Towards physical descriptors of active and selective catalysts for the oxidation of n-butane to maleic anhydride

Article abstract of DOI:10.1002/cctc.201200953

Based on our newly developed microwave cavity perturbation technique, the microwave conductivity of diverse vanadium(III), (IV), and (V) phosphate catalysts was measured under reaction conditions for the selective oxidation of n-butane. The conductivity response on the gas phase was identified as a very sensitive measure for the redox kinetics, reversibility, and stability of the samples, which are important prerequisites for highly selective and active catalysts. The sensitivity achieved by our method was comparable to surface-sensitive methods such as X-ray photoelectron spectroscopy, whereas more conventional analytic techniques such as X-ray diffractometry or Raman spectroscopy only indicated the stability of the bulk crystal phase under the same reaction conditions.

Full text of DOI:10.1002/cctc.201200953

CHEMCATCHEM
FULL PAPERS
DOI: 10.1002/cctc.201200953
Towards Physical Descriptors of Active and Selective
Catalysts for the Oxidation of n-Butane to Maleic
Anhydride
Maik Eichelbaum,*[a] Robert Glaum,[b] Michael Hꢀvecker,[a, c] Knut Wittich,[b] Christian Heine,[a]
Heiner Schwarz,[a] Cornelia-Katharina Dobner,[d] Cathrin Welker-Nieuwoudt,[d]
Annette Trunschke,[a] and Robert Schlçgl[a]
Based on our newly developed microwave cavity perturbation
technique, the microwave conductivity of diverse vanadium(III),
(IV), and (V) phosphate catalysts was measured under reaction
conditions for the selective oxidation of n-butane. The conduc-
tivity response on the gas phase was identified as a very sensi-
tive measure for the redox kinetics, reversibility, and stability of
the samples, which are important prerequisites for highly se-
lective and active catalysts. The sensitivity achieved by our
method was comparable to surface-sensitive methods such as
X-ray photoelectron spectroscopy, whereas more conventional
analytic techniques such as X-ray diffractometry or Raman
spectroscopy only indicated the stability of the bulk crystal
phase under the same reaction conditions.
Introduction
Heterogeneously catalyzed selective oxidation reactions are of
particular importance, as 25% of all industrial organic chemi-
cals and intermediates are produced by this way.[1,2] In this
contribution, we will focus on the selective oxidation of n-
butane to maleic anhydride (MA) over vanadium phosphorus
oxides. This reaction was the first example of a commercially
applied selective oxidation of an alkane to produce a bulk
chemical,[3] and hence is of both academic and industrial rele-
vance. The partial oxidation to MA is catalyzed by chemically
relatively simple and homogeneous compounds such as the in-
dustrially used phase-pure vanadyl pyrophosphate (VO)2P2O7
(VPP). Accordingly, the bulk properties of the material are very
well known making it an ideal system for the investigation of
the working mode of a high-performance catalyst. However,
despite tedious optimization attempts, VPP only allows for MA
yields not higher than 65 mol%. Albeit the great effort made
in studying this catalyst, its atomic or electronic surface struc-
ture, which is, in particular, the reason why VPP comprises
such a unique (though not perfect) catalytic system, for exam-
ple, in comparison to alternative vanadium phosphorus oxides,
is still a matter of intense debates.[4–13]
The working mode of VPP has frequently been described as
the Mars–van Krevelen (MvK) mechanism or its kinetics by the
MvK rate expression.[4,5,9,14] Although the original work by Mars
and van Krevelen did not include any atomic description of
the elementary steps,[15] it is nowadays very often associated
with the participation of lattice oxygen from the catalyst in the
oxidation reaction. Herein, we will use the term lattice oxygen
for oxygen atoms that induce oxygen vacancies in the metal
oxide catalyst upon their removal, as suggested by Morrison,[16]
though the discrimination from adsorbed oxygen is not always
straightforward. Within the MvK mechanism, which should be
rather called redox mechanism, the catalyst has to fulfill several
functions. First, the catalyst surface has to provide binding
sites for the substrate and intermediates. Second, the catalyst
has to be able to exchange oxygen with the gas phase, as the
catalyst, or more precisely its active site, has to adsorb molecu-
lar oxygen and convert (reduce) it to an active form that is
able to oxidize the alkane. For this oxygen reduction process,
the catalyst has to provide electrons, which eventually have to
be transferred back to the catalyst before the oxidation prod-
uct can desorb into the gas phase. Consequently, the third pre-
requisite for an oxidation catalyst is the ability for electron ex-
change, whereupon one site could in principle manage all
three functions. Both the capability of redox exchange (i.e.,
[a] Dr. M. Eichelbaum, Dr. M. Hꢀvecker, C. Heine, H. Schwarz, Dr. A. Trunschke,
Prof. R. Schlçgl
Department of Inorganic Chemistry
Fritz-Haber-Institut der Max-Planck-Gesellschaft
Faradayweg 4–6, 14195 Berlin (Germany)
Fax: (+49)(0)3084134401
E-mail: me@fhi-berlin.mpg.de
[b] Prof. R. Glaum, K. Wittich
Institute for Inorganic Chemistry
University of Bonn
Gerhard-Domagk-Straße 1, 53121 Bonn (Germany)
[c] Dr. M. Hꢀvecker
Solar Energy Research
Helmholtz-Zentrum Berlin/BESSY II
Albert-Einstein-Straße 15, 12489 Berlin (Germany)
[d] C.-K. Dobner, Dr. C. Welker-Nieuwoudt
Process Research and Engineering
BASF SE
Carl-Bosch-Straße 38, 67056 Ludwigshafen (Germany)
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cctc.201200953.
