A study of some bivalent metal divanadates and their catalytic activity in the oxidation of propane

Article abstract of DOI:10.1039/b000447m

The oxidative dehydrogenation of propane to propylene is conceivable as a new process fulfilling the demand for propylene (forecast to increase in the near future) at the expense of a component of natural gas without consuming oil fractions. The divanates M₂V₂O₇ of a virtually unreducible and unoxidizable cation, Mg²⁺, of a reducible cation, Cu²⁺, and of an oxidizable cation, Mn²⁺, were prepared and characterized by XRD, surface area measurements, skeletal IR and UV-visible NIR spectroscopies, and differential thermal analysis-TGA. These materials were evaluated as catalysts for propane oxidative dehydrogenation at two different contact times. Magnesium divanadate was more active and more selective than manganese divanadate, while copper divanadate was more selective than the Mg compound. Magnesium divanadate was the most active in converting propane in the presence of oxygen. However, the yield and productivities to propylene were low, due to a low intrinsic activity with respect to supported vanadia catalysts, and also to a low selectivity. An oxidative way to ethylene and the thermal cracking to ethylene plus methane were the predominant byreactions, largely occurring in the gas phase, particularly at low contact times. True catalytic phenomena were seen at higher contact times and lower temperatures.

Full text of DOI:10.1039/b000447m

View Article Online / Journal Homepage / Table of Contents for this issue
A study of some bivalent metal divanadates and their catalytic
activity in the oxidation of propane
Carlo Resini, Fabio Milella and Guido Busca
L aboratorio di Chimica delle SuperÐci e Catalisi Industriale, Dipartimento di Ingegneria Chimica
`
e di Processo, Universita, P.le J.F. Kennedy, I-16129, Genova, Italy
Received 17th January 2000, Accepted 6th March 2000
Published on the Web 7th April 2000
The divanadates M V O of a virtually unreducible and unoxidizable cation, Mg2`, of a reducible cation,
2
2 7
Cu2`, and of an oxidizable cation, Mn2`, have been prepared and characterized by X-ray di†raction, surface
area measurements, skeletal IR and UV-vis-NIR spectroscopies, and DTA-TG analyses. These materials have
also been tested as catalysts for propane oxidative dehydrogenation at two di†erent contact times. Mg
divanadate is more active and more selective than Mn divanadate, while Cu divanadate is a little bit more
selective than the Mg compound. The activity is likely mainly associated to the redox behavior of V cations
which could be improved by reducible centers. Gas phase phenomena predominate at relatively low contact
times and higher temperatures while true catalytic phenomena are observed at higher contact times and lower
temperatures.
where couples near vanadium centers are certainly present in
the bulk and are expected to exist also at the surface. We
chose to compare the activity of Mg divanadate (already
studied by several authors and found to perform well5h7) with
divanadates of reducible metals for which the bivalent state is
the highest stable oxidation state (copper) or is a low oxida-
tion state (manganese).
1
. Introduction
The oxidative dehydrogenation of propane to propene1,2 is
conceivable as a new process fulÐlling the demand for propene
(
forecast to increase substantially in the near future3) at the
expense of a component of natural gas without consuming oil
fractions. Several catalysts have been tested for this reaction,
but the performances obtained up to now are still not suffi-
cient for industrial exploitation.4 New more active and selec-
tive materials have to be developed. According to Albonetti et
al.1 vanadium is a key element for performing this reaction.
Magnesium vanadates, Ðrst proposed by Kung et al.,5,6 have
been shown to act as performance catalysts for the oxydehy-
drogenation of alkanes to the corresponding alkenes although
several authors dispute which is the most active phase among
orthovanadate Mg (VO ) , divanadate Mg V O and meta-
2. Experimental
The preparation of M V O (with M \ Mn, Mg, Cu) was
2
2 7
carried out by two conventional methods: citrate method
(samples denoted MV1 in Table 1) and coprecipitation
(samples denoted MV2 in Table 1). In the coprecipitation case
the samples were prepared by adding stoichiometric amounts
(M : V atomic ratio 1 : 1) of Mn(CH COO) Æ 2H O (Acros,
3
4 2
2 2 7
vanadate MgV O .5h7 Several metal orthovanadates have
2
6
3
3
2
also been tested and found to be quite active.8 Additionally,
several studies have focused on vanadia supported on alumina
and titania.9h12
98%) or Mg(NO ) Æ 6H O (Carlo Erba, 98%) or
3
2
2
Cu(NO ) Æ 3H O (Carlo Erba, 99.5%) to a limpid NH VO
3
2
2
4
3
(Aldrich, 99]%) solution, hydrolysing with deionized water,
The presence of vanadyl center couples have been reported
to be the key structural element of the so-called VPO cata-
lysts, which represent the active phase for the oxidation of
butane to maleic anhydride. This material also allows the oxy-
dehydrogenation of butane to butenes.13 For this reason, we
decided to investigate the behavior of some metal divanadates,
followed by correcting the pH value at ca. 8 with NH OH
4
30% solution, aging, Ðltering and drying at 393 K for
several h.
