HREM microstructural studies on the effect of steam exposure and cation promoters on vanadium phosphorus oxides: New correlations with n-butane oxidation reaction chemistry

Article abstract of DOI:10.1021/jp971683g

In situ environmental high-resolution electron microscopy (in situ EHREM) under different gaseous environments and ex situ HREM have been used to directly probe commercially important vanadyl pyrophosphate ((VO)₂P₂O₇) cata

Full text of DOI:10.1021/jp971683g

9916
J. Phys. Chem. B 1997, 101, 9916-9925
HREM Microstructural Studies on the Effect of Steam Exposure and Cation Promoters on
Vanadium Phosphorus Oxides: New Correlations with n-Butane Oxidation Reaction
Chemistry
Pratibha L. Gai,* Kostantinos Kourtakis, D. Robert Coulson, and George C. Sonnichsen
Central Research & DeVelopment, DuPont Science & Engineering Laboratories, Experimental Station,
Wilmington, Delaware 19880-0356
ReceiVed: May 20, 1997; In Final Form: September 30, 1997X
In situ environmental high-resolution electron microscopy (in situ EHREM) under different gaseous
environments and ex situ HREM have been used to directly probe commercially important vanadyl
pyrophosphate ((VO)2P2O7) catalysts used in the oxidation of n-butane to maleic anhydride. These fundamental
studies of catalyst-adsorbate interactions have shown that novel microstructural changes occur in vanadyl
pyrophosphate catalysts when the catalyst is reduced, partially oxidized, or exposed to N2 and steam
environments for extended periods of 100-300 h for the first time. The microstructural changes of vanadyl
pyrophosphates are related to the development of extended misfit glide shear plane defects accommodating
nonstoichiometry. The complex characteristics of long-term effects of water vapor and the effect of cation
promoters of iron and antimony in vanadyl pyrophosphates are explored to improve our understanding of
their effects on the catalysts toward n-butane oxidation. Correlations of microstructural changes with reaction
chemistry suggest that the changes are associated with catalyst reactivity and lattice oxygen diffusion rates.
Introduction
of the P-compound were not described, whereas others describe
steam treatment in the presence of cations such as Bi or other
gases. In all these studies microstructural investigations of these
catalysts18-24 were not reported. Therefore despite several
reports, the effect of water on VPO catalysts over longer periods
is not well understood. In the present study, base (undoped)
VPO catalyst was treated with water vapor at elevated temper-
atures for extended periods of ∼100-300 h, to understand the
changes in the catalyst microstructure and their effect on the
performance and long-term stability. Concurrently, cation
promoter effects on the catalyst composition, microstructure,
and stability in n-butane oxidation were also explored.
Although VPO catalysts have been characterized extensively
by a variety of techniques, these studies are primarily postre-
action examinations of static samples which are not the same
as directly observing dynamic catalysts under operating environ-
ments at elevated temperatures. Therefore considerable uncer-
tainty exists as to the nature of reaction mechanisms, evolution
of surface disorder (defects) due to reaction, redox couples, and
active sites. Recently, a novel in situ environmental cell high-
resolution electron microscope (EHREM) has been developed
that permits direct probing of live gas-catalyst interactions at
elevated temperatures at the atomic level.8 Dynamic transfor-
mation of the precursor VOHPO4‚1/2H2O to VPO has been
carried out directly for the first time in the EHREM.12 The
atomic scale studies provide fundamental evidence concerning
the nature of the topotactic transformation of VHPO to VPO,
with no intermediate amorphous, or other, phases and reveal
that atomic periodicity is maintained throughout the material.
In the present study, dynamic effects of steam and other
environments on the catalyst structure are examined to under-
stand the evolution of stuctural disorder and are compared.
The selective oxidation of n-butane to maleic anhydride (MA)
over vanadium phosphorus oxides is an important commercial
process.1-12 In the selective oxidation of the alkane to MA,
the most active and selective phase is believed to be vanadyl
pyrophosphate, (VO)2P2O7 (hereafter referred to as VPO), with
active centers located at the exposed (010) basal planes.2 There
are also some claims that the active phase is a combination of
VPO and VOPO4 phases, or a P-rich phase, or a mixture of
VPO with unknown or amorphous phases.5-7
In addition, the effects of exposure to water vapor on the
microstructure and performance of VPO remain a relatively
unexplored area. The effect of water vapor on the catalyst
equilibration is of importance since water is one of the dominant
products in the selective oxidation of n-butane. The general
consensus is that the effect of water vapor on VPO catalysts is
quite complex. A detailed search of the literature (including
patent literature) has shown that there are several previous
chemical studies relating to the use of steam, where the reaction
times used were generally short (between 30 min to ap-
proximately a few hours; some slightly longer)18-24 with
conflicting suggestions and conclusions. In some of these
studies,21 butane/air gas feed was bubbled through water and
the moist feed was passed over the catalyst for short reaction
times (less than a few hours), and it was suggested that the
treatment could improve catalytic activity. However because
of the different components in the feed (including the presence
of possible impurities or solid structures), the observed improve-
ment could not be solely attributed to the presence of water.
