid. Due to the
needle-like shape of the particles, the surface area of (001) planes
comprises only a minor fraction of the total surface area. Thus, in-
creased activity after grinding of M1 needles has been attributed to
the generation of additional (001) surface area [11,12]. Even though
Selective oxidation of light alkanes requires multifunctional cat-
alysts capable of simultaneous activation of C–H bonds, insertion of
oxygen atoms, and prevention of COx formation by further oxida-
tion of intermediates or desired reaction products, which usually
are much more reactive than the inactive reactant. The neces-
sary functional diversity is implemented in chemically and struc-
turally complex MoVTeNbOx catalysts, which show high activity
and selectivity in partial (amm)oxidation of propane to acrylic acid
and acrylonitrile, respectively [1,2]. Usually, these catalysts con-
sist of phase mixtures composed of the so-called “M1” and “M2”
phases [3], as well as minor amounts of phases such as Mo5O14
-
type structures or binary MoV and MoTe oxides. Propane activation
and high selectivities toward acrylic acid and acrylonitrile gener-
ally are attributed to the presence of the orthorhombic M1 phase
[4–6]. The M1 structure (ICSD 55097) consists of corner-sharing
MO6 octahedrons (M = Mo, V), which are assembled in the (001)
plane forming characteristic hexagonal and heptagonal rings host-
ing Te–O units. Niobium is preferentially located in a pentagonal
bipyramidal environment [7–9]; however, due to a certain chemical
flexibility of the phase, octahedral positions also may be occupied
Corresponding author.
E-mail address: trunschke@fhi-berlin.mpg.de (A. Trunschke).
*
0021-9517/$ – see front matter © 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2008.05.028

36
A. Celaya Sanfiz et al. / Journal of Catalysis 258 (2008) 35–43
Scheme 1. Silylation and mechanical treatment of phase-pure M1.
an increased specific surface area is observed [12], the downside of
the grinding process is the well-known effect of mechanical treat-
ment on the nature and concentration of defects on the overall
surface of transition metal oxides [19].
The spray-dried material was calcined in flowing air at 548 K
(heating rate, 5 K/min) for 1 h. The calcined mixed oxide was then
heated for 2 h at 773 K (heating rate, 5 K/min) and 20 MPa in the
presence of steam. The resulting solid was finally crystallized to
M1 in flowing argon at 873 K (heating rate, 15 K/min) for another
2 h (catalyst ID 3030). The atomic ratio of the metals in the fi-
nal catalyst, Mo/V/Te/Nb = 1/0.34/0.08/0.14, as determined by EDX
was close to the nominal ratio, with the exception of the reduced
tellurium content due to evaporation of elemental Te during ther-
mal activation of the catalyst.
Consequently, here we have pursued a different strategy, as
illustrated in Scheme 1. A batch of crystalline, phase-pure M1
prepared by hydrothermal synthesis was divided into two parts.
One part was fully coated with silica. The complete coverage of
the mixed-metal oxide surface by SiO2 was verified by HS-LEIS.
Both the coated material and the noncoated reference M1 were
pressed into pellets, crushed, and sieved to prepare sieve frac-
tions, which were then loaded into a microreactor and studied
in the partial oxidation of propane to acrylic acid. Preparation
of the sieve fraction represents a gentle mechanical treatment
that generates new M1 surfaces, as was quantified by LEIS for
the silica-coated catalyst [18] and by nitrogen adsorption for the
reference material. Special attention was given to a comprehen-
sive microstructural characterization of the model catalysts before
and after propane oxidation. Scanning electron microscopy (SEM)
and transmission electron microscopy (TEM) revealed that com-
plete coverage of M1 followed by breakage of the SiO2-covered
needles generated a model catalyst that predominantly exposes
(001) planes (Scheme 1). In addition, LEIS was applied to com-
pare the chemical composition of the newly formed M1 surface of
the coated catalyst with that of the total surface of the M1 refer-
ence. Based on the results of the microreactor and LEIS study, the
relevance of particular surface terminations are discussed in view
of the catalytic properties of phase-pure M1 catalysts in the partial
oxidation of propane to acrylic acid.
2.2. Silylation of M1
Silylation of M1 was performed by adding 5.8 g of HSi(OEt)3
(Sigma–Aldrich) to a suspension of 0.5 g of phase-pure M1 (cata-
lyst 3030) in 100 mL of toluene. The mixture was kept for 16 h at
384 K under reflux. Cross-linking of anchored silanol groups was
achieved by treating the resulting solid in a mixture of 90 mL
of ethanol, 10 mL of bidistilled water, and 1 mL of concentrated
sulphuric acid for 4 h under reflux. The two-step procedure was
repeated three times to ensure complete silylation of the entire
M1 surface (catalyst ID 3159).
2.3. Activity measurements
For testing of the catalysts in selective oxidation of propane
to acrylic acid, 0.5 g of each material was pelletized (8 ton on a
round, 3-cm-diameter surface) and sieved to 200–400 μm. This
shape-forming procedure, referred to as mechanical treatment,
causes partial disruption of the needles, changing the size distri-
bution of the primary particles in the original M1 and making
catalytic testing of completely silylated M1 impossible. But pure
silica (Aerosil 300) was shown to be inactive in propane oxidation
under the reaction conditions applied. The sieve fractions of the
mechanically treated M1 and silylated M1 were loaded into quartz
reactors with a diameter of 4 mm. The feed was composed of
propane/oxygen/nitrogen/steam in a molar ratio of 0.85/1.9/15.2/12.