ꢁ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 2318 – 2329 2318

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
electron exchange) and exchange of surface oxygen should be TPO/TPR techniques to validate its structural and electronic
prerequisites for selective oxidation catalysts as defined within properties. Notably, MA, CO, and CO2 were the only products
the MvK concept. Experimental evidence has often been based found in the oxidation of n-butane with VPP.
on analyses of near-stoichiometric reactions of the catalyst
In Figure 1a the microwave conductivity s and the real part
with either oxygen or the alkane[4,14,17,18] or reactions under un- of the permittivity e1 of VPP is shown, starting with a gas mix-
usual conditions (e.g., pulse mode, vacuum).[8,19] However, such ture containing 1.4% n-butane and 20% O2, followed by a reac-
experiments do not necessarily describe the mode of opera- tion feed with lower oxygen content (10% O2, 1.4% n-butane).
tion under commercial steady-state reaction conditions (except As a result, the simultaneously measured n-butane conversion
for in the riser reactor), that is, in the presence of both oxygen dropped from 32 to 26% whereas the selectivity to MA in-
and the substrate, probably defining a different chemical and creased only slightly from 74 to 76% upon changing from the
electronic surface structure (and working mode) of the first to the second gas mixture (Figure 1b). Regarding conduc-
catalyst.[20]
tivity, a small decrease from 4.3 to 4.1ꢂ10ꢀ5 Scmꢀ1 was ob-
Recently, the microwave cavity perturbation technique served between these two conditions (Figure 1a). Furthermore,
(MCPT) has been introduced to the field of catalysis as a non-
contact alternative to conventional 2-contact conductivity
measurements for probing charge transfer and redox proper-
ties of catalysts.[21–23] The contactless and integral nature of this
method allowed the reaction to be conducted in a convention-
al fixed-bed flow-through reactor with a powder catalyst under
high-performance catalytic (nearly technical) conditions with
deduction of the electrical and dielectric material properties of
the catalyst during the reaction. Herein, we applied MCPT to
determine the complex permittivity and conductivity of differ-
ent representative vanadium(III), (IV), and (V) phosphorus
oxides under n-butane oxidation conditions: VIIIPO4,
(VIVO)2P2O7, Ag6(VIVO)2(PO4)2(P2O7), Ag2VIV,VP1.6O7+d
,
Ag(VIVO)-
(VVO)(PO4)2, b-VVOPO4, and a-VVOPO4. The aim of this study
was to validate the two “pillars” of selective oxidation cataly-
sis,[24] lattice oxygen exchange and redox behavior, and to
identify electronic properties that are characteristic for selec-
tive n-butane oxidation catalysts. By measuring the microwave
conductivity at different chemical potentials of oxygen and n-
butane, the nature of band bending,[16] hence, the majority
charge carrier type, was identified for each sample. The con-
ductivity was also measured under the extreme cases of
strongly oxidizing (in air) and reducing (n-butane in N2) condi-
tions, respectively. However, such extreme oxidized and re-
duced states could be directly related to the electronic state of
the working catalyst under normal reaction conditions and cor-
related with the catalytic properties of the studied samples. In
particular, the reversibility and the kinetics of the observed
redox processes were studied, thus defining physical descrip-
tors for selective catalysts that go beyond the simple terms
redox and oxygen exchange capability. As for the established
catalyst VPP, these results were rationalized by complementary
analytical techniques such as in situ near-edge X-ray absorp-
tion fine structure spectroscopy (NEXAFS), in situ Raman spec-
troscopy and temperature-programmed oxidation/reduction
(TPO/TPR).
Figure 1. a) Microwave conductivity s, real part of the permittivity e1, and
b) catalytic performance of VPP in different gas atmospheres at 3858C with
GHSV=4240 cmꢀ1. Gas mixtures: C4/O2 (1:14): 1.4% n-C4H10, 20% O2; C4/O2
(1:7): 1.4% n-C4H10, 10% O2; O2: 20% O2; C4: 1.4% n-C4H10; residual gas: N2.
a strongly oxidizing atmosphere containing 20% O2 in N2 had
a severe impact on the VPP conductivity, inducing an increase
to 6.7ꢂ10ꢀ5 Scmꢀ1 within 1 h. The subsequent treatment in
a reducing atmosphere of 1.4% n-butane in N2 initially re-
versed this behavior, that is, the conductivity fell back to 4.0ꢂ
10ꢀ5 Scmꢀ1. However, this was followed by an increase to 6.0ꢂ
Results and Discussion
Benchmark catalyst vanadyl pyrophosphate
The benchmark catalyst vanadyl pyrophosphate was investi- 10ꢀ5 Scmꢀ1. The conductivity measurements during the final
gated. A technical sample provided by BASF SE was studied by treatment in the lean reaction gas mixture disclosed a fully re-
in situ MCPT, in situ NEXAFS, in situ Raman spectroscopy, and versible behavior of the electronic properties of the catalyst,
ꢁ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 2318 – 2329 2319

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
that is, it exhibited nearly the same conductivity as it was ob-
served at the beginning of the reaction. Regarding the catalyt-
ic properties, n-butane conversion and MA selectivity (30 and
74%, respectively; Figure 1b) were in a similar range to those
observed in the first cycle under lean reaction conditions,
except for a slight deactivation of the activity from 35 to 30%
n-butane conversion over the whole measurement time. Nota-
bly, the real part of the permittivity e1 remained rather con-
stant over the whole reaction time irrespective of the gas at-
mosphere, exhibiting values between 11 and 12, the small var-
iations of which were within the measurement error. These re-
sults suggest that the lattice of the bulk catalyst remained
stable upon the different treatments and that the conductivity
changes (owing to bound and/or free charge carriers) were
not caused by bulk phase transformations. Notably, no
changes in the X-ray powder diffractograms and lattice param-
eters of VPP were observed under oxidizing or reducing condi-
tions (see the Supporting Information, Table S1), which gives
further evidence for the stability of the bulk phase under all
conditions applied.
The measured conductivity values were at the upper limit of
the VPP conductivity measured with alternating current con-
tact methods, which usually give values between 10ꢀ5 and
10ꢀ9 Scmꢀ1 under similar conditions.[25–29] However, compari-
sons of conductivity measurements are not straightforward, as
a strong frequency dependence of the conductivity has to be
considered if a charge-carrier hopping mechanism occurs,[30] as
has been suggested for V2O5.[31]
Figure 2. a) In situ VL3-edge NEXAFS spectra of VPP in different gas mixtures
after the respective TOS; b) PTR–MS signal of MA and applied temperature
program and c) first momenta of the normalized spectral distributions
(Ecenter =sE·TEY(E)dE) at the VL3-edge of VPP in different gas mixtures. Gas
mixtures: C4/O2 (1:10): 0.5 standard cubic centimeters per minute (sscm) n-
C4H10, 5.0 sccm O2; C4/O2 (1:4): 0.5 sccm n-C4H10, 2 sccm O2, 3 sccm He;
O2: 5.0 sccm O2, 0.5 sccm He; C4: 0.5 sccm n-C4H10, 5.0 sccm He.
