In situ studies of atomic, nano- and macroscale order during VOHPO4·0.5H2O transformation to (VO)2P2O7

Article abstract of DOI:10.1016/S1381-1169(01)00162-5

Transformation of VOHPO₄·0.5H₂O precursor to well-crystallized (VO)₂P₂O₇, for n-butane oxidation to maleic anhydride was studied by in situ Raman and XRD techniques. Atomic scale changes observed in the precursor structure at 583 K provided new insights into its transformation to (VO)₂P₂O₇. In addition to (VO)₂P₂O₇, nanocrystalline oxidized δ-VOPO₄ invisible to XRD was detected during transformation in n-butane/air, possibly due to the specificity of the in situ conditions. Under catalytic reaction conditions the disordered nanocrystalline (VO)₂P₂O₇ in the fresh catalysts (ca. 10-20 nm domains) gradually transformed into well-crystallized (VO)₂P₂O₇ in the equilibrated VPO catalysts (>30 nm domains) with time on stream. Simultaneously, a disordered layer ca. 2 nm thick which was covering the surface (100) planes of (VO)₂P₂O₇ disappeared yielding a solid with high steady-state catalytic performance. Only (VO)₂P₂O₇ was observed both at room temperature and reaction temperature in the equilibrated VPO catalysts. Specific surface termination of the (100) planes of (VO)₂P₂O₇ in the equilibrated VPO catalysts is believed to be responsible for high activity and selectivity of these catalysts for maleic anhydride formation.

Full text of DOI:10.1016/S1381-1169(01)00162-5

Journal of Molecular Catalysis A: Chemical 172 (2001) 265–276
In situ studies of atomic, nano- and macroscale order during
VOHPO4·0.5H2O transformation to (VO)2P2O7
V.V. Guliantsa,∗, S.A. Holmesa, J.B. Benzigerb, P. Heaneyc, D. Yatesc, I.E. Wachsd
a
Department of Chemical Engineering, University of Cincinnati, Cincinnati, OH 45221-0171, USA
b
Department of Chemical Engineering, Princeton University, Princeton, NJ 08544, USA
c
Department of Geology, Princeton University, Princeton, NJ 08544, USA
d
Department of Chemical Engineering, Zettlemoyer Center for Surface Studies, Lehigh University, Bethlehem, PA 18015, USA
Received 23 November 2000; accepted 14 March 2001
Abstract
Transformation of VOHPO4·0.5H2O precursor to well-crystallized (VO)2P2O7, for n-butane oxidation to maleic anhydride
was studied by in situ Raman and XRD techniques. Atomic scale changes observed in the precursor structure at 583 K
provided new insights into its transformation to (VO)2P2O7. In addition to (VO)2P2O7, nanocrystalline oxidized ␦-VOPO4
invisible to XRD was detected during transformation in n-butane/air, possibly due to the specificity of the in situ conditions.
Under catalytic reaction conditions the disordered nanocrystalline (VO)2P2O7 in the fresh catalysts (ca. 10–20 nm domains)
gradually transformed into well-crystallized (VO)2P2O7 in the equilibrated VPO catalysts (>30 nm domains) with time on
stream. Simultaneously, a disordered layer ca. 2 nm thick which was covering the surface (1 0 0) planes of (VO)2P2O7
disappeared yielding a solid with high steady-state catalytic performance. Only (VO)2P2O7 was observed both at room
temperature and reaction temperature in the equilibrated VPO catalysts. Specific surface termination of the (1 0 0) planes of
(VO)2P2O7 in the equilibrated VPO catalysts is believed to be responsible for high activity and selectivity of these catalysts
for maleic anhydride formation. © 2001 Elsevier Science B.V. All rights reserved.
Keywords: Vanadium pyrophosphate; In situ studies; Raman; XRD; Rietveld refinement
1. Introduction
and (VO)2P2O7 have been recently elucidated in
several studies [5–10]. The VOHPO4·0.5H2O and
(VO)2P2O7 phases possess similar crystal struc-
Vanadyl(IV) hydrogen phosphate hemihydrate,
VOHPO4·0.5H2O, is a precursor of the commercial
vanadium-phosphorus-oxide (VPO) catalysts for par-
tial oxidation of n-butane to maleic anhydride [1].
Vanadyl pyrophosphate, (VO)2P2O7, obtained by the
thermal transformation of the precursor phase has
been identified as critical for the activity and selectiv-
ity of n-butane oxidation over the VPO catalysts [1–5].
The single crystal structures of VOHPO4·0.5H2O
tures: the [VOHPO4] layers are hydrogen-bonded
2−
via HPO4
groups in the precursor phase [5,6]
and become covalently bonded via pyrophosphate
(P2O74−) groups during the thermal transformation to
vanadyl pyrophosphate [7–10]. On the macroscopic
scale, the crystal morphology is preserved during
this transformation. Several models for this so-called
topotactic transformation of the precursor phase into
vanadyl pyrophosphate have been previously proposed
[11–14]. According to more recent models [11,12],
the HPO4 groups present in the precursor structure
∗ Corresponding author. Fax: +1-513-556-3473.
E-mail address: vguliant@alpha.che.uc.edu (V.V. Guliants).
1381-1169/01/$ – see front matter © 2001 Elsevier Science B.V. All rights reserved.
PII: S1381-1169(01)00162-5