The reaction was performed at atmospheric pressure, a GHSV of
4800 h−1, and a reaction temperature of 673 K. The products were
analyzed by gas chromatography. A molecular sieve column and
a Porapak column coupled with a thermal conductivity detector
were used to analyze O2, N2, CO, CO2, and hydrocarbons (C1–C3).
The oxygenated products were analyzed with a HP-FFAP column
coupled with a flame ionization detector. The detection limit of
the individual products depended on the detector used but was
generally >0.2 vol%. A mass balance of 100 10% was calculated.
2. Experimental
2.1. Preparation of M1
A phase-pure M1 catalyst with nominal atomic ratio Mo/V/Te/
Nb = 1/0.3/0.23/0.13 was prepared starting from 18.71 g of MoO3
(Fluka) suspended in 300 mL of bidistilled water. After the ad-
dition of 19.90 g oxalic acid at 363 K, a light-yellowish solution
was obtained. The mixture was stirred for 25 min and then cooled
to 313 K. Then 6.86 g of telluric acid (Sigma-Aldrich) dissolved in
30 mL of bidistilled water at 313 K was added to the molybdenum
oxalate solution. A vanadium solution was synthesized by carefully
mixing 7.05 g of oxalic acid and 3.55 g of V2O5 (Riedel–de-Haën)
in 30 mL of bidistilled water at 338 K. The resulting blue solu-
tion was cooled to 313 K and added to the binary MoTe solution.
Finally, a solution of 7.35 g ammonium niobium oxalate (Sigma–
Aldrich) in 30 mL of bidistilled water was added to this mixture
at 313 K. The clear quaternary solution was stirred for 10 min and
then spray-dried in a Büchi B-191 spray-dryer at an inlet temper-
ature of 423 K. The delivery rate of the pump and the aspirator
were tuned to an outlet temperature of 383 K.
2.4. X-ray diffraction
The phase purity of the materials was verified by X-ray diffrac-
tion (XRD) using
a STOE STADI-P transmission diffractometer

A. Celaya Sanfiz et al. / Journal of Catalysis 258 (2008) 35–43
37
Table 1
Results of particle shape analysis based on SEM images and BET surface areas measured by nitrogen adsorption
Catalyst
Number
of
particles
lmean
(nm)
dmean
(nm)
d/l
Calculateda total
surface area of
Calculateda surface
area of the (001) plane
of one needle (nm2)
Calculateda
BET surface
area
specific surface
one needle (nm2)
area (m2/g)
(m2/g)
M1b
379
241
298
336 189
277 119
275 114
184 99
164 85
167 71
0.56
0.59
0.61
2.47 × 105
1.85 × 105
1.88 × 105
0.53 × 105
0.42 × 105
0.44 × 105
6.3
7.2
7.1
6.7
7.6
n.d.
M1c
Silyl. M1c
a
Calculation under consideration of mean diameter, mean length and under assumption of cylindrical geometry with circular basal plane.
M1 as synthesized.
Catalyst after propane oxidation.
b
c
equipped with a focusing primary Ge (111) monochromator and
a position sensitive detector, using CuKα1 radiation. For data anal-
ysis, the Topas program, v.2.1 (Bruker AXS) was used to fit the
diffraction patterns of the activated materials.
2.5. Electron microscopy
The microstructure of the catalysts was investigated by TEM
using a Philips CM 200 FEG transmission electron microscope op-
erated at 200 kV, equipped with a Gatan CCD camera for image
acquisition and an EDAX Genesis EDX system. Morphology stud-
ies and shape analysis were performed by SEM, using a Hitachi
S-5200 with a PGT Spirit EDX system and a Hitachi S-4800 with an
EDAX Genesis EDX detector. EDX studies in the SEMs were carried
out with an accelerating voltage of 10 kV, with images acquired at
2 kV to optimize surface resolution. For TEM, the specimens were
prepared by dry dispersing the catalyst powder on a standard cop-
per grid coated with holey carbon film. For SEM investigations, the
samples were deposited on carbon tape. The shape of the M1 nee-
dles was analyzed by measuring the length and diameter of more
than 200 needles on SEM images. Based on the average length
and diameter of the needles, the total surface area and the surface
area of the (001) plane were calculated for an average needle for
each catalyst (Table 1). These calculations were done under the as-
sumptions that the material was 100% crystalline and each crystal
was cylindrically shaped. Furthermore, the cylinder was assumed
to have a circular basal plane. A SEM-derived specific surface area
was calculated using the surface area of the average needle and
the crystallographic density of M1 (4.4 g/cm3 [8]).
+
Fig. 1. 3 keV He scattering. References SiO2 and V2O5.
2.6. Nitrogen adsorption
The specific surface areas of the catalysts were measured using
an AUTOSORB-1-C physisorption/chemisorption analyzer (Quan-
tachrome). Eleven points in the linear range of the nitrogen ad-
sorption isotherm measured at 77 K were used to calculate the
BET surface area. Before adsorption, the samples were degassed at
353 K for 2 h.
+
Fig. 2. 5 keV Ne scattering. references V2O5, Nb2O5, MoO3 and TeO2.
2.7. High-sensitivity LEIS
removal of organic contaminations at room temperature without
sputtering the surface. All samples were cleaned in this manner
(200 W, 5 min) before analysis. Prolonged treatment had no influ-
ence on the LEIS spectra, confirming the cleanliness of the samples.