The behavior of a conductivity increase in oxidizing gases
and decrease in reducing gases (except for the subsequent rise
after longer reducing times that will be explained later) can be
described best by a p-type semiconductor band model[16,32]
and is in perfect agreement with the behavior observed by
contact methods.[25,26,28,33] This dynamic conductivity can be ex-
plained by the exchange of lattice oxygen (OxO), the associated
formation/filling of oxygen vacancies (VOꢁꢁ), and the coupled
generation/depletion of electron holes (hꢁ), which represent
the majority charge carriers [as described in Krçger–Vink nota-
tion[32] in Eqs. (1) and (2), assuming that all involved oxygen
atoms and charge carriers are supplied by the catalyst]:
states of the catalyst in different atmospheres. The first mo-
menta of the V L3-edge spectra (Ecenter =sE·TEY(E)dE; TEY=total
electron yield) in different gas mixtures are plotted in Fig-
ure 2c. As a result, the oxidation state was highest under oxi-
dizing, and lowest under reducing conditions, whereas inter-
mediate values of the first momentum were obtained under
alkane-poor and rich reaction conditions. The differences in
the first momenta between alkane-poor (C4/O2=1:10) and
rich (C4/O2=1:4) conditions were small. This result is in good
agreement with the in situ X-ray photoelectron spectroscopy
(XPS) results of VPP published earlier,[23] which disclosed a gas-
phase-dependent average surface V oxidation state between
4.0 and 4.3 measured at a photoelectron kinetic energy of
160 eV corresponding to an information depth of 1 nm. More-
over, these results resembled the dynamic microwave conduc-
tivity behavior shown in Figure 1. NEXAFS in the total-electron-
yield mode was generally more bulk-sensitive than XPS. The
first momentum of the V L3-edge spectrum changed by
0.16 eV between reaction conditions O2 and C4 (Figure 2c),
which corresponds in a first approximation[34] to a change in
the oxidation state by 0.1–0.2 (e.g., from 4.0 to 4.1–4.2). This
observed change is in agreement with the maximum oxidation
1O2ðgÞ þ Vꢁꢁ ! OxO þ 2hꢁ
C4H10 þ 7OxO þ 14hꢁ ! C4H2O3 þ 4H2O þ 7VOꢁꢁ
ð1Þ
ð2Þ
2
O
In addition, the results of the in situ V L3-edge NEXAFS
measurements with the same catalyst under similar reaction
conditions are depicted in Figure 2a, and the results of the si-
multaneous measurements of the MA production rate (detect-
ed by proton-transfer mass spectrometry, PTR–MS) are dis-
played in Figure 2b. Both measurements proved that the cata-
lyst was indeed working despite the reduced pressure condi-
tions (total reaction cell pressure: 0.5 mbar). It has been shown
earlier that the overall peak position at the V L3-absorption
edge is related to the formal oxidation state of vanadium (the
higher the first momentum Ecenter, the higher the oxidation
state),[34] and thus can be used to compare the V oxidation
ꢁ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 2318 – 2329 2320

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
state changes between 4.0 and 4.15 determined from a V2p3/2
XPS spectrum measured at a photoelectron kinetic energy of
760 eV corresponding to an information depth of 4 nm.[23] This
indicates a similar information depth for the NEXAFS experi-
ment, which can be expected if Auger electrons contribute pri-
marily to the signal. These results and the corresponding con-
ductivity behaviors suggest that the concentration of electron
hole charge carriers is in equilibrium with the V4+/V5+ ratio, as
indicated in Equation (3):
peared, proving that the sample consisted of phase-pure VPP
that was stable in all gas mixtures applied. However, in the re-
ducing mixture containing 2% n-butane in He, an additional
broad feature at 1580 cmꢀ1 appeared, the intensity of which
increased over time. In this spectral region graphitic carbon
shows a specific Raman signal known as G-band.[39] In addition,
a very weak and broad signal around 1360 cmꢀ1 appeared that
can be assigned to the D-band of defective carbon species.[39]
Thus, the new Raman signals, evolving under reducing condi-
tions within minutes, can be explained by coke formation on
the catalyst. Coking is a very frequently observed phenomenon
in alkane oxidation reactions,[38] in particular, in strongly reduc-
ing environments, and was also reported for VPO catalysts and
even given as a principal source for CO and CO2 by-product
formation.[4] As expected, the coke can be removed by intro-
ducing an oxidizing gas flow containing 20% O2 in He. Under
such oxidizing conditions the bands at 1360 and 1580 cmꢀ1
disappeared (Figure 3), obviously because of the oxidation of
the coke to COx. Therefore, the observed microwave conduc-
tivity behavior of the sample shown in Figure 1a could now be
understood. Graphitic carbon is a rather strong electronic con-
ductor, exceeding the conductivity of VPP by orders of magni-
tude (at 10 GHz, the conductivity of graphite is 0.5–1 Smꢀ1,
whereas that of carbon black is 0.04–0.2 Smꢀ1),[40] perfectly ex-
plaining the increasing conductivity in the strongly reducing
environment.
x
ðVVOÞVꢁO Ð ðVIVOÞVO þ hꢁ
ð3Þ
and support the often discussed[16] role of a narrow V3d band
or local V surface states as the acceptor band/states, giving
rise to the p-type conductivity of VPP. However, the pro-
nounced decrease of the vanadium oxidation state in n-butane
(C4) was not reflected in the conductivity response (see
Figure 1). This discrepancy could be an effect of the different
total pressures and, thus, chemical potentials in the two ex-
periments (n-butane conversion in the NEXAFS experiment
was below 1%) or owing to the formation of electronically iso-
lated reduced vanadium species on the surface (such as V3+).
A third (but closely related) explanation would be the preven-
tion of a strong reduction of the catalyst owing to coke forma-
tion under reducing conditions at ambient pressures and high
conversions, which are discussed below.