Nitrates and residual organic compounds were decom-
posed, in air, in an electronically controlled furnace at 723 K
for 4 h. The heating and cooling rate was 40 K min~1.
Table 1 Summary of surface area and XRD data for the catalyst samples
Cell parameters/A
Ž
T calc/
K
S
/
BET
Sample
MgV1
MgV2
m2 g~1
XRD phases
a
b
c
a
b
c
773
873
35
a-Mg V O
b-Mg V O
6.608
8.407
5.418
8.412
3.495
8.006
8.015
8.447
8.740
7.966
8.724
9.491
4.921
9.499
6.736
È
81.20
È
È
È
È
È
È
88.84
È
100.68
106.90
100.59
111.77
110.30
110.19
È
103.64
72.36
103.80
È
130.40
È
È
È
È
È
È
84.31
È
2
2
7
7
7
13.773
6.606
9.281
7.674
7.674
20.579
6.720
6.899
6.718
2
2
21
a-Mg V O
2
MgV O
2
2
6
CuV1
CuV2
773
673
6
9
b-Cu V O
10.073
10.072
6.515
2
2
7
b-Cu V O
2
2
7
7
a-Cu V O
2
2
MnV1
MnV2
973
773
19
10
b-Mn V O
4.981
2
2
7
7
7
a-Mn V O
10.917
4.975
2
2
b-Mn V O
2
2
DOI: 10.1039/b000447m
Phys. Chem. Chem. Phys., 2000, 2, 2039È2045
This journal is ( The Owner Societies 2000
2039

View Article Online
The main steps of the citrate method are: (i) Preparation of
transparent solutions of M2` cations (Mn2`, Mg2` and
Cu2`) and NH VO (Aldrich, 99]%) in a suitable concentra-
92-12 apparatus, in air, up to 1273 K, with a heating and
cooling rate of 10 K min~1.
Catalytic tests were carried out at atmospheric pressure in a
continuous Ðxed bed Ñow tubular quartz reactor. Variable
amounts of catalysts calculated to have the same volume of
powder in the reactor were loaded in the form of Ðne powder
(60È70 mesh) whose total volume was 0.677 cm3.
The feed composition was 3% of hydrocarbon, 3% of
oxygen and 94% of helium. The total gas Ñow was 100 ml
min~1 for the high contact time experiments and 300 ml
min~1 for the low contact time experiments. The resulting
contact times of the propaneÈoxygen mixtures (He excluded)
with the catalyst beds were 6.7 and 2.2 s. The reactants and
the reaction products were analysed using on-line gas chro-
matographs (HP 5890) equipped with a PORAPLOT Q (HP)
capillary column, connected to a TCD detector in series with
a FID detector. Separating them was a Nickel Catalyst Tube
4
3
tion to obtain the atomic ratio M2` : V 1 : 1 of the M V O
2
catalysts; (ii) addition of a small amount of 65% HNO to the
3
2 7
solution of NH VO in order to avoid precipitation when the
4
3
two solutions are mixed; and (iii) addition of a suitable
amount of citric acid to the resulting solution in such a
manner that the molar number of equivalent anions (3 ions
per molecule of citric acid) is equal to that of the cations (M2`
and V5`).
The resulting solution was evaporated until a spongy solid
(
similar to a meringue) was obtained. The resulting powder
was calcinated at 773 K for 24 h, in order to obtain complete
decomposition of the organic compounds.
The XRD spectra were recorded on a Philips PW 1710 dif-
fractometer (Cu-Ka radiation, Ni Ðlter; 45 kV, 35 mA). Unit
cell parameters were calculated from the experimental XRD
patterns using dedicated least square software. The phases
have been identiÐed on the basis of a comparison with the
data reported in the ICDD (International Centre for Di†rac-
tion Data) and ICSD (Inorganic Crystal Structure Database)
tables.
(HP) to reduce CO to CH . A six-port valve with a 0.5 cm3
4
loop was used for the gas sampling.
3. Results and discussion
The FT-IR spectra were recorded using a Nicolet Magna
50 Fourier transform instrument. For the region above 400
3.1. Catalyst characterization
7
cm~1 a KBr beam splitter was used with a DTGS detector.
For the FIR region (600È50 cm~1) a ““solid substrateÏÏ beam
splitter and a DTGS polyethylene detector were used. KBr
pressed disks (IR region) and polyethylene pressed disks (FIR
region) were used.
Di†use reÑectance spectra in the range 2500È200 nm were
obtained with a Jasco V-570 spectrophotometer at room tem-
perature. The BET surface areas were measured with a con-
ventional volumetric instrument by nitrogen adsorption at
liquid nitrogen temperature.