The Calipso LEIS uses a double toroidal analyzer, which com-
bines a large acceptance angle with parallel energy analysis of the
backscattered ions. This gives orders of magnitude higher sensitiv-
ity (HS-LEIS) than that of conventional LEIS equipment [20]. The
required ion dose for analysis is so low that HS-LEIS analysis is
essentially nondestructive (“static”).
2.8. Calibration of the LEIS signals
The extreme surface sensitivity of LEIS requires that organic
contaminations due to transport and storage as well as carbona-
ceous deposits from the catalytic reactions be removed from the
catalysts before analysis. This was done with an oxygen atom
source (Oxford Applied Research, type MPD 21) that produces O-
atoms of thermal energy. In contrast to molecular oxygen and
ozone, the chemical energy of these O-atoms enables the complete
It has been shown previously that roughness has a very small
effect on the LEIS signals [21]. Thus, the precise dispersion is not
of great importance for a good reference for calibrating the LEIS
signals of the catalysts. Thus, high-purity powder samples of SiO2,
V2O5, Nb2O5, MoO3, and TeO2 were used as references to quantify
the surface concentrations of Si, V, Nb, Mo, and Te. Because the
surfaces were cleaned with O atoms, the cations in the outer sur-

38
A. Celaya Sanfiz et al. / Journal of Catalysis 258 (2008) 35–43
+
Fig. 4. 5 keV Ne scattering. Nonsilylated and silylated.
+
Fig. 3. 3 keV He scattering. Nonsilylated, silylated and silylated × 10.
face were in their highest oxidation state (e.g., TeO3 for Te). The
LEIS spectra are given in Figs. 1 and 2. The lighter elements (Si,
+
V) were quantified with 3 keV 4He ions; the heavier elements
+
(V, Nb, Mo and Te), with 5 keV 20Ne ions. As can be seen in
Figs. 1 and 2, the SiO2 surface was contaminated with a trace of
iron oxide, whereas the TeO2 contained some cobalt or nickel oxide
(mass 59). Even with the heavier Ne ions, there was an overlap of
the Nb and Mo peaks. The elements are neighbors in the periodic
table, and their isotopes (Nb: 93; Mo: 92–100) overlap. For the cat-
alysts, the relative concentrations of Nb and Mo were determined
by curve fitting. The large width of the Te peak originates from iso-
topic broadening (atomic mass 122–130). Because these masses are
very different from those of the other elements, Te can be readily
quantified. V can be analyzed using both He and Ne ions, and the
two analyses give very similar results. The values given for V are
the average of the two measurements.
LEIS calibration gives the fractions of the surface that are com-
posed of the various oxides. To obtain atomic concentrations, these
fractions must be corrected for the surface densities of the cations
in the references. These surface densities have been estimated as
(ρ·NAv/M)2/3, where ρ is the bulk density, NAv is Avogadro’s num-
ber, and M is the molar mass. It has been shown previously [22]
that this estimate agrees closely with the estimate from the surface
unit-cell taking the close-packed surface plane, which is dominant
on powders. For V2O5, Nb2O5, MoO3, and TeO3 these densities are
9.96, 9.36, 7.28, and 7.68 (×1014 ) metal atoms/cm2, respectively.
Fig. 5. XRD patterns of as synthesized M1 (a), and calculated patterns assuming
phase-pure M1 [8] (b); XRD patterns of M1 after propane oxidation (c), and of sily-
lated M1 before (d) and after (e) propane oxidation.
backscattered by Nb and Mo/Te in deeper layers. The most likely
process for this is that the ions were neutralized at the initial
interaction with the surface; lost some energy along the ingoing
(straggling) trajectory; backscattered in a hard binary collision with
a Nb, Mo, or Te atom; and lost some more energy while straggling
back to the surface. Because only backscattered ions are detected
in LEIS, the backscattered He atoms must be reionized in an in-
teraction with a surface atom just before leaving the surface. For
3. Results
3.1. Silylation of M1
+
3 keV He ions scattered by this type of oxide, the straggling en-
The surface of a MoVTeNbOx mixed oxide composed mainly of
the M1 phase was covered with a layer of silica by silylation using
HSi(OEt)3. The silylation was optimized using LEIS. Figs. 3 and 4
show the LEIS spectra of the catalyst before and after silylation
ergy loss contributes about 160 eV for backscattering from a depth
of 1 nm. The shape of the 10× magnified background thus gives
information on the thickness distribution of the silica coating. It
shows that a small fraction of the coating is only just covered
(1–2 atoms thick), while the maximum thickness is about 2 nm.
Consistently, TEM (see below) shows that the silica layer is not ho-
mogeneous, and that its thickness varies.
+
according to the process described in Section 2. The 3 keV He
spectrum before silylation shows peaks for O and V and a com-
bined peak for Nb, Mo, and Te, along with some Na contamination.
+
The 5 keV Ne spectrum gives a better mass resolution for the
heavier elements. It shows the peaks for V and Te along with a
broad peak due to Nb and Mo. After silylation, only the peaks for
O and Si (He spectrum) are present. No peaks are observed in the
Ne spectrum. This confirms complete silylation of the catalyst.