The amount of exchangeable oxygen in VPP between room
temperature and 4508C was quantified by temperature-pro-
grammed oxidation (TPO) in 5% O2 in He (TPO1 in Figure 4),
which was succeeded by temperature-programmed reduction
in 5% H2 in Ar (TPR1). This oxidation/reduction cycle was re-
peated (TPO2 and TPR2, respectively). From the integrals of
the measured absolute consumption of O2 and H2, respectively,
the relative amount of exchanged oxygen of VPP was calculat-
ed and plotted versus the temperature (Figure 4). As for TPO,
at 3858C the determined amount of consumed oxygen would
correspond to 0.22% of VPP lattice oxygen. The TPR experi-
ment revealed that the consumption of hydrogen up to 3858C
would correspond to the extraction of 0.15% of VPP lattice
oxygen. For both TPO and TPR, a relative total oxygen ex-
change of 0.35% was obtained at 4508C. These results showed
that 1) the as-received catalyst (run under lean reaction condi-
tions for more than 100 h) was in a reduced state, and that
2) the redox processes were at least within one redox cycle re-
versible. The differences between the TPO and TPR curves
below 4508C might be owing to the different kinetics of the
reduction and oxidation processes. Although both TPR meas-
urements showed within the measurement error no differences
and, hence, indicate the reversibility of the oxygen exchange,
the second TPO curve was shifted to lower temperatures and
resulted in a higher absolute O2 consumption. An explanation
could be the formation of OHꢀ and H2O and/or the irreversible
formation of small amounts of thermodynamically very stable
vanadium(V) phases (not yet observable with Raman or XRPD)
that might also give rise to the slight but constant decrease of
the catalytic activity and conductivity (see Figure 1). In con-
As described above, after an expected sharp initial decline,
the conductivity rose again in the gas mixture containing 1.4%
n-butane in N2 (see Figure 1a). To find an explanation for this
unexpected behavior, in situ Raman spectroscopy was per-
formed. In Figure 3 Raman spectra of VPP under reaction, oxi-
dizing, inert, reducing, and again oxidizing conditions are
shown. During all treatments, signatures at 929, 1131, and
1178 cmꢀ1 were observed. The first band can be assigned to
the pyrophosphate PꢀOꢀP stretching vibration, whereas the
two latter signals can be attributed to VꢀOꢀP stretching bands
of VPP.[35–38] A very weak shoulder around 1000 cmꢀ1 might be
assigned to the V=O bond, which could indicate a very small
concentration of V5+ species,[37] probably located mainly on
the surface. Besides, no signals attributable to alternative vana-
[35–38]
dium oxide or phosphate bulk phases such as VOPO4
ap-
Figure 3. In situ Raman spectra of VPP in different gas atmospheres at
4008C, initial VPP weight=80 mg, gas flow=10 mLminꢀ1. C4/O2: 2% n-
C4H10, 20% O2; O2: 20% O2; C4: 2% n-C4H10; He: 100% He; residual gas: He.
ꢁ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 2318 – 2329 2321

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
capacity of 66 mgmꢀ2, assuming a V4+/V5+ redox couple. With
a BET surface area of 24 m2 gꢀ1, a surface oxygen capacity of
159 mg per 100 mg catalyst was obtained. Accordingly, 0.22%
of exchanged oxygen determined at 3858C would correspond
to 103 mg oxygen per 100 mg catalyst, complying with 65% of
the theoretical total surface oxygen capacity. At 4508C, 164 mg
oxygen per 100 mg catalyst was exchanged, which was close
to the calculated maximum surface capacity. These data sug-
gest that the reversible oxygen exchange mainly involved the
catalyst surface. This result fits nicely to the observed surface
restriction of the vanadium oxidation state change proven by
NEXAFS and XPS, the stability of the bulk phase shown by
XRPD and Raman spectroscopy, and points to the surface sen-
sitivity of the measured microwave conductivity of the sample
depending on the applied gas-phase chemical potentials.
Figure 4. Results of consecutive TPO and TPR experiments of VPP. The rela-
tive amounts of inserted (TPO) and extracted (TPR) (theoretical) VPP lattice
oxygen are shown, respectively, measured during each temperature ramp.
1st TPO run (TPO1): c, 1st TPR run (TPR1): c, 2nd TPO run (TPO2): b,
and 2nd TPR run (TPR2): b.
Vanadium(III), (IV), and (V) phosphorus oxides
Having identified the electronic properties of the benchmark
catalyst VPP, its gas-sensitive electronic behavior was to be
compared with other selective, unselective, or inactive vanadi-
um phosphorus oxides. For this purpose, the microwave con-
ductivity and permittivity of the unselective samples VIIIPO4, a-
VVOPO4 and b-VVOPO4 (crystal structures shown in Figure 5)
were measured (Figure 6a). Notably, aI- or aII-VOPO4 were
spectroscopically hardly discernible from the VOPO4·2H2O
phase. Moreover, the layered a-VOPO4 structure (Figure 5d,e),
which was most likely present at the high reaction tempera-
tures, easily intercalated water between the layers and trans-
formed reversibly into the dihydrate at lower temperatures.
trast, the reversible TPR cycles likely reflected only the reversi-
ble exchange of oxygen on the surface.
One of the strongest X-ray powder diffractometry (XRPD)
signals of VPP could be found at 2q=23.08 (see the Support-
ing Information, Figure S1) and was assigned to {200} plane re-
flections. This might suggest that {100} cleavage planes were
dominating the surface, which are also discussed as active
layers for MA formation.[17,25] The density of V atoms on this
plane was about 8.3 mmolmꢀ2 according to the determined
unit cell parameters (see the Supporting Information, Table S1).
This density would correspond to a theoretical surface oxygen
Figure 5. Crystal structures of a) (VIVO)2P2O7, b) Ag6(VIVO)2(PO4)2(P2O7), c) Ag(VIVO)(VVO)(PO4)2, d) aI-VVOPO4, e) aII-VVOPO4, f) b-VVOPO4.
ꢁ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 2318 – 2329 2322

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
There are only very few catalysts known that exhibit repro-
ducible, stable, and at least moderate selectivities to MA. Re-
cently, Karpov et al. identified a sample with the empirical for-
mula Ag2VIV,VP1.6O7+d (0<d<0.5), which supposedly had
a vanadyl(IV,V) orthophosphate–pyrophosphate layered struc-
ture, as its XRPD pattern (Figure S2) resembled the recently
characterized Ag6(VIVO)2(PO4)2(P2O7) structure (Figure 5b).[41]
The former sample exhibited promising selectivities to MA,
and thus was included as selective catalyst in our study. For
comparison, the phase-pure samples Ag6(VIVO)2(PO4)2(P4O7) and
Ag(VIVO)(VVO)(PO4)2, the latter originally synthesized and char-
acterized by Raveau et al.,[42] were tested as well (Figure 6a).