Other studies22,23 report no direct benefit by the addition of
water/steam, especially to MA selectivity beyond 50% butane
conversion.23 Some patents reveal treatment of VPO with steam
with a “P-compound” where the treatment was carried out
sequentially for a few hours over longer periods24 to increase
the catalyst effectiveness, but effects of the addition or the nature
Experimental Procedures
A variation of an organic method13 was used to synthesize
the catalyst precursors. Vanadium pentoxide and anhydrous
phosphoric acid (prepared by combining P2O5 with aqueous
phosphoric acid) were refluxed in 11:1 isobutyl alcohol/benzyl
* To whom correspondence should be addressed.
X Abstract published in AdVance ACS Abstracts, November 1, 1997.
S1089-5647(97)01683-0 CCC: $14.00 © 1997 American Chemical Society

HREM Microstructural Studies
J. Phys. Chem. B, Vol. 101, No. 48, 1997 9917
alcohol solvent mixtures under nitrogen. Enough phosphoric
acid to satisfy a 1.1/1 P/V atomic ratio was used. Soluble
pentavalent vanadyl species formed during reflux are reduced
and precipitate as a blue tetravalent vanadium precursor. In
the case of pure vanadium phosphorus oxides, VO(HPO4)‚1/
2H2O is precipitated. In the case of doped VPO catalysts,
promoter cations were added as the soluble chlorides and
alkoxides, which coprecipitate with the hemihydrate precursor.
The following calcination procedure was employed to trans-
form the catalyst precursor to the final active phase, which in
the case of pure VPO was identified as the active orthorhombic
(VO)2P2O7 phase, by powder X-ray diffraction. The unit cell
data are given below. Calcinations were performed in a small,
3.5 cm fluidized bed on 60 g samples. Samples were heated to
390 °C in air for 1 h followed by heating at 460 °C for 18 h in
1.5% butane/air. Both HREM and X-ray diffraction (XRD) did
not reveal any residual vanadium hydrogen phosphate hemi-
hydrate precursor in the calcined and activated VPO catalysts.
The structure of (VO)2P2O7 is shown in Figure 1a.14 It is
generally agreed that the structure is orthorhombic,14 with a )
16.594 Å, b ) 7.76 Å, and c ) 9.588 Å (V(Å3) ) 1235) and 1
Å ) 0.1 nm. Pairs of edge-sharing VO6 octahedra are connected
along the b-axis to form double chains of VO6 octahedra sharing
opposite oxygen corners. Pyrophosphate groups link the double
chains into a three-dimensional network.
Studies of Effects of Exposure to Steam and Other
Environments on VPO Catalysts and Correlations with
n-Butane Oxidation Reaction Chemistry. We have probed
dynamic reduction and partial oxidation reactions on the calcined
and activated complex VPO catalysts using our recently
developed novel environmental cell high-resolution electron
microscope (EHREM) under controlled gas environments and
temperatures. For EHREM, a high-resolution 300 kV Philips
CM30 has been extensively modified with the introduction of
a fully integrated and permanently mounted environmental cell
(ECELL) system which is a chemical microreactor placed inside
the EM column.8 The method allows direct (in situ) visualiza-
tion, on the atomic scale, of live catalytic reactions under gas
environments and at elevated temperatures and records dynamic
events in real time (with a low light level TV camera and video
recorder connected to the EHREM). Very low beam currents
and imaging conditions normally used for zeolites15 were used
to eliminate any beam effects, and none were detected. Blank
experiments (without the beam) were also performed on catalyst
samples (with the beam switched on for a few seconds only for
recording) to confirm the in situ data. A conventional chemical
reactor type gas manifold is used to select gas or vapor input
to the ECELL, to control flow rates, and to exchange gases or
to purge and evacuate the reaction chamber. Both in situ
EHREM and parallel chemical reactivity tests were performed
in butane, and alternatively in He, N2, and steam as well as in
partial oxidation environments of 1.5% butane/air mixture, at
temperatures of ∼400 °C, to directly explore nucleation and
development of dynamic catalyst microstructure and its cor-
relation with reaction chemistry. Electron diffraction, electron-
stimulated energy-dispersive X-ray spectroscopy (EDX) for
high-precision composition analysis and high-resolution low-
voltage SEM (LVSEM) were also employed with HREM to
complement the data. Exposures of the catalyst to steam for
extended periods of hundreds of hours were carried out in a
tube furnace to understand the relationship between the micro-
structure, performance, and long-term stability of steam treated
VPO catalysts using the following procedures.
Figure 1. (a) Structural schematic of (VO)2P2O7 (VPO). Pairs of edge-
sharing VO6 octahedra are connected along the b-axis, with pyrophos-
phate groups linking the chains of VO6 octahedra. (b) Environmental
high-resolution electron microscopy (EHREM) atomic structure of (010)
VPO in butane environment, with the associated electron diffraction
(ED), inset (RT), showing an ordered structure.
the (VO)2P2O7 phase. Experiments were performed in a 3 in.
fritted quartz tube placed in a vertical tube furnace. Steam was
generated by using an Isco syringe pump to feed water into a
vaporizer. Following each experiment, the tube was purged with
nitrogen for 20 min and quenched to room temperature (5-10
min) under nitrogen flow. Surface area measurements were
carried out according to the BET method, using N2 at -195
°C.