266
V.V. Guliants et al. / Journal of Molecular Catalysis A: Chemical 172 (2001) 265–276
are inadequately oriented to form the regular P–O–P
linkages observed in the single crystal structure of
(VO)2P2O7 [7–10]. These recent transformation mod-
els involve either the concerted displacement of the
hydrogen phosphate layers [12] or the inversion of the
phosphate tetrahedra in the precursor phase to align
the hydrogen phosphate groups sufficiently to form
the pyrophosphate groups [11]. However, the former
model [11] described a different P–O–P connectivity
than the one established via single crystal diffraction
studies [7–10]. On the other hand, the latter model [12]
cannot account for several experimental observations:
(i) the formation of a disordered intermediate during
this transformation, which coexists with neither the
starting nor product phases; (ii) the significant crystal-
lographic disorder on the nanoscale; and (iii) the pres-
ence of extended defects in vanadyl pyrophosphate
transformed under catalytic reaction conditions. These
experimental observations indicate that the transfor-
mation process is more complex than suggested by
the proposed models, and that the crystal structure
of the catalytically active vanadyl pyrophosphate
obtained at low temperatures (653–723 K) is much
more disordered than the structure of single crystal
vanadyl pyrophosphate obtained from the melt above
1173 K.
Hiroi et al. [15] suggested that there were in actual-
ity two possible symmetries available for (VO)2P2O7,
depending on preparation conditions. Electron, X-ray
and neutron diffraction studies on polycrystalline
samples, prepared from the melt at 1023 K indicated
the presence of solely orthorhombic material, space
group Pca21. In this paper, defects were suggested
to be attributable to the positional disorder of P2O7
groups, and that two distinct crystalline forms of
(VO)2P2O7 coexist with the same symmetry, or that
one exists as a defect in the other.
result in the presence of sites in which the apical
oxygen of the VO5 groups formed is unshared.
This present paper reports the results of an in situ
XRD and Raman study of the evolution and crystal-
lization of vanadyl pyrophosphate in VPO catalysts
over time under catalytic reaction conditions from the
crystalline VOHPO4·0.5H2O precursor. This study
addresses: (i) the nature of the precursor transforma-
tion on the atomic, nano-, and macroscales, and (ii)
the phase composition of the fresh and equilibrated
catalysts under realistic catalytic reaction conditions.
2. Experimental
2.1. Synthesis
1. Aqueous system. The VOHPO4·0.5H2O precursor
was prepared in aqueous medium according to a
previously reported procedure [4].
2. Organic system. Preparation of the organic
VOHPO4·0.5H2O precursor possessing the synthe-
sis P/V ratio of 1.18 has been previously described
[4]. The VOHPO4·0.5H2O precursor obtained was
transformed into the catalytic (VO)2P2O7 phase
and conditioned on stream in 1.2 vol.% n-butane
in air at 653 K for 30 days. This catalyst exhibited
the selectivity to maleic anhydride of 69 mol%
at 76 mol% n-butane conversion in 1.2 vol.%
n-butane in air at 673 K.
3. Model organic system. The preparation of the
model VOHPO4·0.5H2O precursor possessing
platelet morphology has been described in detail
elsewhere [4].
2.2. Characterization
The modeling of extended defects in the crystalline
structure of vanadyl pyrophosphate has also been
published recently [16]. These defects were proposed
as being due to a bc stacking fault, i.e. a change in
2.2.1. In situ XRD
The XRD measurements were made with a Scintag
PAD-X Θ–Θ X-ray diffractometer with an Anton Paar
HTK-10 heating stage. Step scans were conducted us-
ing Cu K␣ radiation with a scintillation detector, and
scans ranged from 10 to 80◦ 2Θ with a step size of
0.03◦ 2Θ and a count rate of 10 s per step. Powdered
aqueous VOHPO4·0.5H2O precursors were mixed
with acetone, and the suspensions were deposited as
a thin layer on the heating stage, which consisted of
=
=
=
=
=
=
orientation from –V O–V O–V to V–O V–O V–
at the stacking fault boundary. This was attributed
4−
to the incomplete dimerisation of the PO3 groups
present in the layered structure of the precursor during
the topotactic transformation to the pyrophosphate.
Stacking faults may also be used to rationalize the
presence of V5+ sites in (VO)2P2O7, since such faults

V.V. Guliants et al. / Journal of Molecular Catalysis A: Chemical 172 (2001) 265–276
267
a Pt strip that heated in response to increased elec-
trical current. The heating stage was enclosed within
a vacuum chamber and purged with dry nitrogen gas
throughout the XRD measurements to prevent the ox-
idation of V(IV) in VOHPO4·0.5H2O. In the case of
the (VO)2P2O7 catalyst, the heating stage was purged
with 0.9 vol.% n-butane in air. Diffraction data were
collected at room temperature (293 K) and at 583 K in
the case of VOHPO4·0.5H2O, and room temperature
and 673 K in the case of (VO)2P2O7, as determined
by a Pt-10% RhPt thermocouple spot-welded to the
Pt strip heater. Temperature gradients across the sam-
ple area have been minimized by the attachment of a
0.254 mm thick piece of Pt foil in the center 15 mm of
the strip heater to reduce the resistance of the heater
in the sample area [17]. For the high temperature
experiments, the sample temperature was raised to
583 and 673 K in the case of VOHPO4·0.5H2O and
(VO)2P2O7, respectively, at a rate of 10 K/min and
maintained to within 5 K with a Microstar 828D
controller and Scintag 150 A dc power supply.
Systematic intensity errors due to angular diver-
gence for reflections below 55◦ 2Θ were corrected
using empirically derived divergence function that
was successfully tested with a LaB6 standard (NIST
SRM 660). In order to assess the possibility of the
irreversible structural changes at high temperature,
VOHPO4·0.5H2O specimens were heated separately
in furnace at 583 K for 8 h. Comparison of powder
XRD patterns from the material before and after
heating revealed no discernible differences. Like-
wise, the similarity of the XRD patterns collected at
room temperature before and after this high tempera-
ture diffraction experiment indicated no evidence for
dehydration or reconstructive transformation of the
VOHPO4·0.5H2O sample.
cylindrical heating coil surrounding the quartz cell was
used to heat the sample and the temperature was mea-
sured by an internal thermocouple. The quartz cell was
capable of operating up to 873 K, and flowing gas was
introduced into the cell at a rate of 50–300 cm3/min at
atmospheric pressure. The 514.5 nm line of the Ar+
laser using between 10 and 100 mW of power was
focused on the sample disc in a right angle scattering
geometry. An ellipsoid mirror collected and reflected
the scattered light into the spectrometer’s filter stage
to reject the elastic scattering component. The result-
ing filtered light, consisting primarily of the Raman
component of the scattered light, was collected with
an EG&G intensified photodiode array detector which
was coupled to the spectrometer and was thermo-
electrically cooled to 238 K. The photodiode array
detector was scanned with an EG&G OMA III optical
multichannel analyzer (Model 1463).
The in situ Raman spectra were obtained over the
273–773 K range employing the following procedure.
The VOHPO4·0.5H2O precursor or VPO catalyst
samples were placed into the cell and the Raman spec-
tra were collected at room temperature. The samples
were then heated up to successively higher temper-
atures in ca. 50 K increments in either flowing pure
helium or a gaseous 1:4 mixture of oxygen and 1.49%
n-butane in helium with a flow rate of 100 cm3/min,
and the Raman spectra were collected immediately
after stabilization at each set-point temperature.
2.2.3. 31P spin echo NMR
The detailed description of the 31P NMR experi-
ments can be found elsewhere [4].
3. Results
3.1. VOHPO4·0.5H2O structure refinement
2.2.2. In situ Raman spectroscopy
The in situ Raman spectrometer system consisted
of a quartz cell and a sample holder, a triple-grating
spectrometer (Spex, Model 1877), a photodiode array
detector (EG&G, Princeton Applied Research, Model
1420), and an argon ion laser (Spectra-Physics, Model
165). The sample holder was made from a metal alloy
(Hastalloy C), the 100–200 mg sample disc was held
by the cap of the sample holder. The sample holder
was mounted onto a ceramic shaft that was rotated
by a 115 V dc motor at a speed of 1000–2000 rpm. A
Rietveld refinement of the XRD data was accom-
plished using the General Structure Analysis System
(GSAS) of Larson and Von Dreele [18]. Starting pa-
rameters for VOHPO4·0.5H2O were obtained from
Leonowicz et al. [6]. Refinement for both structures
was performed for values of 2Θ between 20 and 80◦.
Five reflections produced by the Pt strip heater and
a minor unidentified peak with dhkl = 2.12 Å were
excluded from the analysis. The following parameters