3.1.1. XRD studies. In Fig. 1 the XRD powder patterns of
the three pyrovanadates are reported. The Mg divanadate
produced by the citrate method (MgV1) appears to be a
mixture of the two polymorphs a-Mg V O , monoclinic,
2
2 7
space group P2 /c, no. 14, Z \ 4 (ICDD no. 31-0816) and b-
1
Mg V O , triclinic, space group P-1, no. 2, Z \ 2 (ICSD no.
2321). The MgV2 sample, produced by coprecipitation, is con-
2
2 7
stituted of a mixture of the divanadate a-Mg V O , and the
2
2 7
metavanadate MgV O , monoclinic, space group C2/m, no.
2
6
DTA-TG experiments were performed in a Setaram TGA
12, Z \ 2 (ICSD no. 10391). The CuV1 sample is quite a pure
Fig. 1 XRD patterns of the samples. MgV1: (…) a-Mg V O , ()) b-Mg V O , (|) unknown. MgV2: (…) a-Mg V O , ()) MgV O , (|)
2 2 2 2
unknown. CuV1: (]) b-Cu V O , (|) unknown. CuV2: (]) b-Cu V O , (…) a-Cu V O , (>) V O , (=) CuO, (|) unknown. MnV1: (])
2 2 2 2
2
2
7
7
7
2 6
2
2
7
7
7
2 5
b-Mn V O , (…) a-Mn V O . MnV2: (]) b-Mn V O .
2 2 2 2
2
2
7
7
7
2040
Phys. Chem. Chem. Phys., 2000, 2, 2039È2045

View Article Online
divanadate b-Cu V O , monoclinic, space group C2/c, no. 15,
2
2 7
Z \ 4 (ICSD no. 23479), while the coprecipitated sample
CuV2 is less pure, also containing traces of a-Cu V O ,
orthorhombic, space group Fdd2, no. 43, Z \ 8 (ICSD no.
4756). In the case of MnV samples, the coprecipitated
2
2 7
3
material (MnV2) appears to be quite a pure divanadate b-
Mn V O , monoclinic, space group C2/m, no. 12, Z \ 2
2 2 7
ICSD no. 24321), while the sample prepared by the citrate
(
method (MnV1) is less pure due to the presence of the poly-
morph phase a-Mn V O , triclinic, space group P-1, no. 2.14
2
2 7
The observed crystal phases and the measured unit cell
parameters are reported in Table 1. b-Mg V O adopts a dis-
2
2 7
torted form of the so called ““thortveititeÏÏ structure
(
thortveitite \ Sc Si O ), with sheets of the divanadate ions
2
2 7
and the bivalent cation in an octahedral conÐguration.
However, the pyrovanadate bridge15 has a bent geometry
which is instead linear in thortveitite. The structure of the
other polymorph, a-Mg V O , assumed to be the same as
Co V O ,16,17 is not thortveitite-like because the pyrovana-
date entities do not form sheets. The structure of the metava-
2
2 7
2
2 7
nadate MgV O , present as a mixture with a-Mg V O in
2
6
2 2 7
the preparation MgV2, is built up of double zig-zag chains of
edge-shared octahedra. Mg coordination is octahedral in all
cases.
Fig. 2 DTA curves of: MgV1 and CuV1.
The main structure in the CuV samples is b-Cu V O
2
2 7
where cupric ions are in tetrahedral coordination and the
pyrovanadate bridge is bent (ICSD no. 23479). In the
thortveite-like structure of orthorhombic a-Cu V O , Cu
the reverse c ] b phase transformation. Thus the DTA-TG
data conÐrm that the coprecipitated sample contains
MgV O and a-Mg V O , as shown by XRD.
Results of DTA and TG study of CuV sample prepared by
the citrate method, during heating, show three endothermic
peaks: the Ðrst at 892 K, without weight change, and the
other ones at 1055 and 1080 K associated to a weight loss of
ca. 5%. The endothermic peak at 892 K, occurring without
any weight change, could be attributed to the melting of the
CuV O ÈCu V O eutectic reported to occur at 911 K.20 At
2
2 7
2 6
2 2 7
coordination is a distorted trigonal bipyramid.18
The b-Mn V O adopts the thortveitite structure contain-
2
2 7
ing edge-sharing MnO octahedra, forming (MnO ) layers
3 n
6
connected together by V O 4~ divanadate groups, which
2
7
have a staggered conformation and linear VÈO ÈV.14 The
b
structure of a-Mn V O is a distorted form of thortveitite b-
2
2 7
Mn V O , the distortion being mainly due to the bending of
2 7
the VÈO ÈV angle in the V O moieties. This results in a tri-
2 7
clinic cell with a double volume.