Results
Microstructural data and reaction chemistry of the catalysts
following treatments in the various environments are described
and compared in the following paragraphs. The EHREM atomic
structure of (010) VPO in butane at room temperature (RT) is
shown here in Figure 1b, with the electron diffraction (ED).
The dynamic reduction of the catalysts in butane at ∼390-400
°C showed two types of extended structural defects at ∼80° to
each other (e.g. dark microstructural features shown at A and
The steam treatment experiments were initially performed at
∼360 °C on VHPO which was calcined and activated to form

9918 J. Phys. Chem. B, Vol. 101, No. 48, 1997
Gai et al.
Figure 2. EHREM in situ direct probing of butane molecules-VPO catalyst interactions at ∼400 °C, elucidating defect formation along 201
directions and growth in VPO after a few hours: (a) the two sets of symmetry-related extended defects (dark microstructural features arrowed) in
diffraction contrast; (b) the same defects in high resolution; (c) the associated ED showing streaks (indicative of disorder, arrows) around the Bragg
reflections, along the 201 directions; (d) the structural model for the extended defects deduced by detailed analyses of the in situ EHREM studies.
The defects are glide shear plane defects formed by a novel glide shear mechanism and accommodate the misfit between the reduced surface layers
of the catalyst containing anion vacancies and the underlying VPO;30 vacant anion sites are indicated by filled circles; (e) in situ EHREM studies
of the VPO catalyst in 1.5% butane/air environment, showing the presence of smaller concentrations of defects (imaged in a 201 reflection, arrowed),
indicating a slight reduction of the catalyst surface. The figure shows disordered VPO.
TABLE 1: Steam Treatment of VPO Catalyst
sample no. % steam temperature, °C time, h surface area, m2/g
B in Figure 1b), in an otherwise clean VPO matrix. They are
along the two symmetry-related 201 directions (which are
∼80° to one another). The defects were observed to nucleate
1
2
3
4
31
31
42
0
457
355
355-362
355-362
24
92
312
312
14
25
23
n/a
30
near the surface and extend into the bulk as a function of
reduction time. Figure 2a,b shows the defects after a few hours,
in diffraction contrast and in high resolution, respectively. The
associated ED shows streaks around the Bragg reflections and
are along 201 (c). Under the reducing conditions, lattice
oxygen loss due to the catalyst reduction leads to the formation
of anion vacancies. Detailed analysis of the nature of these
complex structural defects has been reported elsewhere11 and
has shown that the extended defects are (201) misfit glide shear
plane defects, formed by a novel glide shear mechanism to
accommodate the shape misfit between the reduced surface
layers of VPO containing anion vacancies and the underlying
bulk VPO. Novel misfit glide shear defect process in operating
catalytic oxides was first discovered by Gai16,17 (Gai’s mech-
anism, which now appears to be a fundamental, effective, and
pivotal process in selective oxidation catalysis by oxides
generally for accommodating nonstoichiometry11 and this misfit
glide shear mechanism, based on glide defects relieving misfit
strains between catalyst surface layers containing anion vacan-
cies and the underlying unreduced bulk oxide, preserves active
Lewis acid sites at the catalyst surface for catalysis). The
formation of these defects was also observed by in situ EHREM
in N2 and He, which are mildly reducing environments.
5 (VO)2P2O7
regions which maintain the anion vacancies (V3+ phases) do
not change the overall VPO structure under the operating
conditions, and therefore these local changes are not clearly
visible in X-ray diffraction. Electron diffraction and electron
microscopy (real-space crystallography) are essential to observe
these reaction mechanisms directly. Following in situ studies
of the catalyst in 1.5% butane/air, the defects are still observed
in some areas (Figure 2e, in a (201) reflection), suggesting that
a slight surface reduction of the catalyst occurs even in the
presence of gas oxygen. The defects however are in signifi-
cantly lower concentrations compared to the catalysts reduced
with n-butane. This may be due to the reoxidation of many of
the defects upon exposure to cofed oxygen in the gas mixture.
These structural defects were also observed following exposure
of the catalyst to steam in in situ experiments. The effect of
steam on the catalyst is described in the following section.
Table 1 summarizes the conditions for the steam treatment
and surface area data. A parallel heating experiment in nitrogen
for 312 h (sample 4) served as a control sample for the 312 h
steam calcination to determine the effect of water vapor during
the heating procedure.
The anion vacancies lie in the basal plane between corner-
sharing vanadyl octahedra and phosphate tetrahedra. The defect
regions thus correspond to a slightly anion-deficient defect-
pyrophosphate phase, the structure of which is shown in Figure
2d, with pyrophosphate groups sharing corners similar to
metaphosphate (PO3)n groups. These local glide shear defect
In powder X-ray diffraction (XRD) data of samples treated
for shorter time periods (sample 2) the effects of steam treatment
were less noticeable, which may relate to the slow crystallization

HREM Microstructural Studies
J. Phys. Chem. B, Vol. 101, No. 48, 1997 9919
Figure 3. XRD of samples 3 and 4.