In addition to the peaks in the He spectrum, a background ex-
tending to higher energies can be seen, due to ions that were
3.2. Phase composition of the MoVTeNb oxide
In regard to the bulk structure, a material of high phase purity
was exposed to the silylating agent (Fig. 5a), as is evident from
comparing the measured XRD patterns with the patterns calcu-

A. Celaya Sanfiz et al. / Journal of Catalysis 258 (2008) 35–43
39
(a)
(b)
(c)
(d)
Fig. 6. SEM images of phase-pure M1 before (a) and after propane oxidation (b); SEM images of silylated M1 before (c) and after propane oxidation (d).
lated under the assumption that the solid is composed exclusively
of M1 [8] (dotted line in Fig. 5b). Within the method’s range of ac-
curacy, the material is phase-pure. The phase composition of the
mixed oxide remained unchanged after catalytic reaction (Fig. 5c)
and also after the silylation procedure (Figs. 5d and 5e). The mate-
rials are highly crystalline, as also confirmed by TEM.
3.3. Microstructure of the model catalysts
The typical needle-shape morphology of the original phase-
pure M1 material is observed in the SEM image shown in Fig. 6a.
Only very few particles are found that clearly differ in morphology,
indicating the presence of traces of other phases not detectable by
XRD or minor amounts of amorphous fractions. The needle-shape
morphology is maintained after catalytic testing (Fig. 6b) and after
silylation (Fig. 6c). In the latter material, some spherical particles
are observed that have been assigned to pure SiO2 by EDX anal-
ysis. The needle-like shape of the silylated M1 particles (Fig. 6c)
also is not significantly affected by propane oxidation (Fig. 6d).
Shape analysis of the original M1 catalyst reveals a broad dis-
Fig. 7. Distribution of the needle length in as synthesized phase-pure M1 (a), in M1
tribution of the needle length with a maximum between 100 and
400 nm (Fig. 7a). The abundance of shorter needles, especially
those with length shorter than 400 nm, clearly increases after
propane oxidation (Fig. 7b), which is attributed to breakage of the
needles due to mechanical manipulation during preparation of the
sieve fraction before the catalytic test. A similar length distribu-
tion is seen with the silylated catalyst after the catalytic reaction
(Fig. 7c), indicating a comparable impact of mechanical treatment
for silylated and nonsilylated M1.
Table 1 summarizes mean needle length, mean needle diam-
eter and the mean diameter/length ratio of the needles obtained
by shape analysis of about 300 needles for the original M1 ma-
terial, the M1 catalyst after propane oxidation, and the silylated
M1 after catalytic testing. Because the M1 needles are naturally
crystals with a preferred growth direction perpendicular to the
{001} planes, a preferential breaking perpendicular to the length
axis would be expected during mechanical treatment. This pref-
after propane oxidation (b), and in silylated M1 after propane oxidation (c).
erential breaking is reflected in a slight increase of the mean di-
ameter/length ratio from 0.56 for the original M1 catalyst before
propane oxidation to 0.59 and 0.61 for the nonsilylated and the
silylated M1, respectively, after propane oxidation. Assuming that
all needles have cylindrical geometry with circular basal planes
and mean geometric parameters (diameter, length) as given in Ta-
ble 1, the total surface area and the surface area of the (001)
planes of an average needle have been calculated for the three
model catalysts (Table 1). Taking into account a unit cell density of
M1 of 4.4 g/cm3 [8], specific surface areas can be calculated that
agree well with the BET surface areas measured (Table 1), con-
firming the high crystallinity of the materials. If each needle were
disrupted once, an increase in surface area of about 20% would be
expected from the cylindrical geometry. The actual increase in BET

40
A. Celaya Sanfiz et al. / Journal of Catalysis 258 (2008) 35–43
surface area of the nonsilylated M1 before and after propane oxi-
dation of 13% agrees well with the increase in specific surface area
of 14% calculated based on the shape analysis, confirming the reli-
ability of the method. The BET surface area of the silylated M1 was
not measured because it was expected that the presence of silica
particles would falsify the result. However, considering the similar
size distribution of nonsilylated and silylated needles after the cat-
alytic testing (Fig. 7), an increase in surface area of 13% could be
expected for the silylated catalyst as well.
In summary, the combined shape and BET surface area analysis
clearly indicate a preferential disruption of the M1 needles per-
pendicular to the [001] direction by mechanical treatment before
catalytic testing. Comparing nonsilylated and silylated M1 nee-
dles, the similar impact of the mechanical treatment on particle
morphology provides further evidence for the applicability of the
present approach.
Provided that no silica is removed from the sides of the
needles, the new surface area formed by mechanical treatment
of completely silylated M1 is composed mainly of (001) planes
(Scheme 1). Needles that exclusively expose the (001) plane have
been detected by high-resolution transmission electron microscopy
in the silylated M1 catalyst after the catalytic test reaction (Fig. 8a).
The image in Fig. 8a shows one needle with the sides covered by
silica and with the cross-sectional plane free of SiO2, confirming
that the needles can break perpendicular to the longitudinal axis.
Furthermore, needles completely covered by a layer of silica are
also seen (Fig. 8b), indicating that not every needle is broken. This
is in agreement with the results of shape analysis and nitrogen ad-
sorption. The possibility of partial removal of silica from the sides
of the needles cannot be excluded, as evidenced by the TEM image
shown in Fig. 8c.
(a)
(b)
(c)
3.4. Catalytic properties of the model catalysts in propane oxidation
Pure silica shows no catalytic activity in propane oxidation
under the reaction conditions applied. Thus, completely silylated
M1 is expected to be inactive. But experimentally verifying this
statement for a completely silylated model catalyst is impossi-
ble. Mechanical manipulation of the catalyst by pressing and siev-
ing is required to prepare a sieve fraction before catalytic test-
ing. This causes partial damage of aggregates and even of the
primary particles, resulting in exposure of a new MoVTeNbOx
surface in both the silylated and nonsilylated reference catalysts
(Scheme 1).