2
2 6
2 2 7
973 K the DTA curve also shows a small peak, likely due to
the a ] b phase transition; our sample is predominantly com-
posed of b-Cu V O and this explains the small intensity of
b
2
2 7
3.1.2. Surface areas and DTA-TG studies. The measured
this peak. In the range 750È900 K two strong endothermic
peaks are present (at 1055 and 1079 K), which correspond to a
weight loss of ca. 5%. These are likely due to the melting of
b-Cu V O reported to occur at 1053 K18,20 and to its
BET surface areas of the samples calcined at 773 K are sum-
marized in Table 1. In the case of MnV and MgV materials
powders prepared by the citrate method have distinctly higher
surface area, while for CuV both powders have a surface area
lower than 10 m2 g~1.
2
2 7
further decomposition which is associated to oxygen loss.
The DTA-TG curves upon cooling shows six exothermic
peaks and a continuous weight growth in the range 1120È673
K. The higher temperature DTA peaks can be attributed to
the reverse phenomena: reabsorption of oxygen by the melt
(1010 K), solidiÐcation of b-Cu V O (980 K) and solidiÐca-
The DTA curves of MgV1 show upon heating two strong
endothermic peaks at 1057 and 1196 K (Fig. 2). This fully
agrees with the DTA curve reported by Clark and Morley16
that showed they are due to a ] b and b ] c polymorphic
phase transformations. In agreement with this, the TG curves
do not present any weigh variation. Again in agreement with
Clark and Morley16 the DTA curve upon cooling shows com-
plete reversibility of the b ] c polymorphic phase transform-
ation (evidenced by an exothermic peak at 1158 K during
cooling) while the a ] b phase transformation is only slightly
evident as a very broad exothermic peak near 873 K. These
data show that our material is actually stoichiometric
Mg V O , and also justify the presence of the b phase which
2
2 7
tion of the CuV O ÈCu V O eutectic (860 K). The lower
2
6
2 2 7
temperature peaks (a triplet in the range 550È450 K) could be
due to the solidiÐcation of Cu`ÈV5` phases (as for example
Cu V O , Cu VO , CuVO 21) in agreement with the trend
2
2 6
3
4
3
of the TG curve which shows a progressive weight gain.
The sample obtained with the coprecipitated method shows
the same behavior at the TG-DTA analysis with small shifts
of the peaks and changes in the relative intensities. According
to the above discussion and to XRD data, the peak at 973 K,
due to the a-Cu V O ] b-Cu V O phase transition is a
2
2 7
tends to stay as a metastable phase also below the transition
temperature.
2
2 7
2 2 7
little more intense here than for CuV1. These data suggest
that small amounts of CuV O , not observed by XRD, are
also present in the samples.
The behavior of the coprecipitated sample has the same
trend. In fact the same endothermic peaks, as described above,
are observed in the DTA curve at 1044 and 1196 K upon
heating and an exothermic one, on cooling. According to
Clark and Morley16 and to Kerby and Wilson19 the endo-
thermic peak at 1044 K is due to the peritectic melting of
MgV O , which is in fact observed in the XRD spectrum, and
2
6
The DTA and TG curves of both MnV materials do not
show any peak or weight change in the range 773È1250 K, in
agreement with the stability of b-Mn V O from the a ] b
2
2 7
transition temperature (296 K) up to the melting point near
1350 K.14,22
2
6
to the a ] b phase transition of Mg V O ; while the other
2
2 7
endothermic peak, at 1196 K, in agreement with Clark and
Morley16 is due to the b ] c polymorphic phase transform-
ation; so on cooling, the exothermic peak at 1158 K is due to
3.1.3. UV-vis spectra. Fig. 3 shows the di†use reÑectance
UV-vis spectra of the metal divanadate samples. The spectra
of the two MgV are similar showing a strong adsorption at
Phys. Chem. Chem. Phys., 2000, 2, 2039È2045
2041

View Article Online
have far more complex electronic structures dominated by the
dÈd transitions of Mn2` (d ) and of Cu2` (d ). The spectra
5
9
suggest that these transition metal divanadates also present a
shift of the charge transfer transitions towards lower energies
resulting in a lower energy gap.
3
.1.4. Skeletal FT-IR spectra. The results of the FT-IR and
FT-FIR experiments are reported in Fig. 4. An interpretation
can be given in agreement with previous work.25,26 In partic-
ular, the strong absorptions found in the region above 750
cm~1 are assignable to the stretchings of the terminal VÈO
bonds of the pyrovanadate units, while the features in the
region 750È500 cm~1 should be predominantly due to the
asymmetric and symmetric stretchings of the VÈOÈV bridges.
The multiple peaks in the region below 500 cm~1 are due to
the deformations of the VÈO bonds as well as to the lattice
modes predominantly due to MÈO (M \ Mg, Cu or Mn)
stretching modes.