TABLE 2: Continuous Fixed Bed Microreactor
Measurements
TABLE 3: Pulsed Microreactor Data for Steam-Treated
Vanadium Phosphate
1.5% butane/air
(360 °C)
% selectivity at
after 20 butane pulses
sample
k, s-1
conversion
MA sel.
1.5% butane/N2
2
3
4
2.65
2.05
3.68
5.98
16
11
21
20
21
9
33
33
k,
60% maleic capacity,
sample
k, s-1a
2.75
3.07
3.35
g/(m2 s) 20% 40% conv. mmol/g catalyst
5 (base VPO)
2
3
4
5
0.110
0.134
78
79
77
82
75
78
77
80
71
73
74
77
2.22
1.23
6.06
4.95 ( 0.58
marked effects are observable. The maleic anhydride capacities
measured in 1.5% butane/N2 also are diminished for the steam-
treated catalysts and correlate with reactor data described in the
following section.
3.39 ( 0.22 0.113
a The rate constants have been normalized to T ) 360 °C assuming
an activation energy of 25 kcal/mol.
Pulsed Microreactor Measurements. Table 3 summarizes
results for the evaluation of these catalysts in a pulse microre-
actor at 380 °C. The pulse microreactor evaluations were carried
out by injecting 0.05 mL pulses of n-butane using a gas sampling
valve contained in an oven at 170 °C into a stream of helium
flowing over 0.5 g of catalyst in a reactor heated to 380 °C.
The effluent of the reactor passes through a thermal conductivity
detector and then through a sample loop. When the pulse was
in the sample loop, as determined by the thermal conductivity
detector, it was injected into a gas chromatograph for analysis
of the reaction products. The rate constants are obtained by
introducing pulses of butane over the catalyst at three different
flow rates, while the catalyst is reoxidized after each butane
pulse with eight pulses of air at a constant flow rate. The
behavior of the surface-reduced catalyst is determined by
introducing successive butane pulses at 10 min intervals over
the catalyst without reoxidation. Butane conversion and
selectivity to maleic anhydride decrease after each pulse.
n-Butane conversion and selectivity to maleic anhydride of the
catalyst as it is reduced with successive butane pulses is shown
in Figure 4a,b. The oxygen removed, or the degree of total
catalyst reduction, can be calculated from n-butane conversion
and selectivities to various oxdation products. The oxygen
removed at each n-butane pulse is tabulated in units of
micromoles oxygen reacted per gram of catalyst. Initial
vanadium oxidation states were measured by redox titrimetry
using a procedure similar to that of Nakamura et al.25 The initial
of the catalyst in the presence of water vapor at elevated
temperatures. In samples reacted for extended periods in steam
a modest decrease in the diffraction line widths was apparent
(e.g. in sample 3), qualitatively correlating with the surface area
decrease measured by N2 BET. Figure 3a,b shows examples
of XRD data for samples 3 and 4, respectively.
Rate constants were measured in a microreactor on 1 g of
catalyst at temperatures between 358 and 363 °C, in 1.5%
butane/air. The rate constants represent a pseudo-first-order rate
constant for the disappearance of butane. Catalyst bed volumes
were used in these calculations. Changes in bed porosity should
be minor and should not change the relative rate constants for
comparisons between catalysts. Where possible, the intrinsic
rate constants are also tabulated based on N2 BET surface area
measurements. “Maleic anhydride capacities” are also tabulated,
and represent the total amount of maleic anhydride liberated
by reduction of each sample with a constant flow of 1.5%
butane/N2 at 360 °C. Prior to this reduction, the sample had
been treated in air at 360 °C for 30 min.
A comparison of the apparent first-order rate constants
tabulated above reveals that the activity of the VPO catalyst
(sample 5) and the control VPO sample calcined in nitrogen
for 312 h (sample 4), are nearly equal. Both steam-treated
catalysts show diminished activities in 1.5% butane/air, but their
intrinsic rate constants based on the BET surface area measure-
ments are essentially equal. In the absence of cofed oxygen,

9920 J. Phys. Chem. B, Vol. 101, No. 48, 1997
Gai et al.
Figure 4. (a) Conversion; (b) selectivity to maleic anhydride, for successive n-butane pulses.
vanadium oxidation states of both steam-treated and nitrogen-
calcined catalysts, as determined by this procedure, are 4.00 (
0.01. In Figure 4a,b, a calculated value of 82 mmol oxygen
(O) reacted/g catalyst is equivalent to a vanadium oxidation state
change of 0.025 (3.975 ( 0.01).
Rate constant data obtained from this reactor system indicate
a loss in activity for the nitrogen-calcined catalysts (sample 4)
and for catalysts heated in water vapor (samples 2 and 3).
Significant differences in activity between sample 4 (nitrogen-
calcined VPO) and sample 5 are observed from these experi-
ments. It is interesting that differences in catalyst activity
between these samples are not observed from microreactor
evaluations performed in 1.5% butane/air, described in Table
2. The lack of cofed oxygen during the pulse reactor rate
constant evaluations should lead to catalyst reduction during
the measurements and could account for these activity differ-
ences.