Table 2 compares the catalytic performance of the two me-
chanically treated catalysts. Propane oxidation was performed in
parallel reactors with the same temperature, space velocity, and
feed composition. Under steady-state conditions, the conversion of
propane is fivefold greater over the reference M1 than over the
silylated model catalyst. Whereas only acrylic acid is formed over
silylated M1, the selectivity is lower for the reference material, pro-
ducing 12% CO2 and 15% CO as byproducts. These differences in
selectivity most likely are related to the different propane con-
version levels. The carbon oxides may be formed in consecutive
reactions, thus lowering the selectivity to acrylic acid over the ref-
erence M1. Table 2 gives average rates. Due to the high conversion
of the nonsilylated reference catalyst, the average consumption
rate of propane is underestimated for this catalyst. Nevertheless,
Fig. 8. HRTEM images of a silylated M1 needle disrupted along the (001) plane (a),
of a completely silylated M1 needle (b), and of a silylated M1 needle with silica
scratched from the sides (c).
Table 2
Catalytic properties of M1 and silylated M1 in propane oxidation to acrylic acid (AA) after 32 h time on stream at 673 K
XC
SAA
YAA
C3H8 consumption rate
Formation rate AA
Catalyst
3H8
(%)
(%)
(%)
mmolC /(h gcat
)
mmolC /(h m2
)
mmolAA/(h gcat
)
mmolAA/(h m2
)
exp.surf.
3H8
3H8
exp.surf.
M1
Silylated M1
49
9
73
100
36
9
>2.94
0.54
>0.39
0.24
>2.15
0.54
>0.28
0.24

A. Celaya Sanfiz et al. / Journal of Catalysis 258 (2008) 35–43
41
Table 3
Metal composition of the bulk (EDX) and topmost layer (LEIS) in at%. The numbers
in parentheses represent molar ratios of the metals normalized to molybdenum
Catalyst
M1a
Method
Mo
V
Te
Nb
EDX
LEIS
64 (1)
n.d.
22 (0.34)
n.d.
5 (0.08)
n.d.
9 (0.14)
n.d.
M1b
EDX
LEIS
64 (1)
65 (1)
20 (0.31)
12 (0.18)
6 (0.09)
12 (0.18)
10 (0.16)
11 (0.17)
Silylated M1a
EDX
LEIS
63 (1)
0
22 (0.35)
0
6 (0.09)
0
9 (0.14)
0
Silylated M1b
EDX
LEIS
62 (1)
55 (1)
21 (0.34)
20 (0.37)
6 (0.09)
15 (0.27)
11 (0.18)
10 (0.17)
a
As synthesized.
Catalyst after propane oxidation.
b
+
Fig. 9. Background-subtracted 3 keV 4He spectra from catalytically tested samples.
+
Fig. 11. Normalised 5 keV 20Ne spectra (Mo = 1.00) from catalytically tested sam-
ples (after subtraction of background).
To obtain the atomic surface concentrations, the surface frac-
tions of the oxides must be corrected for the differing metal
atoms/cm2 in these oxides. Table 3 gives the atomic composi-
tions (in parentheses normalized to Mo) for the treated nonsily-
lated sample and the new surface of the silylated sample, along
with the results of the bulk analyses by EDX for comparison. In
agreement with earlier studies [15], the chemical composition of
the outermost surface layer differs significantly from that of the
bulk. The LEIS measurement of the M1 reference provides impor-
tant information, including the basal plane and the lateral surface,
showing an enrichment of tellurium at the expense of vanadium.
The situation on the freshly formed surface of the silylated model
catalyst, which is dominated by the (001) plane, is different, with
an increase in tellurium at the expense of molybdenum. No depth
profile and no differences between the catalysts with respect to the
Nb concentration are noted. Fig. 11 also illustrates the clear differ-
ences in the surface compositions, with the Ne spectra normalized
on the maximum of the Nb/Mo peak.
+
Fig. 10. 5 keV 20Ne spectra from catalytically tested samples.
it is fivefold greater than that of the silylated catalyst. The larger
MoVTeNbOx surface area that is exposed to the reactants in case
of the nonsilylated reference catalyst is a possible explanation for
this finding. The surface area of the reference M1 was measured
by nitrogen adsorption (Table 1). LEIS was used to determine the
accessible MoVTeNbOx surface area of the silylated catalyst, as de-
scribed in the following section.
3.5. LEIS after catalysis
To properly compare the catalysts, the silylated and nonsilylated
samples were both investigated with LEIS after the same mechan-
ical treatment (pelletizing, crushing, sieving) and catalytic testing
(see Scheme 1). Figs. 9 and 10 show the He and Ne spectra after
background subtraction. Due to partial disruption of the needles,
significant signals can be seen for V, Nb, Mo, and Te in the silylated
sample. Again, some Co/Ni contamination is present. The surface
fractions of the elements in Figs. 9 and 10 can be quantified us-
ing the reference spectra in Figs. 1 and 2. It is seen that the sum
of the surface fractions of V2O5, Nb2O5, MoO3, and TeO3 in the
treated silylated sample amounts to 30% of that in treated nonsi-
lylated sample. The other part of M1 (70%) remains covered by
SiO2. The 30% thus represents the new surface generated by the
mechanical treatment.