The spectra of MgV samples are in full agreement with lit-
erature data25,27 for the a-Mg V O phase, for which a com-
2
2 7
plete interpretation has also been reported in ref. 25. The only
di†erences are the presence of a shoulder at 1025 cm~1 and a
very small peak at 530 cm~1 evident in the spectrum of the
sample prepared by the citrate method. These are likely due to
the presence of traces of V O possibly in an amorphous
2
5
state28 and so not detectable with XRD powder di†raction.
In the case of CuV samples the two spectra are quite di†er-
ent, especially in the region above 700 cm~1 where inversions
in the intensity ratios of some bands and slight shifts are
observed. Furthermore, in the spectrum of the coprecipitate
sample a shoulder can be evidenced at 1025 cm~1 likely due
to the presence of amorphous V O . Thus, the spectra show
2
5
di†erences in the phase composition of the two samples, while
the strength of the absorptions in the region 750È450 cm~1
Fig. 3 UV-vis spectra of samples: (a) MnV, (b) CuV and (c) MgV.
00È315 nm with an onset near 500 nm, according to its pale
3
yellow color. This spectrum agrees fairly well with that report-
ed by Hanuza et al.23 This absorption is certainly associated
to the O2~ ] V5` charge transfer transitions involving the
tetrahedrally coordinated vanadium centers.
The electronic spectra of the coprecipitated CuV sample
shows a main peak at 415 nm with components at 260 nm
and a wide peak in the range 700È1300 nm. The absorptions
of CuV1 are shifted at slightly higher wavelengths, i.e. 315 and
450 nm while the higher wavelength wide band is deÐnitely
stronger and centered at 860 nm. The spectra agree with the
deep brown color of the powders. The lower wavelength
absorption is due to charge transfer transitions involving both
Cu2` and V5` empty orbitals. The higher wavelength absorp-
tion is to be assigned to dÈd transitions of Cu2`. The di†erent
intensity of the dÈd band and the shift of the charge transfer
absorption can be associated to the di†erent composition in
terms of a and b polymorphs (which have a di†erent coordi-
nation for Cu2`) and of crystallinity.
The UV-vis spectrum of MnV1 presents two resolved com-
ponents at 420 and 625 nm, while that of MnV2 presents an
unresolved absorption centered at 430 nm with a pronounced
shoulder near 600 nm. The position of these absorptions lie in
the typical range of dÈd transitions of Mn2` in oxide environ-
ments.24 It seems likely that the broad absorption also con-
tains at its low wavelength side the charge transfer transitions.
These spectra reveal the strongly di†erent electronic struc-
ture of the materials under study. According to its spectrum,
Mg divanadate, involving only d metal centers, is expected to
0
behave as a semiconductor with an approximate energy gap of
2.5 eV (500 nm). On the contrary, Mn and Cu divanadates
Fig. 4 FT-IR spectra of samples: (a) MgV, (b) CuV and (c) MnV.
2042 Phys. Chem. Chem. Phys., 2000, 2, 2039È2045

View Article Online
agrees with the bent structure of the VÈOÈV bridge of the
pyrovanadate units in both polymorphs a-Cu V O and b-
homogeneous reactions. For this reason we stopped these
experiments at 910 K.
The propene productivity under these conditions is deÐ-
2
2 7
Cu V O . Interestingly, we can note that the number of
2
2 7
peaks identiÐable in the spectrum of the sample prepared by
the citrate method (Fig. 4, spectrum b), nearly corresponds to
the number of IR active modes forecast on the basis of the
crystal structure of b-CuV O . The unit cell of b-Cu V O in
nitely low, 0.067 kg
kg ~1 h~1. To see if higher pro-
C3H6
Cat
ductivities are possible, we also performed experiments at
higher space velocities, by using a total Ñow of 300 ml min~1
corresponding to q \ 2.2 s. In these cases the conversion is
smaller, as expected. On the other hand, the selectivities to
2
7
2 2 7
fact contains the two smallest Bravais cells, so that the
number of atoms in the smallest Bravais cell is 22, which cor-
responds to 3 acoustic modes and 63 optical modes. 30 IR
active modes are expected (A and B symmetries).
both propene and CO is also lower while the selectivity to
x
ethylene increases strongly. However, no methane is copro-
duced in the range up to 880 K. This suggests that a way to
ethylene from propane exists which is parallel both to oxy-
u
u
In the case of MnV samples the two spectra are quite di†er-
ent too, in agreement with the di†erent a-Mn V O to b-
dehydrogenation
ethylene ] methane. This way could be intermediate to CO
formation. By further increasing the temperature oxygen con-
version becomes total and methane is also produced in signiÐ-
cant amounts. Selectivity to ethylene rises 53%, that to
propene is less than 15%. In these conditions, the maximum
yield to propene is 12.7%, the productivity to propene is 0.195
to
propene
and
cracking
to
2
2 7
Mn V O ratio observed by XRD. The spectrum of the cop-
2 7
recipitated sample (nearly pure b-Mn V O ) is simpler also
2
x
2
2 7
with respect to those of Cu and Mg pyrovanadates. This is in
agreement with very small dimension of the smallest Bravais
cell in this crystal structure. We actually can observe at least
13 IR bands. Here we can observe the absence of deÐnite
peaks in the region 500È700 cm~1 likely due to the linearity of
kg
C3H6 Cat
in similar conditions with vanadiaÈalumina (3.3 kg
kg ~1 h~1, still very low with respect to that obtained
the VÈOÈV bridges, typical of the b-Mn V O phase. The
2
2 7
C3H6
spectrum of the citrate sample shows, instead, a weak peak at
kg ~1 h~112) which has, however, much higher surface area.