Marked differences in catalytic behavior upon reduction are
more dramatically indicated from the pulse reactor during
catalyst reduction with consecutive pulses of n-butane. At the
20th pulse of n-butane, sample 3, which was treated in an
atmosphere of 42% steam for 312 h, is significantly less active
and selective than VPO (sample 5) or VPO calcined in nitrogen
(sample 4). These results qualititatively agree with continuous
fixed bed evaluations (described in the previous section) in 1.5%
butane/nitrogen.

HREM Microstructural Studies
J. Phys. Chem. B, Vol. 101, No. 48, 1997 9921
This is further demonstrated in Figures 4a,b, in which the
micromoles oxygen reacted per gram of catalyst is plotted
against n-butane conversion and selectivity to maleic anhydride.
(Plots of surface area against n-butane conversion and selectivity
to MA were identical.) It is clear that important differences in
both conversion and selectivity are observed. Samples 2 (92
h, steam) and 3 (312 h, steam) are clearly different from the
base “VPO” catalyst, sample 5 ((VO)2P2O7) and the control
VPO catalyst calcined for 312 h in nitrogen. Both steam-treated
samples (samples 2 and 3) are less active and selective to maleic
anhydride than samples 3 and 4 as they are reduced, indicating
that the active sites of these have been depleted of oxygen. The
plateau observed from these reduction graphs reflects a depletion
of reactive surface oxygen, so that the reaction is limited by
lattice oxygen diffusion. The lower plateaus observed for both
steam-treated catalysts suggest lower oxygen anion diffusion
rates for the steam-treated catalysts compared with the VPO
control samples. The relative positions of the plateaus reached
with these curves should not be dependent on small changes in
the initial oxidation states of the catalyst samples.
Microstructures of Steam-Calcined VPO and Correlation
with Reaction Chemistry. Electron microscopy investigations
have revealed that steam plays an important role in the
development of the disordered VPO structure. The key
observation from in situ electron microscopy experiments in
steam at ∼390 °C is the evolution of extended glide shear
defects of the type we have observed in vanadyl pyrophosphate
(VPO) reduced in butane in {201} orientations. The defect
concentration was found to be higher in steam than in samples
reacted in butane and N2 under similar conditions. This novel
result is discussed in the following sections.
Comparisons of the control sample and steam-treated samples
from the reactor studies indicate that the steam/VPO samples
contain a higher concentration of the defects, confirming in situ
observations that steam accelerates their formation. The defect
concentration in steam also increases considerably as a function
of time: for example, steam/VPO samples reacted for 312 h
have a higher defect density than the sample reacted for 92 h.
In the steam/VPO samples steamed for 24 and 92 h, streaking
(disorder) from the defect regions is found to be predominant,
suggesting that long-range order has not yet been achieved. By
measuring the number of defects per formula unit (in Figure
2b), the local oxidation state of V in these defective regions
may be estimated to be ∼3.7-3.8 (i.e. close to 4+), dispersed
in a VPO matrix with V4+, if the V3+ states are not terminated
by hydroxyl groups. The percentage of misfit glide shear defects
observed, the anion deficiency, and the averaged oxidation state
(of the defects and bulk) changes fall in the range of 0.02
valence units. Vanadium oxidation state measurements, by
redox titrimetry which give an overall oxidation state, are not
capable of distinguishing changes in vanadium oxidation states
smaller than these valence units. However, it is also possible
that although the misfit glide shear defects form by a redox
mechanism, anion vacancies and V3+ states are oxidized to V4+
and terminate with hydroxyl species.
Figure 5. Steam interactions with VPO catalyst: (a) low-voltage SEM
(LVSEM) of base VPO; (b) LVSEM of the steamed-VPO catalyst with
microcrystals (plates, arrows); and (c) HREM of VPO crystals covered
with the extended glide shear defects and porosity (arrows).
in some areas rosettes were decomposed into microcrystals,
resulting in porosity. Figure 5 summarizes microstructural
observations in steam: part a shows LVSEM of base VPO; part
b shows LVSEM images of steamed catalyst with microcrystals/
plates (arrows); part c shows a HREM image of crystals covered
with glide shear defects and porosity (arrows) in steam. The
inset shows a large concentration of glide defects at lower
magnification. Furthermore, the study of the VPO sample
heated in N2 has shown the presence of glide shear defects, but
in lower concentrations compared to those treated in steam under
similar conditions of temperature and time. Electron diffraction
showed diffuse streaking due to the defects (similar to that
shown in butane in Figure 2c), suggesting disorder.
In the steam/VPO samples reacted for 312 h, platy formation
of crystallites is indeed visible in some areas of the samples
along the 201 defect directions: this suggests that in a
prolonged exposure long-range order (leading to a new, slightly
anion-deficient phase) is essentially reached in those crystals.
The misfit glide shear is a mechanism for these larger micro-
structural changes which occur with time in the catalyst and
may be important in the long-term stability. Since such plates
are also formed in prolonged reduction in butane,11 they appear
to be anion-deficient structures (VPO4.33). LVSEM showed that

9922 J. Phys. Chem. B, Vol. 101, No. 48, 1997
Gai et al.