4. Discussion
It has been well documented that the M1 phase occurring
in MoVTeNb mixed-oxide catalysts plays a key role in the selec-
tive oxidation and ammoxidation of propane [6,14,23]. The crystal
structure hosts four different metals, three of them with variable
oxidation states, meeting the requirements for catalyzing redox re-
actions by exposing or bearing structural arrangements at the sur-
face that easily meet the demands of propane activation and the
transfer of 8 electrons in its selective oxidation to acrylic acid or
in its ammoxidation to acrylonitrile, respectively. Surface and sub-

42
A. Celaya Sanfiz et al. / Journal of Catalysis 258 (2008) 35–43
surface reduction and reoxidation of M1 crystals without structural
collapse seem to be possible [24]. The occupancy of the 13 metal
sites in the unit cell of the crystal structure is variable [9,10,14].
Accordingly, the catalytic properties may be controlled by synthe-
sis or posttreatment procedures. Crystallinity seems to be required
to obtain catalytic activity for selective (amm)oxidation of propane
[6,25,26]. The possible role of defects remains an open question
[24,27]. Typically, nanocrystalline M1, composed of fine needles a
few hundred nanometers long, exhibits excellent catalytic behav-
ior.
any other damage. The relevance of shape analysis is important,
because in the silylated catalyst, BET surface area measurements
cannot be relied on for calculation due to an expected increase in
surface area caused by the possibly not smooth SiO2 layer and the
occasional formation of pure silica particles.
Both the original M1 needles and the silylated needles have
a low probability of breaking (Table 1, Fig. 7), and long needles
break more frequently than short needles (Fig. 7). Fig. 7 confirms
that nonsilylated and silylated needles break in a similar way, be-
cause the two materials have essentially the same particle size
distribution after the catalytic test. For the silylated catalyst, the
increase in surface area (i.e., the yield in newly formed basal plane
surface area) corresponds to 12.7% when the mean particle size
determined by shape analysis is used for the calculation (Table 1).
In summary, based on shape analysis and/or nitrogen adsorp-
tion before and after catalytic testing, the breakage of the M1
needles benefits in newly formed (001) surface area of approxi-
mately 15% for both the original and the silylated M1.
As an independent method, LEIS analysis was applied to de-
termine the freshly formed MoVTeNbOx surface of the silylated
catalyst. This fraction corresponds to 30% of the surface of the
original M1. This value is twice as high as the increase calcu-
lated from the shape analysis, implying that more than silica-free
(001) planes were generated. HRTEM (Fig. 8c) shows that some
silica was scratched from the sides of the needles as well. Assum-
ing that the approximate 15% increase in surface area calculated
from shape analysis (Table 1) corresponds exclusively to the newly
formed silica-free (001) planes, the ratio of the basal plane area
versus other MoVTeNbOx surface area (i.e., lateral surface area) is
about 1 in the silylated M1. In the original M1, the entire surface
of the needles is exposed to the gas phase during propane oxida-
tion and, thus the ratio of (001) surface area to lateral surface area
corresponds to 0.25 (Table 1). Consequently, even though other
surface planes also are to some extent exposed to the reactants
in the silylated M1 during propane oxidation, we can conclude
that the silylated, mechanically treated M1 catalyst represents a
suitable model system for studying the catalytic properties of the
(001) plane, because half of the exposed MoVTeNbOx surface is
composed of (001) planes.
Table 2 compares the catalytic performance of M1 and silylated
M1. As expected, the conversion of silylated M1 is much lower
than that of M1, because less MoVTeNbOx surface is accessible
in the former catalyst. Based on the information obtained from
the HS-LEIS study indicating that actually 30% more MoVTeNbOx
surface is available in the silylated catalyst, consumption rates of
propane normalized to the area of exposed MoVTeNbOx surface
were calculated considering a 7.6 m2/g MoVTeNbOx surface for
M1 (BET surface area given in Table 1) and a 2.3 m2/g MoVTeNbOx
surface for silylated M1 (0.30×7.6 m2/g). The rates are quite simi-
lar for both catalysts, meaning that the intrinsic catalytic properties
of the entire M1 surface are similar to the catalytic behavior of
a surface composed of 50% (001) planes. This finding casts some
doubt on the uniqueness of the basal plane of M1.
The result is especially surprising considering that the chemical
composition of the newly formed M1 surface of the silylated cat-
alyst (previously unexposed surface of M1) has a different chem-
ical composition than the average surface composition of M1, as
shown by the LEIS experiment. Table 3 compares the bulk com-
position measured by EDX and the surface composition measured
by HS-LEIS. Whereas the EDX bulk compositions of the samples
are almost the same, confirming that the silylation procedure does
not affect the bulk composition, the HS-LEIS surface composition
of the newly generated surface shows relatively large differences,
specifically an increased concentration of tellurium at the expense
of molybdenum. However, prior to the present LEIS measurements,
a mild pretreatment with oxygen atoms at room temperature was
The phase-pure M1 synthesized in the present study ex-
hibits the corresponding morphology and has very small amounts
of impurity phases or amorphous fractions, as evidenced by
XRD (Fig. 5) and electron microscopy (Fig. 6a). Compared with
published M1 stoichiometries (e.g., Mo1V0.15Te0.12Nb0.13Ox [8] or
Mo1V0.23Te0.11Nb0.14Ox [9]), the present material is characterized
by increased V content and reduced Te content (Table 3), which
may be the reason for its exceptional catalytic behavior (Table 2).