Cat
5
15 cm~1 and a shoulder at 638 cm~1, likely due to the sym-
Maximum yield to oleÐns is near 57% and productivity to
metric and asymmetric stretching modes, respectively, of the
oleÐns is near 0.6 kg
kg ~1 h~1. The apparent activation
C3H6 Cat
bent VÈOÈV bonds of a-Mn V O .
energy for propane consumption we measured with q \ 2.2 s
is near 60 kcal mol~1. This value is much too high for a true
catalytic reaction for which activation energies of the order of
2
2 7
3.2. Catalytic activity tests
15È35 kcal mol~1 are usually observed.30h34 On the contrary,
Catalytic tests have been performed with the catalysts produc-
ed by the citrate method that are all almost exclusively consti-
tuted of pyrovanadates and have mostly higher surface areas.
The catalytic activity of the Mg pyrovanadate in the conver-
sion of propane in oxygen with q \ 6.7 s (see experimental) is
summarized in Fig. 5 (left-hand side). The propane conversion
rises from less than 20% to more than 40% in the T range
this value is more typical for radical gas-phase processes such
as propane oxidation by dioxygen and propane cracking.35
The increase of the rate of the gas-phase processes relative to
surface catalysed processes can be due to an increase of the
temperature in the empty section of the reactor, according to
the larger amount of propane converted at higher Ñows.
Copper divanadate is poorly active showing a conversion
not exceeding 5% at 887 K and the highest contact times (Fig.
775È910 K. However, the conversion of oxygen grows only up
to 830 K (74%) and later decreases a little bit. The selectivity
to propene is quite stationary in the range between 22 to
slightly more than 30%, so that the best yield is 13%. This
behavior suggests that above 830 K we fall under di†usion
limited conditions and that, in this range, propene can be pro-
duced not only by oxydehydrogenation but also by pure
dehydrogenation without oxygen consumption. The per-
formances at the highest temperature fully agree with those
summarized by Kung6 as typical for Mg vanadates, showing
that our preparation has a similar behavior as that reported
by other authors. The main product in the overall 775È910 K
range is CO2 (up to 47% selectivity) whose selectivity
decreases at the highest temperatures when ethylene pro-
duction becomes important (9% selectivity). CO selectivity is
quite stationary in the range 24È29%. As reported pre-
viously,29 the production of ethylene is evidence of gas-phase
6
, left). Selectivity to propene does not exceed 40% at signiÐ-
cant conversion. By decreasing contact time and increasing
reaction temperature (Fig. 6, right) the selectivity to propene
stays constant in the range 30È40% while selectivity to ethyl-
Fig. 6 Catalytic activity of CuV1 for the oxidation of the propane at
q \ 6.7 s (left) and q \ 2.2 s (right). Symbols deÐned as in Fig. 5.
Fig. 5 Catalytic activity of MgV1 for the oxidation of the propane at
q \ 6.7 s (left) and q \ 2.2 s (right): È+È O (C), È=È CO (S),
2
ÈÈ>È CO (S), ÈÈ]È CH (S), È]]È C H (S), È…È C H (S),
Fig. 7 Catalytic activity of MnV1 for the oxidation of the propane at
q \ 6.7 s (left) and q \ 2.2 s (right). Symbols deÐned as in Fig. 5.
2
4
2
4
2 6
ÈoÈ C H (S) and È=È C H (C). C \ conversion; S \ selectivity.
3
6
3 8
Phys. Chem. Chem. Phys., 2000, 2, 2039È2045
2043

View Article Online
ene grows again up to above 50% when conversion rises 40%.
Methane is also produced with selectivity almost comparable
to that of ethylene, showing that cracking is predominant. The
activation energies for propane consumption at q \ 6.7 s is 34
kcal mol~1, which can be attributed to a true solid catalysed
process over poorly active centers.34 Working with q \ 2.2 s
the apparent activation energy for propane consumption is
very high, near 61 kcal mol~1, providing evidence for the pre-
dominancy of gas-phase phenomena. Comparing the per-
formances of our CuV sample with respect to the MgV sample
we have clearly a lower catalytic activity but also a lower
surface area of the catalyst. By comparing the behavior at the
same conversion, copper divanadate appears to be more selec-
tive to propene than magnesium divanadate. As for example,
working with q \ 2.2 s at the conversion of near 5% we
obtain a selectivity to propene near 40% (at 855 K) on copper
divanadate and near 20% (at 835 K) on Mg divanadate.