Figure 6. High-resolution and analytical electron microscopy of promoted (V0.9Fe0.05Sb0.05O)2P2O7 catalyst ((Fe,Sb)VPO): (a) surface morphology
of base VPO; (b) morphology of the (Fe,Sb) VPO with smoother structure; (c) electron-stimulated high-precision X-ray spectroscopy in the electron
microscope (EDX) for microchemical analysis of the promoted catalyst showing the presence of Fe and Sb in solid solution; (d) the spectral area
at higher magnification; (e) HREM showing a well-ordered single phase formation of (Fe,Sb)VPO; there is no evidence of amorphous or secondary
phases on the surface of the catalyst crystals; (f) the corresponding electron diffraction.
Discussion
lie in the basal plane. The slight reduction of the catalyst surface
leads to restructuring via the formation of (201) glide shear
defects which accommodate the misfit strains between the
reduced layers containing anion point defects and the underlying
bulk VPO. These defect sites may be important in oxidation
and anion diffusion. Under partial oxidation conditions, the
novel glide shear process provides an effective mechanism to
accommodate small deviations in stoichiometry, without the
Our direct in situ electron microscopy observations of steam,
butane, and butane/air interactions with the catalyst and analyses
of the steam-calcined and control experiments indicate that in
all these environments the VPO crystal surface is slightly altered
and the crystal is continuously changing to the (201) lower
energy stable configuration. The studies show that during the
operation of the catalyst some anion vacancies are formed which

HREM Microstructural Studies
J. Phys. Chem. B, Vol. 101, No. 48, 1997 9923
Figure 7. (a) Pulse reactor data for promoted catalyst: n-butane conversion for successive butane pulses compared with those of base VPO. (b)
Selectivity to maleic anhydride for successive butane pulses.
collapse of the VPO lattice. Only the local symmetry is altered
at the defect boundary between the surface and bulk VPO due
to the crystal glide. The microstructural and reactivity data
suggest that steam especially enhances the extent of reduction
and catalyst reconstruction (i.e. the formation and growth of
the glide shear defects leading to crystallite growth). The
reasons for these are unclear to us at present, but we propose
the following speculations for the observed microstructure and
catalyst performance upon long exposures of VPO catalysts to
steam.
As described in the preceding sections, we have observed
increased structural defects upon exposure to steam as shown
in Figure 5. Over extended periods the microstructural changes
leading to crystallite growth would reduce the active VPO
surface content and manifest in the observed decrease in the
catalyst performance affecting the useful life of the catalyst. It
is possible that in the presence of water some of the catalyst
surface sites are occupied by hydroxyl groups. This hydroxyl-
ated surface might inhibit further chemisorption of oxygen and
reduce sites capable of oxidizing the alkane, resulting in a
decrease in the rate of butane oxidation. These possibilities
are discussed below.
Microreactor data show that prolonged steam treatment of
VPO results in a significant loss in acitivity and selectivity as
the catalyst is reduced in n-butane; the activity decline is not
apparent during evaluations in 1.5% butane/air. This implies
that steam treatment has reduced the population of sites capable
of oxidizing butane to maleic anhydride in the absence of gas-
phase oxygen during catalyst reduction. This is accompanied
by a loss of N2 BET surface area and a decrease in XRD line
width, which together are consistent with electron microscopy
data indicating crystallite growth and an increase in the extended

9924 J. Phys. Chem. B, Vol. 101, No. 48, 1997
Gai et al.
Figure 8. Butane reaction for a few hours: (a) (Fe,Sb)VPO; (b) base VPO.
glide shear defects. In the absence of cofed oxygen, it is
possible that the chemisorbed water prevents chemisorption of
any trace oxygen in the butane/N2 stream, inhibiting the partial
reoxidation of the catalyst. It may also be due to the fact that
the redox and recontruction of the catalyst depend upon
hydroxylation state of the catalyst surface. The loss in activity
and selectivity of the catalysts during prolonged reduction
indicates a loss in catalyst lattice oxygen carrying capacity in
the absence of cofed oxygen, which is a consequence of
decreased lattice oxygen diffusion rates. The oxygen diffusion
rates are most diminished for steam-treated catalysts and roughly
correlate with the formation and increased number of glide shear
defects extending throughout the sample in prolonged treat-
ments. In prolonged reduction with n-butane in the absence of
cofed oxygen, the large number of misfit extended glide shear
defects, extending throughout the sample, tend to order into new
plate crystals which correspond to a new, slightly anion-deficient
phase, and may interfere with oxygen anion transport to catalyst
surfaces.
Although glide shear defects are present in all the environ-
ments to a varying degree, the extent of ordering (condensation
into a new phase) of a large number of glide defects especially
in the absence of cofed oxygen may be important in understand-
ing the loss of activity and selectivity in steam compared to the
samples reacted in n-butane and nitrogen. Steamed samples
have more plates due to ordering of a larger number of the
defects. Comparison of the activity data of steam-treated
samples with those of the N2-treated and unmodified VPO
samples suggests that a small number of the disordered glide
shear defects is sufficient to accommodate slight stoichiometric
deviations in selective oxidation reactions and for facilitating
operation of the catalyst. Inhibiting irreversible reduction of
the sites capable of maleic activity by controlling the concentra-
tion and ordering of the defects is largely accomplished in the
presence cofed oxygen in 1.5% butane/air.