All samples were investigated by using the same protocol
under isothermal conditions at 3858C and a gas hourly space
velocity (GHSV) of 4240 hꢀ1 (i.e., contact time=0.85 s): after
2 h in the lean reaction gas mixture containing 1.4% n-butane
and 20% O2, the samples were exposed for 2 h to an oxidizing
mixture (20% O2 in N2), followed by inert-gas treatment in N2
for 1 h. Then, the gas mixture was switched back for 1 h to
lean reaction conditions, followed by a strongly reducing at-
mosphere containing 1.4% n-butane in N2 for 1 h. For the VPP
sample, the latter treatment was restricted to 30 min and the
sample was exposed subsequently to 20% O2, then to a final
treatment in 2% H2 in N2. The latter conditions were applied
to verify the p-type conductivity behavior of VPP in reducing
atmospheres excluding the effect of coking. As a result, the
conductivity behavior of VPP was in perfect agreement with
the above description of the behavior in the p-type semicon-
ductor model. The initial increase of the conductivity in the
first reaction feed C4/O2 owed probably to a healing of defects
(oxygen vacancies). The decrease in conductivity upon the
second treatment in O2, followed by a steady increase could
be explained by the removal of coke (diminishing initially the
conductivity), as already observed by Raman spectroscopy
(Figure 3); and the subsequent increase in conductivity was at-
tributed to the oxidation of the catalyst surface, which is as ex-
pected for a p-type semiconductor. In H2, the conductivity
dropped immediately to the level observed under lean reac-
tion conditions, which further evidenced coking to be the re-
sponsible factor for the unexpected conductivity behavior
under n-butane-rich conditions.
The catalytic properties, conductivities, and permittivities of
all samples measured in different gas atmospheres are depict-
ed in Figure 6 and are summarized for the lean reaction condi-
tion (C4/O2) in Table 1.
Notably, for all active samples, CO and CO2 were the only
products observed in significant amounts (besides MA if indi-
cated). Moreover, none of the investigated materials came
even close to the MA yield obtained with VPP. Only the novel
sample Ag2VIV,VP1.6O7+d showed a moderate and stable MA se-
lectivity of 25%, whereas the n-butane conversion was with
2.5% rather low (but stable). At higher temperatures (4308C)
the selectivity decreased to 13% at a conversion of 24%, as
was shown in an earlier study (with a GHSV of 2000 hꢀ1).[41] In-
terestingly, the monovalent silver vanadyl(IV) orthophosphate–
pyrophosphate Ag6(VIVO)2(PO4)2(P2O7) showed no MA selectivity
at all, despite its structural similarities to Ag2VIV,VP1.6O7+d. Its ac-
*
Figure 6. a) Microwave conductivity s (+), real part of the permittivity e1 ( ),
&
*
and b) n-butane conversion ( ) and MA selectivity ( ) of vanadium phospho-
rus oxides in different gas atmospheres at 3858C, GHSV=4240 cmꢀ1. Gas
mixtures: C4/O2: 1.4% n-C4H10, 20% O2; O2: 20% O2; N2: 100% N2; C4:
1.4% n-C4H10; H2: 2.0% H2; residual gas: N2.
ꢁ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 2318 – 2329 2323

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
state of +5 showed n-type semiconductivity, whereas the V3+
phosphate did not react on the gas atmosphere.
Table 1. Catalytic conversions, selectivities, microwave conductivities, and
majority charge carrier types of the different vanadium phosphorus oxide
samples measured at the end of the second cycle under C4/O2 reaction
conditions (see Figure 6).[a]
Another important result is that the conductivity of the
silver vanadyl phosphates was orders of magnitudes higher
than that of the other samples. In general, Ag+ ions in the dif-
ferent crystal layer structures were located between the layers
(Figure 5). Hence charges could be transferred easily between
the Ag ions, forming a delocalized partially filled conduction
band. This band could have caused the high conductivity be-
cause of the high number of charge carriers and/or very high
mobilities. Ionic conductivity of Ag+ might be another explan-
ation for the high conductivity of the silver vanadyl phos-
phates, though the high microwave frequency excitation gen-
erally privileges highly mobile electric charge carriers over iner-
tial ionic ones. However, the characteristic conductivity behav-
ior of the three different silver vanadyl phosphates upon
changing the gas phase was probably still governed by the va-
nadium, as the dynamic behavior resembled the situation
found for the silver-free samples, in particular with respect to
the vanadium oxidation state.
Sample
XC4 SMA SCOx [%] RateMA
s
Cond.
type
[%] [%] CO2 +CO [molmꢀ2 minꢀ1] [Scmꢀ1
]
(VO)2P2O7
Ag6(VO)2(PO4)2(P2O7) 0.5
Ag2VP1.6O7+d
Ag(VO)2(PO4)2
a-VOPO4
31
80
0
10+10 4.6ꢂ10ꢀ7
100+0
4.0ꢂ10ꢀ5
0.17
1.6
1.1
9.0ꢂ10ꢀ7
1.6ꢂ10ꢀ5
5.0ꢂ10ꢀ5
p
p
p
p
n
n
–
0
2.5 25
1.2 12
(0.2)
1.4 13
(0.1)
38+37 2.7ꢂ10ꢀ8
35+53 1.3ꢂ10ꢀ8
(58+42) 0
0
b-VOPO4
VPO4
46+41 4.3ꢂ10ꢀ8
0
(100+0)
0
[a] Reaction conditions: T=3858C, GHSV=4240 hꢀ1 (5 mLminꢀ1), gas mix-
ture: 1.4% n-butane, 20% O2, residual gas N2.
tivity was very low with a conversion of 0.4–0.8%. In contrast,
the mixed vanadyl (IV,V) silver orthophosphate and b-VOPO4
produced MA, though with a low selectivity of 10–15% and an
n-butane conversion of approximately 1%. The low but repro-
ducible conversions allowed the investigation of the catalysts
under nearly differential conditions, hence avoiding steep con-
centration and conductivity gradients in the fixed bed. This
was an advantage in the light of the integral nature of MCPT
measuring the overall conductivity of the sample, which was
to be correlated with the chemical potential of the gas phase
(that was known only upstream and downstream of the cata-
lyst). The other samples, VPO4 and a-VOPO4, could be regarded
as inactive and unselective, as only minor amounts of COx
were produced.
Moreover, the differences in the gas-phase response of the
permittivity e1 were striking. Whereas no significant changes
were observed for the silver-free samples, the permittivity very
closely followed the conductivity trend for the silver vanadyl
phosphates (Figure 6a). The real part of the permittivity de-
pended on the polarizability—i.e., the (induced) electric dipole
character—of the material. Thus, it is conceivable that large
changes of the absolute number or mobility of charge carriers
(indicated by the conductivity), as observed in the silver sam-
ples upon changing the gas atmosphere, would induce appro-
priate changes in the dipolar character (polarizability) of the
sample. However, this interpretation should be taken with
care, because the resonance frequency shift of the microwave
cavity induced by the sample becomes dependent on the con-
ductivity s (or the imaginary part of the permittivity e2) for
large conductivities in the vicinity of the depolarization
peak,[43] and not only on e1 as assumed in our calculation. The
exact relationships should be determined by means of temper-
ature-dependent measurements of the complex permittivity in
the future.