Mn divanadate is, in spite of a similar surface area, by far
less active than Mg divanadate working at the higher contact
times (Fig. 7). In agreement with this, the apparent activation
energy is, in these conditions, 36 kcal mol~1. Additionally it is
less selective to propene than copper divanadate. Also in this
case selectivity to propene does not grow very much by
decreasing contact time (still below 40%) while ethylene and
methane production grows strongly, with a ratio a little higher
than 1. The apparent activation energy in these conditions is
vanadates to vanadium oxide redox centers. We show here
that composing the divanadate species with oxidizable biva-
lent centers such as Mn2` does not improve the catalytic
activity. On the other hand it has been shown previously that
composing vanadium with manganese oxides does not
improve the performances also in the case of alumina sup-
ported catalysts.12
On the contrary, composing the divanadate species with
quite easily reducible divalent centers such as Cu2` seems to
have little e†ect on the intrinsic activity per surface area
except to improve slightly the selectivity to propene.
The data presented here show that true surface-catalysed
reactions are observed at relatively high contact time (q \ 6.7
s) in the presence of propane and oxygen and suggest that in
this range it is possible to improve the performances of vana-
dium oxide-based oxydehydrogenation catalytic systems by
composing it with reducible cations. At lower contact times
gas-phase phenomena become more important, possibly due
to an increase of the temperature in the empty section of the
reactor, but the catalyst composition still has a role in deter-
mining conversion and product distribution. Such gas-phase
phenomena contribute in producing further propene from
propane but increase signiÐcantly the production of another
potentially useful product such as ethylene.
Acknowledgements
The authors acknowledge MURST (Rome, Italy) and The
University of Genova for funding (CoÐnanziamento nazionale
8
5 kcal mol~1 which is similar to that expected for propane
gas-phase cracking. In any case, selectivity to propene and to
oleÐns is higher, at the same conversion level, with respect to
that observed on Mg divanadate.
1999).
To compare the catalysts activity, we measured the propane
conversion rate in true catalytic conditions i.e. with q \ 2.2 s
References
(
calculated on the basis of the same catalyst weight), at
T \ 830È840 K. The measured values are 0.0868 ] 10~3
1
2
3
S. Albonetti, F. Cavani and F. TriÐro
38, 413.
`
, Catal. Rev. Sci. Eng., 1996,
mol
m
~2 h~1 for MgV1, 0.0250 10~3 mol
m
~2
C3H8 Cat
C3H8 Cat
M. M. Bettahar, G. Constentin, L. Savary and J. C. Lavalley,
Appl. Catal. A, 1996, 145, 1.
h~1 in the case of CuV1 and Ðnally 0.0161 ] 10~3 mol
C3H8
m~2 h~1 for the sample MnV1. It is thus conÐrmed that mag-
nesium divanadate is by far the most active, although copper
divanadate is more selective at the same conversion level.
J. Cosyns, J. Chodorge, D. Commereuc and B. Torck, Hydrocar-
bons Process., 1998, 77, 61.
4
5
F. Cavani and F. TriÐro, Catal. T oday, 1999, 51, 561.
`
M. A. Chaar, D. Patel, M. C. Kung and H. H. Kung, J. Catal.,
1
987, 105, 483.
4. Conclusions
6
(a) H. K. Kung, Adv. Catal., 1994, 40, 1; (b) H. H. Kung and
M. C. Kung, Appl. Catal. A, 1997, 157, 105.
Fairly pure Mg, Cu and Mn divanadates have been prepared
by the citrate method, although these preparations can
contain mixtures of the polymorphs. According to XRD, the
Mg vanadate material produced by coprecipitation is impure
from the metavanadate. DTA-TG analyses and skeletal IR,
however, allow us to characterize also the presence of impu-
rities not detected by XRD. The electronic structures of these
materials are deÐnitely di†erent as is evidenced by UV-vis
7
8
D. Sam, V. Soenen and J. C. Volta, J. Catal., 1990, 123, 417.
K. Seshan, H. M. Suan, R. H. H. Smits, J. G. van Ommen and
J. R. H. Ross, in New developments in selective oxidation, ed.
G. Centi and F. TriÐro, Elsevier, Amsterdam, 1990, p. 505.
`
9
0
J. G. Eon, R. Olier and J. C. Volta, J. Catal., 1994, 145, 318.
J. M. Lopez Nieto, R. Coenraads, A. Dejoz and M. I. Vasquez, in
Proceedings of the 3rd W orld Congress on Oxidation Catalysis, ed.