Effects of Cation Promoters on the VPO Catalyst System
in n-Butane Oxidation Chemistry: (V0.9Fe0.05Sb0.05O)2P2O7.
The role of cation promoters on the microstructure and stability
of VPO catalysts in n-butane catalysis is of major importance.
Although several studies have been cited in the literature26,27
on the specific activity and MA selectivity of promoted VPO
catalysts, no direct evidence for the formation of solid solutions
with VPO has been reported. Understanding of the location
and of the complex role of promoters is therefore lacking. We
have synthesized a novel (Fe,Sb) promoted catalyst system28
and have probed the structure and chemistry of the catalyst by
atomic scale HREM imaging in conjunction with composition
analyses. The extent of solid solution and the promoter effect
on the catalyst microstructure and stability in n-butane oxidation
are examined in the following sections. Chemical composition
spectra calibrated with standards have been obtained in the
electron microscope using electron-stimulated energy-dispersive
X-ray (EDX) spectroscopy from a large number of crystals.
Our studies elucidate that iron and antimony are present in
solid solution in vanadyl pyrophosphate, indicating the formation
of a new promoted phase isostructural with the undoped VPO.
These results are summarized in Figure 6: part a shows surface
morphology of base VPO; surface morphology of (Fe,Sb) VPO
shown in Figure 6b appears to be smoother than VPO (Figure
6a); part c shows a high spatial resolution (2 nm) EDX
microanalysis spectrum from the catalyst, indicating that Fe and
Sb are present throughout in the catalyst forming a novel
composition, isostructural with (VO)2P2O7 (peaks from the Cu
sample holder are also visible); part d shows the magnified
spectral region showing Fe and Sb peaks more clearly; and part
e shows a HREM image indicating a well-ordered promoted
catalyst with the associated electron diffraction pattern (f). There
is no evidence for amorphous or additional phases on the surface
of the catalyst crystallites. The new composition can therefore
be described as (V0.9Fe0.05Sb0.05O)2P2O7.

HREM Microstructural Studies
J. Phys. Chem. B, Vol. 101, No. 48, 1997 9925
Pulse reactor data of the catalyst system are shown in Figure
7a,b. It is clear that compared to pure (VO)2P2O7, both
sites, resulting in lattice collapse, are secondary to cataly-
sis16,17,29,31 and glide shear is the most effective defect mech-
anism in accommodating nonstoichiometry and facilitating
continued operation of oxide catalysts in selective oxidation
reactions.
conversion and selectivity are improved for (V0.95Fe0.05
-
Sb0.05O)2P2O7 as oxygen is removed from the catalyst. In both
cases, the catalyst is exposed to consecutive n-butane pulses
and is not reoxidized between pulses. As observed for the
steam-treated catalysts, once surface oxygen is removed by
reduction with n-butane, the reaction becomes limited by lattice
oxygen diffusion and is indicated by plateaus in both conversion
and selectivity for both catalyst systems during further reduction.
However, the higher conversion and selectivities noted in the
diffusion-limited regime which are observed to the right of
Figure 7a,b clearly show that the rate of anion diffusion to the
surface of (V0.9Fe0.05Sb0.05O)2P2O7 is enhanced compared to
VPO. One explanation for the enhanced oxygen diffusion rates
may be that cooperative behavior between the FeII/FeIII and
Sb(III)/Sb(V) (where the promoter cations may be pinning anion
vacancies), anion defects, and vanadium redox couples is
important in the diffusion.
Acknowledgment. P.L.G. thanks C. R. A. Catlow and A.
K. Cheetham for the kind invitation to prepare this article. She
continues to enjoy a very fruitful collaboration with Sir John
Meurig Thomas on structural chemistry and catalysis of novel
zeolites. She and her colleagues thank H. S. Horowitz and E.
D. Boyes for technical discussions and L. G. Hanna, D. L.
Smith, J. J. Borowski, D. Cline, Jr., M. Harrison, T. Flecken-
stein, F. G. Gooding, and E. Jones for technical assistance.
References and Notes
(1) Hodnett, B. K. Catal. ReV.-Sci. Eng. 1985, 27, 323.
(2) Bordes, E. Catal. Today 1987, 1, 499.
(3) Centi, G. (ed. sp. vol.) Catal. Today 1993, 16, 1.
(4) Horowitz, H. S.; et al. Appl. Catal. 1988, 38, 193.
(5) Adelouahabab, F.; Volta, J. C.; et al. J. Catal. 1991, 128, 248.
(6) Koyano, G.; Okuhara, T.; Misono, M. Catal. Lett. 1995, 32, 205.
(7) Kiely, C.; Sajip, S.; Ellison, I.; Santanes, M.; Hutchings, G. J.; Volta,
J. C. Catal. Lett. 1995, 33, 357.
(8) Boyes, E. D.; Gai, P. L. Ultramicroscopy 1997, 67, 219. In Situ
Microscopy in Materials Research; Gai, P. L., Eds.; Kluwer Academic
Publishers: Dordrecht, 1997.