268
V.V. Guliants et al. / Journal of Molecular Catalysis A: Chemical 172 (2001) 265–276
Table 1
Table 2
Refinement data for VOHPO4·0.5H2O structure at room temper-
Fractional coordinates for VOHPO4·0.5H2O structure at room tem-
ature and 583 K
perature and 583 K (origin at −1)
Temperature
Atom Temperature (K)
x
y
z
Room temperature a 293 K b
583 K b
V(1) 293
583
0.0431(6) 0.25
0.0440(6) 0.25
0.0322(5)
0.0334(5)
Space Group
a (Å)
b (Å)
Pmmn Pmmn
Pmmn
7.420(1) 7.4223(6)
7.4509(6)
9.6058(7)
5.7371(3)
410.62(6)
2.765
0.0794
0.1138
2.099
P(1) 293
583
0.25
0.25
0.5327(7)
0.5316(7)
0.2144(8)
0.2078(8)
9.609(2)
5.6931(7)
9.6109(8)
5.6993(3)
c (Å)
O(1) 293
583
0.25
0.25
0.5137(17)
0.5082(15)
0.4879(9)
0.4777(9)
Volume (Å3)
Density (g/cm3)
exp Rw
405.9(4)
406.55(2)
2.81
2.792
0.041
0.042
0.0679
0.0968
2.035
O(2) 293
583
0.0830(13) 0.6084(10)
0.0834(12) 0.6056(10)
0.1521(13)
0.1499(12)
Rwp
Reduced χ2
O(3) 293
583
0.25
0.25
0.3889(15)
0.3855(15)
0.1133(19)
0.1123(17)
a Leonowicz et al. [6].
b This work.
O(4) 293
583
−0.0623(19) 0.25
−0.0502(20) 0.25
0.2768(15)
0.2753(18)
O(5) 293
583
0.25
0.25
0.25
0.25
−0.2852(21)
−0.2844(25)
0.629
0.629
were refined: (i) scale factor; (ii) specimen displace-
ment; (iii) background, which was modeled by 18
refinable coefficients in a cosine Fourier series; (iv)
lattice parameters; (v) preferred orientation parallel to
(0 0 1) to account for the strong basal cleavage exhib-
ited by the material; (vi) profile shapes, which were
fitted to symmetric pseudo-Voigt functions (previ-
ous refinement of a LaB6 standard (NIST SRM 660)
provided starting values for the profile parameters,
Gaussian Caglioti coefficients were held constant,
and only the Lorentzian components and the Scher-
rer coefficient for Gaussian broadening were varied);
and (vii) fractional coordinates for V, P, and O. In
both experiments, the P(1)–O(1) distance refined to
unreasonably high values (ca. 1.64 Å), presumably
as a result of O(1)–H(2) bonding. Consequently, the
P(1)–O(1) bond was fixed at 1.56 Å by soft constraint.
Upon convergence, the goodness of fit as measured
by χ2 was 2.035 and 2.099 for the room temperature
and the 583 K analyses. Refinement results are pre-
sented in Table 1, and refined fractional coordinates
are included in Table 2.
Refinement of the room temperature structure of
VOHPO4·0.5H2O from powder XRD data was in
strong agreement with the single crystal results pre-
sented by Torardi and Calabrese [5] and Leonowicz
et al. [6]. The structure consists of face-sharing VO6
octahedra that are corner-linked by phosphate tetrahe-
dra to form sheets parallel to (0 0 1) (Fig. 1). Within
each VO6 octahedron, a single V–O(4) double bond
is indicated by a refined bond length of ca. 1.56 Å.
H(1)a 293
583
0.25
0.25
0.178
0.178
a Position determined from observed Fourier synthesis maps.
The two previous studies of VOHPO4·0.5H2O dis-
cerned positional disorder for O(1) and O(4). The
powder diffraction data in this study did not support
a determination of anisotropic temperature coeffi-
cients, but observed Fourier synthesis maps do reveal
an anisotropic electron density distribution for these
atoms parallel to (1 0 0) and [0 1 0], respectively. In
addition, O(5) displayed elongated electron density
contours parallel to (1 0 0) and perpendicular to the
H(1)–O(5)–H(1) plane. As noted in the earlier studies
[5,6], these slight displacements from mirror planes
suggest that the oxygen atoms are either statically or
dynamically disordered between proximal sites.
Neighboring layers in VOHPO4·0.5H2O are hydro-
gen-bonded via molecular water groups and hydroxyl
groups. This structural hydrogen accounts for the
refined V–O(5) length of 2.37 Å, which is unusu-
ally long due to two H(1) atoms bonded to O(5).
Leonowicz et al. [6] identified, but did not refine
the position of H(1) at (0.25, 0.166, 0.668). This
location results in reasonable H–O bond distances,
but the H(1)–O(5)–H(1) bond angle is nearly 140◦,
somewhat removed from the ideal tetrahedral angle
of 109.5◦. Examination of observed Fourier maps in