It was now interesting to compare the simultaneously mea-
sured (dynamic) conductivities and permittivities under the re-
spective conditions (Figure 6a). First, all vanadyl(IV) catalysts,
including the mixed vanadyl(IV,V) phosphates, showed an in-
creasing conductivity in O2 and a decreasing conductivity in
the n-butane atmosphere. As described above, this could be
interpreted in terms of a p-type semiconducting behavior. Al-
though Ag6(VIVO)2(PO4)2(P2O7) was obviously unstable in the
presence of n-butane (or more likely its surface), exhibiting
a constantly falling conductivity, it was stabilized in inert or ox-
idizing conditions, thus justifying its classification as a p-type
semiconductor. In contrast, the conductivities of the two vana-
dyl(V) orthophosphates showed the opposite trend, decreasing
values under oxidizing conditions and rising conductivity
under reducing or reaction conditions. Hence, this could be ex-
plained by the fact that electrons were the majority charge car-
riers, which classified these samples as n-type semiconductors.
The p-type and n-type conductivities of VPP and the two
VOPO4 samples, respectively, were in agreement with contact
measurements at low frequencies.[26] The other materials have
not yet been measured by using direct or alternating current
methods. Interestingly, the conductivity of VPO4 remained basi-
cally constant and did not respond systematically to the differ-
ent gas mixtures. Hence it can be summarized that the investi-
gated compounds with a bulk vanadium oxidation state of +4
(or mixed +4/+5) exhibited p-type, samples with an oxidation
For a better comparison of the conductivity kinetic trends of
the different samples, the conductivities were normalized to
(i.e., divided by) the respective value measured under lean re-
action conditions after 190 h time on stream (TOS) for each
sample. All data could then be plotted on one graph
(Figure 7). As conductivity s is the product of the charge carri-
er density N and the mobility m [Eq. (4); e=elementary
charge]:
s ¼ eNm
ð4Þ
and, under the assumption that the charge carrier mobilities
did not depend on the external gas phase, the thus normalized
conductivity would allow a direct comparison of the relative
changes of charge carrier concentrations in dependence on
the gas mixture. Although this hypothesis should be proven in
the future by measuring the mobilities, the general differences
ꢁ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 2318 – 2329 2324

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
strongly reducing conditions. Moreover, all vanadium(IV) (and
(IV,V)) samples were p-type semiconductors, whereas all vana-
dium(V) phases were n-type semiconductors. As already indi-
cated by Equation (3), in vanadium(IV) oxides a small amount
of V5+ was needed for the generation of electron holes. It has
been suggested[28] that these electron holes are trapped by lat-
tice oxygen, as described in Equation (5):
OxO þ hꢁ Ð OꢁO
ð5Þ
Hence, the conductivity could be described as a hopping
process in which the electron hole is transferred between adja-
cent Oꢀ and O2ꢀ ions or within the valence band (exhibiting
a strong O character).[44,45] In vanadium(V) oxides or phos-
phates, local V4+ states or a narrow partially filled V3d band
within the band gap have been suggested,[16] and were indeed
identified in V2O5 single crystals.[45] These states can donate
electrons into the conduction band and thus explain the n-
type semiconductivity of the vanadium(V) samples.
Figure 7. Normalized relative microwave conductivities of vanadium phos-
phorus oxides in different gas atmospheres at 3858C, GHSV=4240 cmꢀ1
.
Gas mixtures: C4/O2: 1.4% n-C4H10, 20% O2; O2: 20% O2; N2: 100% N2; C4:
1.4% n-C4H10; residual gas: N2.
Next, we tried to summarize these results exemplarily for
the two selective catalysts VPP and Ag2VIV,VP1.6O7+d in schemat-
ic band diagrams (Figure 8). Owing to the gas phase sensitivity
of the measurements and the finding that only the surface par-
ticipated in the reaction (as proven for VPP), the samples are
best described as solid-state gas sensors (i.e., chemiresis-
tors).[16,46] Then, the V4+/V5+ redox couple could be identified
as surface-state exchanging electrons with the gas phase (ad-
sorbed molecules). The emptied surface state would corre-
spond to V5+, the occupied surface state to V4+. This charge
transfer would induce a band bending and the formation of
a space charge region attributed to the pinning of the Fermi
level to the surface state level (and/or the redox potential of
the adsorbed reaction gas). As a consequence, the band bend-
ing, which is accompanied by changes in the subsurface
charge carrier concentration to screen the surface charge,
would control the conductivity in the subsurface region, as
was observed in our MCPT measurements. From this represen-
tation it is also clear why vanadium(III) phosphates could not
activate n-butane. Most probably, all (empty) V3+/V4+ surface
states lied above the redox potential of n-butane, thus provid-
ing no driving force for the exchange of charge carriers.
between the samples could be already deduced. First, only the
selective samples VPP and Ag2VIV,VP1.6O7+d showed a similar be-
havior, that is, a strong and fast response to the O2 atmos-
phere and a stable and reproducible conductivity under reac-
tion conditions, irrespective of the previous treatment. Under
oxidative conditions, the values for VPP doubled, whereas the
conductivity of Ag2VIV,VP1.6O7+d increased by a factor of 1.5. The
abrupt change of the conductivity of VPP under inert condi-
tions might be attributable to the removal of weakly chemisor-
bed oxygen. Although the unselective sample Ag(VIVO)(VVO)-
(PO4)2 exhibited a similar increasing conductivity in O2, its
slope was much smaller and, even more important, the con-
ductivity remained on the high value under the subsequently
applied reaction conditions. A similar, though reversed, behav-
ior was observed for the inactive a-VVOPO4. Here, the conduc-
tivity showed a strong and steep decrease in O2, which did not
recover in the reaction gas mixture. In contrast, the slightly
active b-VVOPO4 did recover, but the conductivity difference
between oxidizing and reaction gas atmosphere was much
smaller than for the other active samples. As for Ag6(VIVO)2-
(PO4)2(P2O7) and VIIIPO4, the conductivity was not stable under
reaction conditions or it did not respond at all to the gas
phase, respectively. In conclusion, a strongly reversible re-
sponse in the sequence C4/O2, O2, C4/O2 was obviously
a strong indication for the activity and selectivity of the
catalyst.
Hence, the conductivity behavior of the p-type semiconduc-
tor Ag2VIV,VP1.6O7+d could be easily understood, with a high
conductivity in O2, an intermediate conductivity in C4/O2, and
a low conductivity in C4 (Figure 8a; see Figure 6 and 7). This
behavior was in contrast to that of VPP, for which the conduc-
tivity was never below the value measured in the reaction gas
mixture. Here, VPP already existed in a highly reduced state
(probably with strong downward band bending, Figure 8b).