The average formation rate of acrylic acid measured in the sta-
tionary state after 32 h on stream is 2.15 mmol/(h gcat), which
exceeds the rates reported for M1 in the literature [28]. In prepar-
ing a model catalyst with preferentially exposed {001} planes we
have chosen the present material due to its high phase purity and
high catalytic activity. Based on grinding experiments, and under
the assumption that the structural characteristics and the chemical
composition of the surface do not differ from the bulk even under
reaction conditions, it was previously concluded that the catalyt-
ically active sites in propane oxidation [11,12] or ammoxidation
[14,26] are located on terminating (001) planes. This assumption
is not consistent with room temperature LEIS measurements that
give an indication of gradients in elemental composition between
surface and bulk of M1 [29]. Surface texturing of catalyst particles
also has been revealed by high-resolution TEM on the surface of
high-performing MoVTeNbOx catalysts that have been leached with
solvents [27]. In any case, a well-defined and rigid arrangement of
structural units may be rather unlikely at the reaction temperature
and in the presence of steam in the feed.
Exploring this question by in situ spectroscopic methods is
quite challenging, however. Thus, in the present study we used
propane oxidation itself as a probe reaction to ascertain the spe-
cific catalytic activity of the (001) plane compared with the in-
tegral activity of the entire M1 surface. For this purpose, the M1
crystals were covered completely by a thin layer of silica. HS-LEIS
confirms coverage of the whole mixed oxide surface (Fig. 4). The
thickness of the layer ranges from a few monolayers to 20 nm, as
verified by LEIS and TEM. Neither the particle size distribution of
the original M1 (Table 1) nor its chemical composition (Table 3)
is affected by the silylation procedure. But the powders were pel-
letized, crushed, and sieved before being introduced into the cat-
alytic test reactor. Such a gentle mechanical treatment leads to par-
tial disruption of needles and formation of fresh MoVTeNbOx sur-
face. For the silylated M1, the newly emerging MoVTeNb oxide sur-
face was verified by HS-LEIS (Figs. 9 and 10). For the pure, nonsi-
lylated M1, these morphological changes should be reflected in an
increased specific surface area as measured by nitrogen adsorp-
tion. Under the assumption that the mechanical treatment leads
exclusively to breakage of needles perpendicular to the [001] di-
rection, the difference corresponds to the surface of newly formed
(001) planes. In the present experiment, the difference in the spe-
cific surface area is 13.4% (Table 1). A very similar result (14.3%) is
obtained if the mean particle size determined based on the shape
analysis by SEM and the bulk density of M1 (4.4 g/cm3 [8]) are
used to calculate the increase in surface area of pure, nonsily-
lated M1 (Table 1), satisfactorily confirming the reliability of the
shape analysis and demonstrating that the mechanical treatment
applied is associated mainly with breakage of needles and not with

A. Celaya Sanfiz et al. / Journal of Catalysis 258 (2008) 35–43
43
necessary to eliminate environmental contamination from the re-
Acknowledgments
action, transport, and storage. This procedure, which converts all
of the surface metal ions into their highest oxidation states, may
be connected with structural changes. Although the local struc-
ture is affected, it is unclear whether the elemental composition
changes significantly. However, the surface reconstruction holds
for the entire surface of M1, thus justifying a comparison of the
basal plane with the total surface in terms of elemental composi-
tion.
The authors thank Dr. Olaf Timpe for helpful discussions, Dr.
Frank Girgsdies for performing the phase analysis of the catalysts,
Edith Kitzelmann for conducting the XRD measurements, and Kil-
ian Klaeden and Gisela Lorenz for carrying out the nitrogen ad-
sorption measurements.
References
In any case, these uncertainties do not affect the findings that
the intrinsic catalytic properties of the basal plane of M1 in the
selective oxidation of propane to acrylic acid do not differ much
from those of the lateral surface of the M1 needles. If we accept
this result, then the question arises as to why increased catalytic
activity and improved selectivity to acrylic acid were observed in
previous studies after grinding of M1 [11,12]. The details of the
grinding procedure were not given in those reports; however, most
likely the mechanical treatment was less gentle than that in the
present study and probably generated a more defect-rich material
along with extra basal planes. Another possible explanation could
be that contamination, such as residues from the preparation pro-
cedure, which are common on catalyst surfaces, were removed by
extended grinding. Breaking the needles and scratching deposited
impurities away from the sides of the needles would expose fresh,
uncontaminated, and thus active MoVTeNb oxide surface.
[1] T. Ushikubo, H. Nakamura, Y. Koyasu, S. Wajiki, US Patent 5 380 933 (1995), to
Mitsubishi Kasei Corporation.
[2] M. Hatano, A. Kayo, US Patent 5 049 692 (1991), to Mitsubishi Kasei Corpora-
tion.
[3] T. Ushikubo, K. Oshima, A. Kayou, M. Hatano, Stud. Surf. Sci. Catal. 112 (1997)
473.
[4] M. Baca, A. Pigamo, J.L. Dubois, J.M.M. Millet, Top. Catal. 23 (2003) 39.
[5] J.M. Oliver, J.M. Lopez Nieto, P. Botella, Catal. Today 96 (2004) 241.
[6] W. Ueda, D. Vitry, T. Katou, Catal. Today 96 (2004) 235.
[7] H. Tsuji, K. Oshima, Y. Koyasu, Chem. Mater. 15 (2003) 2112.