V.V. Guliants et al. / Journal of Molecular Catalysis A: Chemical 172 (2001) 265–276
269
bond angle close to tetrahedral geometry. In addition,
it maintains a hydrogen-bonded O(1)–H(1)···O(5)
angle near 170◦. As with the earlier studies [5,6], the
location of H(2) could not be determined, though the
refined bond length of 1.65 Å for P(1)–O(1) supports
the speculation of Leonowicz et al. [6] that H(2) is
bonded to O(1) to form a hydrogen bond with O(4).
Heating VOHPO4·0.5H2O to 583 K resulted in
only subtle changes in the structure and yielded no
evidence for a symmetry-changing phase transition.
The unit cell volume increased by 1.00% with the
temperature increase of 290 K. This overall expansion
resulted from increases in the lengths of a and c, as
b contracted with higher temperature. Thus, minor
extension occurred parallel to the (VO6)2 chains and
along the hydrogen-bonded layers. Both the PO4 and
the VO6 polyhedra remained quite rigid with increase
in temperature. At 293 K ꢀP–O
= 1.50 0.04 Å
ave
and ꢀV–O
= 2.01 0.25 Å, and the values at 583 K
ave
are different only in terms of standard deviation, with
ꢀP–O
= 1.50 0.05 Å and ꢀV–O
= 2.01
ave
ave
0.27 Å. The observed rigidity of the PO4 tetrahedra
is consistent with high temperature studies of other
phosphates, such as the AlPO4 polymorph berlinite
[19].
Torardi and Calabresi [5] argued that below ca.
143 K the oxygen atoms in VOHPO4·0.5H2O invert
from positional disorder to an ordered arrangement,
which induces a doubled periodicity along c. Exami-
nation of the elongated electron density contours for
O(1), O(4), and O(5) in the present study at 583 K
revealed no change in eccentricity from room tem-
perature. If the positional disorder is dynamic, rather
than static, the increase in temperature apparently does
not increase the amplitude of the oscillation. More-
over, Fourier observed maps revealed the presence
of H(1) atoms bonded to O(5) at 583 K at close to
the room temperature position. This observation and
the insignificant change in the magnitude of the basal
spacing suggest that little or no dehydration occurred
during the course of these experiments.
Fig. 1. Projections of the structure of VOHPO4·0.5H2O (a) along
the [0 0 1] axis; and (b) off the (1 0 0) axis.
3.1.1. Thermal transformation of VOHPO4·0.5H2O
The in situ Raman spectra of the model organic
VOHPO4·0.5H2O precursor during its thermal trans-
formation in inert helium atmosphere at 298–823 K
are shown in Fig. 2a. Thermal broadening and gra-
dual loss of long and short range order in VOHPO4·
the present study suggests a slightly different posi-
tion for H(1) at (0.25, 0.178, and 0.629). Relocation
of H(1) to this position lowered the value of χ2 by
0.05 in both refinements, and it produces an H–O–H

270
V.V. Guliants et al. / Journal of Molecular Catalysis A: Chemical 172 (2001) 265–276
observed at elevated temperatures. Above 704 K,
even the local order is lost and after further heat-
ing to 823 K, the weak P–O stretch of the poorly
crystalline vanadyl pyrophosphate is observed in the
in situ Raman spectra at ca. 930 cm−1. Nevertheless,
no other crystalline phase was observed for the in situ
Raman spectra at 823 K, even after cooling to room
temperature.
Similar loss of crystallinity was observed during
the VOHPO4·0.5H2O transformation in the reactive
n-butane and oxygen-containing atmosphere (Fig. 2b).
However, in this case vanadyl(IV) pyrophosphate
and oxidized ␦-VOPO4 are already observed along
with VOHPO4·0.5H2O at 708 K, in agreement with
previous observations [21]. The transformation of
VOHPO4·0.5H2O into a mixture of vanadyl(IV) py-
rophosphate and ␦-VOPO4 is complete at 708 K
within 45 min. In a previous study [4], the Raman
scattering cross-section of ␦-VOPO4 was found to
be approximately three times that of vanadyl(IV)
pyrophosphate. Therefore, even a minor ␦-VOPO4
component present in slightly oxidized fresh cat-
alysts may be detected using Raman spectroscopy
(Fig. 2b).
3.1.2. Equilibrated VPO catalysts under reaction
conditions
The changes in the equilibrated organic VPO cat-
alysts as a function of temperature in the reactive
n-butane and oxygen-containing atmosphere were
followed by the in situ XRD and Raman techniques.
Prior to the in situ experiments, these catalysts were
equilibrated for over 30 days using 1.2 vol.% n-butane
in air at 653 K. The in situ XRD patterns and Raman
spectra are shown in Figs. 3 and 4, respectively. Both
of these experimental techniques demonstrated that
the equilibrated VPO catalysts contained only crys-
talline vanadyl pyrophosphate under catalytic reaction
conditions studied. The intensity of the asymmetric
pyrophosphate P–O stretch observed at 921 cm−1
decreased at high temperature as the result of ther-
mal broadening and its position shifted to higher
wavenumbers, 929.5 cm−1 (Fig. 4). The room temper-
ature 31P spin echo NMR spectrum of the equilibrated
organic catalyst (Fig. 5) showed a single peak at ca.
2400 ppm characteristic of vanadyl pyrophosphate
and no minor diamagnetic VOPO4 phases expected
at ca. 0 ppm [4].
Fig. 2. In situ Raman spectra of VOHPO4·0.5H2O at (a) 298–823 K
in flowing He at 100 cm3/min; and (b) 298–708 K in flow-
ing 1:4 mixture of oxygen and 1.49 mol% n-butane in He at
100 cm3/min.
0.5H2O is observed in the in situ Raman spectra as
VOHPO4·0.5H2O is heated to 704 K. The charac-
teristic P–O stretch of VOHPO4·0.5H2O observed
at 981 cm−1 was shifted slightly to 979 cm−1 at
684–704 K; a new shoulder at 1002 cm−1 is also