The absence of oxygen led only to the formation of coke and
not to the formation of oxygenates because of the lack of
charge carriers and/or active oxygen. This behavior could be
explained by a very fast reduction step (oxidation of n-butane),
as already indicated by the conductivity response upon chang-
ing the atmosphere to more reducing conditions (Figure 6 and
7). In a low conducting (highly reduced) p-type semiconductor,
the transport of charge carriers to the surface is limited be-
In addition to the reversibility of the conductivity, the behav-
ior under reducing conditions was unique for VPP. Neither in
n-butane (in which coking disturbed the measurements), nor
in H2 did the conductivity undercut the values observed under
the reaction conditions. In contrast, the conductivity of the p-
type semiconductor Ag2VIV,VP1.6O7+d further decreased, whereas
the n-type semiconductors a-VVOPO4 and b-VVOPO4 exhibited
a
pronounced increase. Interestingly, Ag(VIVO)(VVO)(PO4)2
showed only a very slowly falling conductivity, even under
ꢁ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 2318 – 2329 2325

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
of the (2D) V5+ phase to the surface could also be
understood, as a 3D oxidation or reduction of the
catalyst would enormously increase the band bend-
ing and, hence, the electrostatic repulsion within the
space-charge layer, that is, the system was self-limit-
ing in the segregation of the surface structure.
The electrical conductivity measured by MCPT was
obviously a valuable tool for precise measurements
and characterizations of the electronic response of
the catalytic system on the chemical potential of the
gas phase under close to technical conditions. The
technique was extremely powerful in conjunction
with in situ XPS and NEXAFS as shown for the exam-
ple of VPP. Whereas the latter synchrotron-based
methods gave insights into the concrete oxidation
states, ratios, and respective changes of the redox
active elements involved, the conductivity response
indicated the electronic coupling between adsor-
bates, the active surface, and the bulk structure at
high pressures and (realistic) conversions in a labora-
tory machine without the need for expensive syn-
chrotron beam time. Thus, XPS and NEXAFS are well
suited for fundamental studies of a limited number
of well characterized and well-chosen systems. For
extended and possibly even high-throughput experi-
ments in conjunction with a comprehensive catalytic
investigation under various conditions, in situ MCPT
is the method of choice. In the future, we will work
on the quantification of the involved charge carriers
by MCPT and related methods to improve the under-
standing of charge transfer between gas, surface,
and bulk phase further.
Conclusions
In this contribution, the microwave conductivities of
the vanadium phosphorus oxides VIIIPO4, (VIVO)2P2O7
(VPP), Ag6(VIVO)2(PO4)2(P2O7), Ag2VIV,VP1.6O7+d, Ag(VIVO)-
(VVO)(PO4)2, b-VVOPO4 and a-VVOPO4 were investigat-
ed by the microwave cavity perturbation technique
(MCPT) under different reaction conditions and were
compared with the simultaneously measured catalyt-
ic performance of these samples in the selective oxi-
dation of n-butane to maleic anhydride. It was found
that all vanadium(IV) and (IV,V) samples showed p-
type, vanadium(V) phosphates n-type semiconductiv-
ity, whereas the vanadium(III) phase exhibited no re-
sponse to the applied reaction gas mixtures.
Figure 8. Schematic band diagrams of a p-type semiconductor with upward band bend-
ing upon adsorption of the acceptor molecule O2 and downward bending upon adsorp-
tion of the donor n-butane to a V4+/V5+ surface state. Conductivity responses of
a) Ag2VP1.6O7+d and b) VPP. EF =Fermi energy, Ei =intrinsic energy, ECB =conduction band
edge, EVB =valence band edge, NSS =surface-state density, z=distance. The arrows indi-
cate the transfer of electron holes as a result of the adsorption of molecules with suitable
frontier orbitals and redox energies ERedox at surface states. Additionally, the correspond-
ing local and mean conductivities (as measured by integral methods such as MCPT) of
the semiconductors are shown schematically.
As for VPP, it could be shown by X-ray powder dif-
fractometry and Raman spectroscopy that the bulk
cause of the increasing electrostatic repulsion with the increas-
ing depletion of electron holes (strong downward band bend-
ing). This space-charge-limited charge transfer might prevent
the over-oxidation of the substrate, which is an important pre-
requisite for highly selective oxidation catalysts and was only
observed in the highly selective VPP but in none of the other
investigated (unselective) samples. Furthermore, the restriction
was stable during the redox cycles applied. Moreover, temper-
ature-programmed reduction/oxidation (TPR/O) and in situ
NEXAFS experiments indicated that the redox activity of vana-
dium and the oxygen exchange was restricted to the surface
region of the catalyst. This proves the high sensitivity of MCPT
as an indicator for the redox stability of catalysts under opera-
tion conditions. It is conceivable that MCPT can be developed
ꢁ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 2318 – 2329 2326

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
and (NH4)2HPO4 in the molar ratio 1:0.9:2 for 24 h at 3808C in air.
After complete evaporation of all volatile species, elemental vana-
dium was added as the reducing agent to set the desired composi-
tion. After reaction at 5508C (72 h) in a sealed silica tube, silver va-
nadyl(IV) vanadyl(V) orthophosphate was obtained as a brown, mi-
crocrystalline single-phase powder (XRPD pattern, see Figure S3)
as a rapid test method for evaluating high-throughput elec-
tronic transport and catalytic data without performing a com-
plete structural analysis for predicting promising (even multi-
phase) catalysts.
Moreover, we could identify three general descriptors of se-
lective and active oxidation catalysts: 1) a fast and reversible
change of the conductivity (and surface vanadium oxidation
state) with the chemical potential of the gas phase; b) stability
of the electronic structure (conductivity and vanadium oxida-
tion state) under the reaction conditions, and c) the prerequi-
site that the absolute conductivity under reaction conditions is
as low as possible (compared to strongly reducing conditions
for p-type or oxidizing conditions for n-type semiconductors).
The latter descriptor is interpreted as physical representation
of the “site-isolation” principle of selective oxidation catalysts
because a low amount of active charge carriers should prevent
an over-oxidation of the substrate.
with a BET surface area of 2.7 m2 gꢀ1
.
a-VVOPO4/VVOPO4·2H2O. The dihydrate (Figure S5) was synthesized
according to the literature[50] by heating V2O5 at reflux in concen-
trated H3PO4, filtering the precipitate, washing with water and ace-
tone and drying in air. The anhydrous phosphate was obtained by
dehydrating the dihydrate under vacuum at 7008C. The anhydrous
phosphate was rehydrated reversibly and rather fast. The sample
exhibited a BET surface area of 6 m2 gꢀ1
.