[8] P. DeSanto Jr., D.J. Buttrey, R.K. Grasselli, C.G. Lugmair, A.F. Volpe Jr., B.H. Toby,
T. Vogt, Z. Kristallogr. 219 (2004) 152.
[9] H. Murayama, D. Vitry, W. Ueda, G. Fuchs, M. Anne, J.L. Dubois, Appl. Catal. A
Gen. 318 (2007) 137.
[10] A. Celaya Sanfiz, T.W. Hansen, F. Girgsdies, O. Timpe, E. Rödel, T. Ressler, A.
Trunschke, R. Schlögl, Top. Catal. (2008), doi: 10.1007/s11244-008-9106-z, in
press.
[11] W. Ueda, K. Oshihara, Appl. Catal. A 200 (2000) 135.
[12] K. Oshihara, T. Hisano, W. Ueda, Top. Catal. 15 (2001) 153.
[13] V.V. Guliants, R. Bhandari, R.S. Soman, O. Guerrero-Perez, M.A. Banares, Appl.
Catal. A 274 (2004) 213.
5. Conclusion
[14] R.K. Grasselli, D.J. Buttrey, P. DeSanto Jr., J.D. Burrington, C.G. Lugmair, A.F.
Volpe Jr., T. Weingand, Catal. Today 91–92 (2004) 251.
The model studies presented here suggest that a distinguished
lattice plane of the M1 crystal structure, the (001) plane, is not
solely responsible for its outstanding catalytic activity and se-
lectivity in the partial oxidation of propane to acrylic acid. The
chemical and structural natures of the active ensembles on the
catalyst surface remain unknown. The unique crystal structure
of the M1 phase certainly plays an essential role in the selec-
tive (amm)oxidation of propane, because amorphous material or
MoVTeNb oxide comprising different phases cannot compare with
phase-pure crystalline M1. All experimental experience to date
suggests that this may differ from the situation for related mixed
oxides, such as nanostructured MoVW oxides, which have been
reported to be active in propylene oxidation in a semicrystalline
state [30]. Similar intrinsic reactivity irrespective of the terminat-
ing lattice plane implies that related active ensembles of metal–
oxo clusters are exposed to the reactants on the entire surface
of the M1 needles. The lateral surface of the needles accounts
for the main part of the surface area of M1 (80%). The stepped
morphology of the latter surface may generate similar metal-oxo
arrangements as on the surface of the basal plane of the needles.
HR-TEM studies together with targeted synthesis of M1 showing
different microstructure are currently in progress to verify this as-
sumption.
[15] V.V. Guliants, R. Bhandari, B. Swaminathan, V.K. Vasudevan, H.H. Brongersma,
A. Knoester, A.M. Gaffney, S. Han, J. Phys. Chem. B 109 (2005) 24046.
[16] V.V. Guliants, R. Bhandari, A.R. Hughett, S. Bhatt, B.D. Schuler, H.H. Brongersma,
A. Knoester, A.M. Gaffney, S. Hann, J. Phys. Chem. B 110 (2006) 6129.
[17] J.-P. Jacobs, A. Maltha, J.G.H. Reintjes, J. Drimal, V. Ponec, H.H. Brongersma,
J. Catal. 147 (1994) 294.
[18] H.H. Brongersma, M. Draxler, M. de Ridder, P. Bauer, Surf. Sci. Rep. 62 (2007)
63.
[19] G. Mestl, N.F.D. Verbruggen, H. Knözinger, Langmuir 11 (1995) 3034.
[20] G.J.A. Hellings, H. Ottevanger, S.W. Boelens, C.L.C.M. Knibbeler, H.H. Bron-
gersma, Surf. Sci. 162 (1985) 913.
[21] W.P.A. Jansen, A. Knoester, A.J.H. Maas, P. Schmitt, A. Kytökivi, A.W. Denier van
der Gon, H.H. Brongersma, Surf. Interface Anal. 36 (2004) 1469.
[22] L.C.A. van den Oetelaar, H.E. van Benthem, J.H.J.M. Helwegen, P.J.A. Stapel, H.H.
Brongersma, Surf. Interface Anal. 26 (1998) 537.
[23] M. Baca, A. Pigamo, J.L. Dubois, J.M.M. Millet, Top. Catal. 23 (2003) 39.
[24] R.K. Grasselli, D.J. Buttrey, J.D. Burrington, A. Andersson, J. Holmberg, W. Ueda,
J. Kubo, C.G. Lugmair, A.F. Volpe Jr., Top. Catal. 38 (2006) 7.
[25] R.K. Grasselli, Catal. Today 99 (2005) 23.
[26] R.K. Grasselli, J.D. Burrington, D.J. Buttrey, P. De Santo, C.G. Lugmair, A.F. Volpe
Jr., T. Weingand, Top. Catal. 23 (2003) 5.
[27] J.B. Wagner, O. Timpe, F.A. Hamid, A. Trunschke, U. Wild, D.S. Su, R.K. Widi,
S.B.A. Hamid, R. Schlögl, Top. Catal. 38 (2006) 51.
[28] D. Vitry, Y. Moriwaka, J.L. Dubois, W. Ueda, Appl. Catal. A Gen. 251 (2003) 411.
[29] V.V. Guliants, R. Bhandari, H.H. Brongersma, A. Knoester, A.M. Gaffney, S. Han,
J. Phys. Chem. B 109 (2005) 10234.
[30] H. Hibst, F. Rosowski, G. Cox, Catal. Today 117 (2006) 234.