V.V. Guliants et al. / Journal of Molecular Catalysis A: Chemical 172 (2001) 265–276
271
Fig. 5. 31P spin echo NMR spectrum of the equilibrated organic
VPO catalyst at room temperature.
Fig. 3. In situ XRD patterns of the equilibrated organic VPO
catalyst at (a) room temperature; (b) 653 K; and (c) in flowing
0.9 mol% n-butane in air at 30 cm3/min and 653 K.
The d-spacing increases slightly with temperature
from 5.699 to 5.737 Å, which may indicate slight
weakening of the interlayer hydrogen bonding. In
fact, the V–OH2 distance between the oxygen atom
of the structural water molecule and the vanadium
atom to which it is coordinated increases slightly
with temperature from 2.373 to 2.383 Å. At the same
time, the hydrogen bond between the protons of the
structural water and the oxygens of the P–OH groups
in the adjacent layers shortens from 2.010 to 1.988 Å.
These observations may be interpreted as the early
manifestation of the thermal transformation. Thus, the
d-spacing increases with temperature due to thermal
expansion leading to weakened H-bonding between
the layers (Fig. 1b), while structural water molecules
leave the coordination sphere of the cis-vanadyl
dimers to hydrogen-bond with the P–OH groups
prior to their loss in dehydration. Simultaneously, we
=
may postulate that the V O groups begin to rotate
Fig. 4. In situ Raman spectra of the equilibrated organic VPO
catalyst in flowing 1:4 mixture of oxygen and 1.49 mol% n-butane
in He at 100 cm3/min at (a) 298; (b) 533; (c) 618; and (d) 653 K.
to a parallel arrangement relative to each other in a
cis-vanadyl dimer, observed in a decrease of the angle
between the V O bond and the c-axis from 29.35 to
=
26.80◦ with temperature. In this manner, the distance
between the equatorial oxygens in a cis-vanadyl dimer
along the shared face (Fig. 1a) contracts from 2.670 to
2.603 Å with increase in temperature. This is the same
distance which is observed in the vanadyl(IV) py-
rophosphate structure at room temperature. The V–V
distance in the vanadyl dimer does not change with
temperature (3.070 Å) and is much shorter than the
4. Discussion
4.1. Atomic scale changes in VOHPO4·0.5H2O
structure at high temperature
The high temperature structure of VOHPO4·0.5H2O
shows no evidence of the loss of the structural water.

272
V.V. Guliants et al. / Journal of Molecular Catalysis A: Chemical 172 (2001) 265–276
H2O. . . O(H)P hydrogen bond distance. These obser-
vations are further examined in the light of two recent
structural transformation models [11,12].
The recently proposed models of the VOHPO4·
0.5H2O transformation into (VO)2P2O7 [11,12] are
based on the observed similarities between the sin-
gle crystal structures of these phases. According to
Amorós et al. [12], the dehydration process begins
with concerted displacement of the basal layers of
VOHPO4·0.5H2O along the [1 1 0] direction to align
2−
the HPO4 groups on the adjacent layers for con-
4−
densation into P2O7 groups. The elimination of the
structural water and the condensation of the layers are
also concerted processes. The total synchronism of
this displacement could be easily disrupted by local
defects leading to a disordered intermediate, requiring
high temperatures for complete conversion into crys-
talline (VO)2P2O7 [12]. Amorós et al. detected the
disordered intermediate after elimination of ca. 60% of
water obtained from the hydrogen phosphate groups.
However, such concerted displacement of the layers
relative to one another should also result in disrup-
tion of orthorhombic symmetry in VOHPO4·0.5H2O,
which was not observed in the present study.
Conversely, Torardi et al. [11] proposed a phospho-
rus inversion model, which provided a detailed expla-
nation for the formation of the P–O–P connectivity
without the orthorhombic symmetry change during the
transformation. Therefore, the changes observed in the
precursor structure at high temperature in the present
study are further discussed in terms of the phosphorus
inversion model.
According to the phosphorus inversion model pro-
posed by Torardi et al., the structural water coordi-
nated to the vanadyl dimers is lost first, followed by
a decrease in the interlayer spacing. Next, the VO5
groups pivot about the shared edge in each vanadyl
Fig. 6. Projections of the structure of (a) VOHPO4·0.5H2O onto
the ac plane; and (b) (VO)2P2O7 onto the bc plane.
corresponding distance in the vanadyl(IV) pyrophos-
phate at room temperature, 3.231 Å [8]. A slight
expansion of the a-axis is also observed (Fig. 1a).
Since the V–V distance in the dimer does not change
with temperature, the a-axis expansion occurs at the
expense of flattening of the layers. In fact, the a-axis
expansion due to the increase in the V–V distance
would only contribute 0.32 Å per unit cell or 4.3% of
the a-parameter and should result in c/2 = 7.7423 Å
in vanadyl pyrophosphate. However, the c/2 axis in
vanadyl pyrophosphate is 8.29425 Å, and the addi-
tional increase of 0.55195 Å or 7.4% per unit cell
may be assumed to be due to the flattening of the
undulating HPO4 layers in the precursor phase during
the thermal transformation (see Fig. 6). The distance
between the phosphate tetrahedra along the b-axis
increases with temperature (4.196 Å at 583 K versus
4.177 Å at 293 K) resulting in a shorter structural
=
dimer to an arrangement where all V O bonds are
approximately parallel, leading to the 12% increase
in the a-lattice parameter. Subsequently, the protons
2−
from half of the HPO4 groups transfer to the re-
2−
3−
mainder of the HPO4 groups to yield the PO4
−
and H2O–PO3 units. Rem−oval of further water from
these latter units (H2O–PO3 → PO3−+H2O) allows
the layers to fully condense into the pyrophosphate
structure. Half of the vanadyl oxygens displace in
the c direction to produce edge-sharing trans-vanadyl
dimers. Ultimately, the coordinatively unsaturated