R. K. Grasselli, S. T. Oyama, A. M. Ga†ney and J. E. Lyons, San
Diego, Elsevier, Amsterdam, 1997, p. 443.
1
spectroscopy, as a consequence of the d structure of Mg2`
11 B. Grzybowska-Swierkosz, Appl. Catal. A, 1997, 157, 263.
12 V. Ermini, E. Finocchio, S. Sechi, G. Busca and S. Rossini, Appl.
Catal., A, in the press.
0
and V5` ions and of the d and d nature of Cu2` and Mn2`
9
5
cations, respectively.
1
3
G. Centi, F. TriÐro
Rev., 1988, 88, 55.
`
, J. B. Ebner and V. M. Franchetti, Chem.
The catalytic activity shows that Mg divanadate is the most
active in converting propane in the presence of oxygen.
However, the yield and productivities to propene are relatively
low, due to a low intrinsic activity with respect to supported
vanadia catalysts, and also to a quite low selectivity. An oxi-
dative way to ethylene and the thermal cracking to ethylene
plus methane are the predominant byreactions, largely
occurring in the gas phase, in particular at low contact times.
Copper divanadate is less active mainly due to its low
surface area. However, it is more selective to propene at the
same conversion levels. Mn divanadate is less active than Mg
divanadate but shows higher selectivity to propene at the
same conversion levels.
1
4
J. H. Liao, F. Leroux, C. Payen, D. Guyomard and Y. Pi†ard, J.
Solid State Chem., 1996, 121, 214.
15 R. Gopaj and C. Calvo, Acta Crystallogr., Sect. B, 1974, 30, 2491.
16 G. M. Clark and R. Morley, J. Solid State Chem., 1976, 16, 429.
17 E. E. Sauerbrei, R. Faggiani and C. Calvo, Acta Crystallogr., Sect.
B, 1974, 30, 2907.
1
8
D. Mercurio-Lavaud and B. Frit, Acta Crystallogr., Sect. B, 1973,
9, 2737.
2
1
9
R. C. Kerby and J. R. Wilson, Can. J. Chem., 1973, 51, 1032.
20 (a) P. Fleury, C. R. Acad. Sci., Ser. C, 1966, 263, 1375; (b) P.
Fleury, Rev. Chim. Miner., 1969, 6, 819.
21 L. F. MalÏtseva and A. A. Fotiev, Russ. J. Inorg. Chem., 1978, 23,
1
096.
2
2
2
3
A. A. Fotiev and L. L. Surat, Russ. J. Inorg. Chem., 1982, 27, 608.
J. Hanuza, B. Jezowska-Trzebiatowska and W. Oganowski, J.
Mol. Catal., 1985, 29, 109.
Since Mg2` has virtually non-reducible and non-oxidizable
metal centers, it is reasonable (also on the basis of the large
amount of work done in the Ðeld of alkane oxidation on V-
containing oxides) to attribute the catalytic activity of Mg
24 W. Sjoerd Kijlstra, E. K. Poels, A. Bliek, B. M. Weckhuysen and
R. A. Schoonheydt, J. Phys. Chem., 1997, 101, 309.
2044
Phys. Chem. Chem. Phys., 2000, 2, 2039È2045

View Article Online
2
5
6
G. Busca, G. Ricchiardi, D. Siew Hew Sam and J. C. Volta, J.
Chem. Soc., Faraday T rans., 1994, 90, 1161.
G. Busca and C. Resini, in Encyclopedia of Analytical Chemistry,
Instrumentation and Application, ed. R. A. Meyers, Wiley, New
York, 2000, in the press.
30 J. N. Michaels, D. L. Stern and R. K. Grasselli, Catal. L ett., 1996,
42, 139.
31 Y.-F. Y. Yao, Ind. Eng. Chem. Prod. Res. Dev., 1980, 19, 293.
32 K. Otto, J. M. Andino and C. L. Parks, J. Catal., 1991, 131,
243.
2
2
2
2
7
8
9
X. Gao, P. Ruiz, Q. Xin, X. Guo and B. Delmon, Catal. L ett.,
33 C. P. Hubbard, K. Otto, H. S. Gandhi and K. Y. S. Ng, J. Catal.,
1
994, 23, 321.
1993, 139, 268.
C. Sanchez, J. Livage and G. Lucazeau, J. Raman Spectrosc.,
34 M. Baldi, V. Sanchez-Escribano, J. M. Gallardo-Amores, F.
Milella and G. Busca, Appl. Catal. B, 1998, 17, L175.
35 C. K. Westbrook and J. P. William, Combust. Sci. T echnol., 1984,
37, 117.
1
982, 12, 68.
V. Ermini, E. Finocchio, S. Sechi, G. Busca and S. Rossini, Appl.
Catal. A, 2000, 190, 157.
Phys. Chem. Chem. Phys., 2000, 2, 2039È2045
2045