(9) Gai, P. L.; Kourtakis, K. Science 1995, 267, 661.
(10) Gai, P. L.; Kourtakis, K.; Coulson, R.; Sonnichsen, G. In Catalyst
Materials for High-Temperature Processes; Ramesh, K. S., Misono, M.,
Gai, P. L., Eds.; American Ceramic Soc.: Westerville, OH, 1997; Vol. 73.
(11) Gai, P. L. Acta Crystallogr. 1997, B53, 346.
(12) Gai, P. L.; Torardi, C. C. In In Situ Electron & Tunneling
Microscopies; Sharma, R., Gai, P. L., Gajdarziska, M., Sinclair, R.,
Whitman, L., Eds.; Materials Research Society: Warrendale, PA, 1996;
Vol. 404.
Additionally, our initial HREM studies on the promoted
catalysts cycled between n-butane and air for a few hours have
indicated that the microstructural changes due to the develop-
ment of a large number of extended glide defects are less marked
in these catalysts relative to pure (VO)2P2O7 (Figure 8a,b). This
suggests that the promoter cations may be pinning the anion
vacancies (or oxidize them) and inhibiting the growth and
ordering of a large number of the defects into new platy
structures and thus may act to maintain the optimal anion defect
concentration and surface oxidation state of the vanadium. With
the present HREM methods it is not possible to image isolated
anion point defects. Since the development and ordering of
the defects appear to be symptomatic of further larger micro-
structural changes as a function of time, the presence of these
promoters may help to stabilize the pyrophosphate structure.
(13) Katsumoto, K.; Marquis M. Chevon Co. U.S. Parent 4,132,670.
(14) Gordunova, Yu. E.; Linde, S. SoV. Phys.-Dokl. (Engl. Trans.) 1979,
24, 138.
(15) Gai, P. L.; Thomas, J. M.; Wright, P.; Jones, R.; Natarajan, S.;
Chen, J.; Xu, R. J. Phys. Chem. 1992, 96, 8206.
Conclusions
The in situ EHREM studies coupled with reaction chemistry
have provided insights into the nature of dynamic reaction
mechanisms of VPO catalysts in n-butane, butane/air, and steam
environments. The slight reduction of the catalyst in n-butane
oxidation leads to restructuring via the formation of (201) glide
shear defects, which accommodate the resulting nonstoichiom-
etry. Effects of long-term exposure to steam on VPO catalysts
show a marked increase in the number of glide defects which
order into new anion-deficient plate crystals. Cation-promoted
(16) Gai, P. L. Philos. Mag. A 1981, 43, 841.
(17) Gai, P. L. Catal. ReV.-Sci. Eng. 1992, 34, 1.
(18) Tomita, S.; Kawakami, T.; Suwa, H.; Ihara, T., Mitsubishi Chem.
Ind. Japan Patent JP 05043567 A2 930223, 1992.
(19) Ebner, J.; Andrews,W., Monsanto Co. U.S. US 5137860 A 920811.
(20) Ootake, M.; T. Kawakami, T.; Otake, M., Mitusbishi Kasei Corp.
Japan Patent JP 01201016 A2 890814, 1988.
(21) Arnold, E. W.; Sundaresan, S. Appl. Catal. 1988, 41, 225.
(22) Cornaglia, A.; et al. Appl. Catal. 1991, 74, 15.
(23) Lerou J. J.; Mills, P. L. Proceedings Process Techn. International
Conference; Weijenen, M., et al., Eds.; Kluwer Academic Publ.: Dordrecht,
1993; p 175.
vanadyl pyrophosphate with a nominal composition (V0.9Fe0.05
-
(24) Click, G.; Barone, B., Denka Chem. Corp. U.S. 4515899 A 850507,
1985.
Sb0.05O)2P2O7 has been synthesized and compared with undoped
VPO in the n-butane oxidation. Pulse reactor studies have
indicated higher lattice oxygen carrying capacities in the
promoted catalyst. Microstructural studies indicate that the iron
and antimony might maintain optimal anion point defect
concentration.
The in situ EHREM confirms our earlier results on simple
and mixed-metal oxide catalysts that crystallographic shear
planes formed during catalytic reactions which eliminate vacant
(25) Nakamura, M.; Kawai, K.; Fujiwara, Y. J. Catal. 1974, 34, 345.
(26) Bej, S. K.; Rao, M. S. Appl. Catal. 1992, A83, 149.
(27) Hutchings, G. J.; Higgins, R. J. Catal. 1996, 162, 153.
(28) Kourtakis K.; Sonnichsen, G., U.S. Patent 543532. Kourtakis, K.;
Sonnichsen, G.; Coulson, D. R.; Gai, P. L. Manuscript in preparation.
(29) Gai, P. L.; et al. Nature 1990, 348, 430.
(30) Notes: The studies11
have shown that the displacement vector is
approximately parallel to the trace of (201) planes (lies in the plane of shear)
and to the defect directions.
(31) Gai, P. L. J. Solid State Chem. 1993, 104, 119.