V.V. Guliants et al. / Journal of Molecular Catalysis A: Chemical 172 (2001) 265–276
273
−
PO3 groups invert through a plane of the three basal
phosphate groups. However, proton transfer between
3−
2−
oxygen atoms to bond to PO4 units located above
the HPO4 groups may occur randomly and result
2−
or below the original layer. The resulting vanadyl py-
rophosphate structure is expected to display a disorder
in the arrangement of the pyrophosphate groups that
hold the layers together. Much higher temperatures or
longer times under catalytic reaction conditions are
required to break and reform the P–O–P bonds into
the observed ordered 3D arrangement [7–10]. While
the model of Torardi et al. does not require the com-
plete loss of the crystalline order, some disorder in
the pyrophosphate groups and vanadium atoms along
in an incomplete condensation of all HPO4 groups
into the pyrophosphate network. Assuming a ran-
dom distribution of −half of the protons, any particular
3−
PO4
or PO3OH2 group has 50% probability of
finding a complementary group in the adjacent layer
suitable for initial condensation into the P–O–P link-
ages. Assuming irreversibility of the P–O–P bond for-
mation at low temperatures, there is 75% probability
that any particular phosphate group would eventually
form a P–O–P bond due to further proton diffusion
2−
the c-axis is expected [9,20].
However, these proposed HPO4
according to 2HPO4 → PO43− + PO3OH2−. Such
2−
dehydration
partially transformed intermediate would be charac-
terized by a significant variation of the local order
and the d-spacing leading to its disordered charac-
ter on the nanoscale in agreement with the earlier
observations made by Amorós et al. [12]. It should
be noted that Amorós et al. [12] observed that the
disordered intermediate was formed after evolution of
ca. 60% water from the condensation reaction of the
and phosphorus inversion steps [11] may require
a high energy of activation, since they involve the
generation and −subsequent condensation of a large
3−
number of PO3 and PO4 groups, which bear lo-
cally uncompensated charges of the same sign, to
form the P–O–P bonds. Conversely, the dehydration
of the HPO4
via a nucleophilic attack of the oxygen attached to
the de-protonated O3P–O3− groups in one layer o−n
the phosphorus atom in the di-protonated PO3OH2
moieties in the layer above. For this reaction, proceed-
2−
layers into (VO)2P2O7 may occur
2−
HPO4 groups. If one also accounts for the loss of
the surface-adsorbed moisture, the observed weight
loss (ca. 60%) is indeed close to the 50% mark, cor-
3−
responding to random distribution of the PO4
or
−
−
ing via inversion of the PO3 tetrahedron to form a
PO3OH2 groups on the adjacent layers. This obser-
vation may indicate that the initial condensation step
leading to the disordered intermediate occurs faster
than the accompanying redistribution of the protons,
which would result in the evolution of 75% of water
due to such condensation reactions. Such disordered
P–O–P linkage, the water molecule is a good leaving
group. If we consider the hydrogen bonding scheme
suggested by Leonowicz et al., for the room tempera-
ture structure of VOHPO4·0.5H2O, the proton transfer
2−
between the HPO4 groups that precedes the con-
densation may be mediated by t+he vanadyl oxygens
pyrophosphate structure may contain ca. 25% of
2−
2−
=
shown in Fig. 4 [6]: V(IV) O2 + H–OPO3
→
the isolated HPO4
groups, which were unable to
+
V(IV)–OH3 + OPO33−. This proton transfer is
form pyrophosphate linkages. At higher temperatures,
poorly crystalline (VO)2P2O7 would be formed due
to further proton diffusion, rearrangement and con-
=
likely to decrease the V O bond order, elongate this
bond along the b-axis and bring the vanadyl O atom
closer to the V atom in the layer above, thereby pro-
moting the conversion of cis-vanadyl dimers to the
trans arrangement observed in (VO)2P2O7. Thus, the
OH group would migrate to the V atom directly in
the adjacent layer followed by a proton transfer back
2−
densation of the residual HPO4 groups present on
the phosphate layers. The fully ordered (VO)2P2O7
structure is observed only at higher (near melt) tem-
2−
peratures after a complete dehydration of the HPO4
groups and the P–O–P bond breaking and reformation
[7–10].
3−
2−
to a neighboring PO4
or HPO4
group, result-
ing in the trans-vanadyl oxygen arrangement. This
scheme suggests both the mechanism of the proton
transfer as well as the role of protons as catalysts in
the cis/trans transformation step. The complete con-
densation into the reported pyrophosphate structure
[7–10] requires a perfectly ordered alignment of the
The results of the present high temperature XRD
study do not provide insights into the further steps of
dehydration and condensation into the vanadyl(IV)
pyrophosphate structure. Further changes in the
nanoscale order of the precursor were studied by the
in situ Raman spectroscopy.

274
V.V. Guliants et al. / Journal of Molecular Catalysis A: Chemical 172 (2001) 265–276
4.2. Evolution of nano- and macroscale order during
VOHPO4·0.5H2O transformation
transformation to ␣II-VOPO4; and (ii) the dehydration
of VOHPO4·0.5H2O to (VO)2P2O7 [22]. However,
it seems most likely that the appearance of the oxi-
dized phases in these cases is probably caused by the
poor heat and mass transfer conditions during in situ
Raman experiments.
In agreement with the earlier findings [12], grad-
ual loss of the nanoscale order in VOHPO4·0.5H2O
was observed as the model organic VOHPO4·0.5H2O
precursor was heated from 583 to 704 K under He
atmosphere (Fig. 2a). Above this temperature the
structure was completely disordered and nanocrys-
talline vanadyl pyrophosphate (ca. 10 nm domains by
XRD) was observed for the in situ Raman spectra
only at 823 K, in agreement with the transforma-
tion mechanism proposed above. The precursor and
the pyrophosphate phase did not coexist under these
dehydration conditions in agreement with the pre-
vious observations of Amorós et al. [12] and Volta
coworkers [21,22]. Furthermore, Amorós et al. [12]
observed the transformation of the higher hydrates of
VOHPO4·nH2O (n = 1–4) to (VO)2P2O7 in a powder
neutron diffraction study. Remarkably, all crystalline
hydrates, including VOHPO4·0.5H2O, coexisted with
one another during the sequential dehydration. How-
ever, it was observed that the solid-state transforma-
tion of VOHPO4·0.5H2O into (VO)2P2O7 proceeded
via a disordered intermediate [12,21]. It is interesting
that the structures and the V–O–P connectivity of
the higher hydrates are very different from those of
VOHPO4·0.5H2O. They contain isolated vanadyl oc-
tahedra, which transform to the face-sharing vanadyl
dimer coordination in VOHPO4·0.5H2O upon dehy-
dration. Given the much closer similarity between
these two structures, it is remarkable that the crys-
talline hemihydrate and pyrophosphate structures
were not observed to coexist during the dehydration
experiments of Amorós et al. [12]. However, when this
transformation was conducted in the reactive n-butane
and oxygen-containing atmosphere, vanadyl(IV) py-
rophosphate and oxidized ␦-VOPO4 were readily
observed along with VOHPO4·0.5H2O at 708 K with
in situ Raman (Fig. 2b).
The considerable differences between the ex situ
and in situ catalysts during the activation, and the
presence of VOPO4 phases at the moment of the
transformation of the precursor, were also observed
by Overbeek et al. [23] using X-ray diffraction. They
observed that the VPO catalysts were a mixture of
unknown VOPO4 phases under catalytic reaction
conditions and contained only (VO)2P2O7 after ex-
posure to air at room temperature. However, forma-
tion of crystalline VOPO4 phases detected by XRD
typically occurred under even stronger oxidizing con-
ditions than those employed in the in situ Raman
studies [4]. Transformation of (VO)2P2O7 [24,25]
and various VOPO4 phases [25] into a “new, slightly
anion-deficient phase” [24] and V2O5 [25] during
their exposure to water vapor at high temperatures
has been recently reported. Xue and Schrader [25]
observed transformation of (VO)2P2O7 into V2O5 by
in situ Raman after the VPO catalyst was exposed to
5–10% water vapor at 723 K. Formation of the surface
V2O5 phase was reversible after a short-term expo-
sure to water vapor, and the surface phosphate lost
due to hydrolysis was probably replenished by diffu-
sion of the phosphate anions from the crystalline bulk.
After a prolonged exposure to water vapor and a sig-
nificant phosphate loss, the bulk V2O5 phase formed
irreversibly.
These previous observations [21–25] further em-
phasized the extreme sensitivity of the VPO catalysts
to various reducing, oxidizing and hydrolytic envi-
ronments. However, when the typical dry 1.2 vol.%
n-butane/air feed was used, the VOHPO4·0.5H2O
precursor transformed inside a catalytic microreactor
contained only vanadyl pyrophosphate and no other
nanocrystalline VPO or V2O5 phase [26].
Evolution of the nanocrystalline vanadyl pyrophos-
phate and other oxidized VOPO4 phases coincided
with an onset of n-butane oxidation activity and selec-
tivity of the VPO catalysts [21,22]. However, the selec-
tivity of these fresh catalysts to maleic anhydride was
low (e.g. 30–45 mol%) as compared with the equili-
brated catalysts (e.g. 70–80 mol%) and improved over
Volta and coworkers [21,22] also observed forma-
tion of oxidized ␣II-, ␦- and ␥-VOPO4 phases during
the activation of the VOHPO4·0.5H2O precursor in the
n-butane/air atmosphere by in situ Raman spec-
troscopy. They considered two parallel routes for the
transformation of the precursor: (i) oxidehydration
of VOHPO4·0.5H2O to VOPO4 phases, accompanied
by a slow reduction of ␦-VOPO4 to (VO)2P2O7 and

V.V. Guliants et al. / Journal of Molecular Catalysis A: Chemical 172 (2001) 265–276
275
time during on-stream conditioning [3,8,21,22,26].
5. Conclusion
During this conditioning procedure, the disordered
(VO)2P2O7 phase gradually crystallized in the
bulk VPO catalysts over a period of hundreds of
hours.
In the present paper, the transformation of the cat-
alytic VOHPO4·0.5H2O precursor to well-crystallized
(VO)2P2O7 in the equilibrated VPO catalysts was
studied over several length scales by in situ and ex situ
Raman and XRD techniques. The refinement of the
room temperature structure of VOHPO4·0.5H2O re-
sulted in an improved location of the structural water
molecule coordinated to the edge-sharing vanadyl(IV)
dimers. The structure of the VOHPO4·0.5H2O pre-
cursor was stable up to 583 K in dry nitrogen during
in situ XRD experiments.
High-resolution TEM studies detected a disor-
dered layer ca. 2 nm thick covering the surface (1 0 0)
planes of (VO)2P2O7 in fresh catalysts [26]. This
nanoscale structural feature became progressively
less pronounced with time on stream, and could be
no longer detected by TEM after a few weeks under
catalytic reaction conditions. It was concluded that
a specific surface termination of the (1 0 0) planes
of (VO)2P2O7 was responsible for a significant im-
provement of selectivity to maleic anhydride for the
equilibrated catalysts [26].
Minor atomic scale changes in the precursor struc-
=
ture observed at 583 K, such as the rotation of the V O
moieties, weakening of the hydrogen bonding, migra-
tion of the structural water molecules during dehydra-
tion, and flattening of the HPO4 layers, were indicative
of the early stages of the precursor transformation
and consistent with the phosphorus inversion model
of transformation. Furthermore, the random nature of
The phase analysis of the equilibrated VPO cata-
lysts is typically conducted ex situ under very different
conditions, i.e. in the absence of the catalytic reaction.
The phase composition of the equilibrated organic
VPO catalyst under catalytic reaction conditions was
studied here employing in situ XRD and Raman meth-
ods (Figs. 3 and 4). This catalyst was conditioned in
flowing 1.2 vol.% n-butane in air at 653 K for 30 days.
The in situ XRD and Raman methods and ambient
31P spin echo NMR (Fig. 5) indicated that the equili-
brated VPO catalysts contained only (VO)2P2O7 (ca.
30 nm domains) under both the ambient and n-butane
oxidation conditions, although the intensity of the
XRD and Raman signal decreased at high tempera-
ture due to increased thermal broadening. In addition
to thermal broadening, the asymmetric P–O py-
rophosphate stretch observed at 921 cm−1 at ambient
2−
the proton transfer between the HPO4 groups was
suggested to be responsible for the incomplete con-
densation of these groups in the fresh catalysts under
typical catalytic reaction conditions and the formation
2−
of the disordered intermediate. The HPO4 protons
were also indicated as the catalytic species in the cis- to
trans-vanadyl conversion step. Complete transforma-
tion to well-crystalline (VO)2P2O7 gradually occurred
under catalytic reaction conditions via the breakup and
rearrangement of the P–O–P connectivity.
Nanocrystalline oxidized VOPO4 phases were
detected by in situ Raman spectroscopy when the
precursor was transformed in the reactive n-butane/
air atmosphere in agreement with earlier observations,
possibly due to the specificity of the in situ conditions.
However, the precursors transformed under typical
plug flow conditions in the n-butane/air atmosphere
contained only poorly crystalline (VO)2P2O7 and no
nanocrystalline oxidized phases.
Freshly transformed VPO catalysts were character-
ized by low activity in n-butane oxidation and selectiv-
ity to maleic anhydride. Nanocrystalline (VO)2P2O7
and the disordered component present in the fresh
catalysts gradually transformed into well-crystalline
(VO)2P2O7 in equilibrated VPO catalysts over a
period of hundreds of hours under catalytic reaction
conditions.
temperature shifted to higher frequency, 929.5 cm−1
,
at 653 K (Fig. 4). High temperature X-ray structural
studies of single crystal (VO)2P2O7 may provide new
insights into the pyrophosphate anion geometry and
the origin of this shift. The previous observations of
minor phases other than (VO)2P2O7 under catalytic
reaction conditions may be explained by a reversible
oxidation of (VO)2P2O7 in the VPO catalysts at
high temperature and n-butane lean oxidizing con-
ditions into nanocrystalline ␦- and ␥-VOPO4 phases
[25,27,28]. The best (VO)2P2O7 catalysts were stable
under the typical fixed bed conditions (1.2–1.5 vol.%
n-butane in air at a GHSV of 1000–2000 h−1
and 653–713 K) and did not form nanocrystalline
oxidized phases.

276
V.V. Guliants et al. / Journal of Molecular Catalysis A: Chemical 172 (2001) 265–276
[9] J.R. Ebner, M.R. Thompson, in: R.K. Grasselli, A.W. Sleight
Both the in situ XRD and Raman techniques in-
(Eds.), Structure-Activity and Selectivity Relationships in
Heterogeneous Catalysis, Elsevier, Amsterdam, 1991, p. 31.
[10] P.T. Nguyen, R.D. Hoffman, A.W. Sleight, Mater. Res. Bull.
30 (1995) 1055.
[11] C.C. Torardi, Z.G. Li, H.S. Horowitz, W. Liang, M.-H.
Whangbo, J. Solid State Chem. 119 (1995) 349.
[12] P. Amorós, R. Ibáñez, A. Beltrán, D. Beltrán, A. Fuertes, P.
Gomez-Romero, E. Hernandez, J. Rodriguez-Carvajal, Chem.
Mater. 3 (1991) 407.
dicated that the equilibrated bulk VPO catalysts
contained only crystalline (VO)2P2O7 under catalytic
reaction conditions. The conditioning process in-
creased the surface area of the catalysts and resulted
in a specific surface termination of the (1 0 0) planes
of (VO)2P2O7 in the equilibrated VPO catalysts, a
phase believed to be responsible for high steady-state
activity and selectivity of these catalysts in partial
oxidation of n-butane.
[13] J.W. Johnson, D.C. Johnston, A.J. Jacobson, J.F. Brody, J.
Am. Chem. Soc. 106 (1984) 8123.
[14] E. Bordes, P. Courtine, J.W. Johnson, J. Solid State Chem.
55 (1984) 270.
[15] Z. Hiroi, M. Azuma, Y. Fujishiro, T. Saito, M. Takano, F.
Izumi, T. Kamiyama, T. Ikeda, J. Solid State Chem. 146
(1999) 369.
[16] P.T. Nguyen, A.W. Sleight, N. Roberts, W.M. Warren, J. Solid
State Chem. 122 (1996) 259.
[17] N.E. Brown, S.M. Swapp, C.L. Bennett, A. Navrotsky, J.
Appl. Cryst. 26 (1993) 77.
[18] A.C. Larson, R.B. Von Dreele, Generalized Structure Analysis
System (GSAS): Los Alamos National Laboratory Report
LAUR 86-748, 1994.
Acknowledgements
This work was supported by the Grants from
the University of Cincinnati Research and Faculty
Development Councils, BP AMOCO Corporation
(Princeton), and National Science Foundation CTS-
9417981 (Lehigh).
[19] H.N. Ng, C. Calvo, J. Phys. C. 17 (1976) 1375.
[20] J.R. Ebner, M.R. Thompson, Catal. Today 16 (1993) 51.
[21] G.J. Hutchings, A. Desmartin-Chomel, R. Olier, J.-C. Volta,
Nature 368 (1994) 41.
References
[1] F. Cavani, F. Trifiró, Catal. 11 (1976) 246.
[2] G.J. Hutchings, C.J. Kiely, M.T. Sananes-Schulz, A. Burrows,
J.-C. Volta, Catal. Today 40 (1998) 273.
[3] S. Albonetti, F. Cavani, F. Trifiró, P. Venturoli, G. Calestani,
M. Lopéz Granados, J.L.G. Fierro, J. Catal. 160 (1996)
52.
[22] J.-C. Volta, Catal. Today 32 (1996) 29.
[23] R.A. Overbeek, M. Verslujs-Helder, P.A. Warringa, E.J.
Bosma, J.W. Geus, in: S.V., Bellon, V.C. Corberan (Eds.),
New Developments in Selective Oxidation, Benalmadena,
Spain, Stud. Surf. Sci. Catal. 84 (1994) 183.
[24] P.L. Gai, K. Kourtakis, G.W. Coulston, G.C. Sonnichsen, J.
Phys. Chem. B 101 (1997) 9916.
[25] Z.-Y. Xue, G.L. Schrader, J. Phys. Chem. B 103 (1999) 9459.
[26] V.V. Guliants, J.B. Benziger, S. Sundaresan, N. Yao, I.E.
Wachs, Catal. Lett. 32 (1995) 379.
[4] V.V. Guliants, J.B. Benziger, S. Sundaresan, I.E. Wachs, J.-
M. Jehng, J.E. Roberts, Catal. Today 28 (1996) 275.
[5] C.C. Torardi, J.C. Calabrese, Inorg. Chem. 23 (1984) 1310.
[6] M.E. Leonowicz, J.W. Johnson, J.F. Brody, H.F. Shannon,
J.M. Newsam, J. Solid State Chem. 56 (1985) 370.
[7] N.E. Middlemis, Ph.D. Thesis, McMaster University, Hamil-
ton, Ont., Canada, 1978.
[27] A.D. Soejarto, G.W. Coulston, G.L. Schrader, Can. J. Chem.
Eng. 74 (1996) 594.
[28] G. Koyano, T. Okuhara, M. Misono, J. Am. Chem. Soc. 120
(1998) 767.
[8] S. Linde, L. Gorbunova, Dokl. Akad. Nauk SSSR 245
(1979) 584.