By heating a-VVOPO4 and traces of moisture in a sealed silica am-
poule for 7 d at 5008C, the a-form was transformed to b-VOPO4.
By applying a temperature gradient of 700!6008C, chemical
vapor transport of b-VOPO4 becomes possible, with water acting
as the transport agent.[51] According to its XRPD pattern (Figure S6),
the b-modification was obtained as single-phase product with
The absolute conductivity and the majority charge carrier
type are probably no (simple) descriptors of selective catalysts,
because alternative compounds such as the MoVTeNbOx M1
phase are active and selective in the oxidation of n-butane,
but they indicate, in contrast to VPP, a high and n-type semi-
conductivity, as identified in preliminary in situ MCPT studies.
Thus, besides the studied redox behavior and participation of
(surface) lattice oxygen, more key properties, as summarized in
the form of seven (heuristic) fundamental principles,[24] addi-
tionally comprising metal-oxygen bond strength, host struc-
ture, multifunctionality of active sites, (spatial) site isolation
and phase cooperation (i.e., chemical complexity), are needed
to predict and rationally design lead structures for selective ox-
idation catalysts.
a BET surface area of 1 m2 gꢀ1
.
Vanadium(III) orthophosphate was obtained in sealed silica tubes
by reducing b-VOPO4 with vanadium phosphide VP. VIIIPO4 was pu-
rified and crystallized by chemical vapor transport. Details on the
synthesis and crystallization are described in the literature.[51,52] The
XRPD pattern of the single-phase material is given in Figure S7.
The BET surface area was <1 m2 gꢀ1
.
All powder samples were first pelletized, then the pellets were
crushed and for the catalytic and MCPT experiments, the sieve frac-
tion of 80–200 mm was used.
Characterization
The MCPT setup and the measurement protocol were described in
detail previously.[22,23] In short, as resonator a cylindrical X-band
TM110 silver-plated brass cavity (ZWG Berlin–Adlershof) with
a height of 19.5 mm and a diameter of 38.5 mm was used. A
quartz tube plug-flow reactor with 4 mm outer and 3 mm inner di-
ameter containing the sample under investigation (powders were
filled in with a bed height of 10 mm and embedded within quartz
wool) and surrounded by a 10 mm outer diameter double-walled
quartz Dewar mantle was directly placed in the center of the
cavity. The quartz tube reactor was connected upstream to a gas
delivery manifold equipped with mass flow controllers (Bronkhorst
El-Flow) and downstream to an online gas chromatograph (Agilent
7890A). Heating of the reactor was performed by preheating
a stream of N2 (8 Lminꢀ1) in a resistive furnace consisting of a Sylva-
nia tungsten series I heater. The cavity was cooled with two water-
circuit-cooled copper plates attached to the resonator endplates
and additionally heated by heating wires to maintain a constant
resonator temperature of 258C. The cavity was connected to
a vector network analyzer (Agilent PNA-L N5230C-225) to record
resonance spectra of S11-parameters in reflection mode (reflected
power versus frequency) and to determine the resonance frequen-
cy and quality factor of the cavity with and without sample. The
microwave power attenuation was set to 11 dBm (12.6 mW).
Experimental Section
Catalyst preparation
The synthesis and characterization of the benchmark catalyst va-
nadyl pyrophosphate was described previously.[23] Its X-ray powder
diffraction pattern is shown in Figure S1 of the Supporting Infor-
mation. This catalyst exhibited a BET surface area of 24 m2gꢀ1
.
Ag2VIV,VP1.6O7+d (0ꢂdꢂ0.5) was synthesized by the reaction of
silver acetate with V2O5 and phosphoric acid in water, using citric
acid as the reducing agent under reflux and constant stirring. The
resulting suspension was filtered and the recovered solid washed
with water and dried under vacuum at 908C. The XRPD pattern of
Ag2VIV,VP1.6O7+d (Figure S2), in particular, of the post-mortem
sample measured after the whole measurement cycle, showed sim-
ilarities to the XRPD pattern of Ag0.43VIV,VO(PO4)·2H2O,[47] Ag2(VVO2)-
(PO4)[41,48] and Ag6(VIVO)2(PO4)2(P2O7)[49] (Figure S4). The sample ex-
hibited a BET surface area of 7 m2 gꢀ1
.
In a very similar procedure,[49] Ag6(VIVO)2(PO4)2(P2O7) was obtained
from AgNO3, V2O5, and (NH4)2HPO4 with vanadium as the reducing
agent. The same temperatures as reported for Ag(VIVO)(VVO)(PO4)2
were used. Thus, the mixed pyrophosphate–orthophosphate was
obtained as single-phase product (Figure S4) with a BET surface
In situ NEXAFS was performed at the synchrotron radiation facility
BESSY II of the Helmholtz-Zentrum Berlin by using monochromatic
radiation of the Innovative Station for In Situ Spectroscopy (ISISS)
beamline as a tunable X-ray source. High-pressure NEXAFS spectra
were obtained in the presence of reactive gases at elevated tem-
area of 0.3 m2 gꢀ1
.
Ag(VIVO)(VVO)(PO4)2 was synthesized according to a slightly modi-
fied procedure described in literature[42] by reacting AgNO3, V2O5,
ꢁ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 2318 – 2329 2327

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
[7] G. J. Hutchings, A. Desmartin-Chomel, R. Olier, J. C. Volta, Nature 1994,
368, 41–45.
[8] U. Rodemerck, B. Kubias, H. W. Zanthoff, M. Baerns, Appl. Catal. A 1997,
153, 203–216.
[9] B. Chen, E. J. Munson, J. Am. Chem. Soc. 2002, 124, 1638–1652.
[10] M. Hꢀvecker, R. W. Mayer, A. Knop-Gericke, H. Bluhm, E. Kleimenov, A.
Liskowski, D. Su, R. Follath, F. G. Requejo, D. F. Ogletree, M. Salmeron,
J. A. Lopez-Sanchez, J. K. Bartley, G. J. Hutchings, R. Schlçgl, J. Phys.
Chem. B 2003, 107, 4587–4596.
perature using the high-pressure end station designed and con-
structed at the Fritz-Haber-Institut. Details of the setup are de-
scribed elsewhere.[53] In brief, VPP powder (5 mg) was pressed into
a self-supporting disc (pelletizing pressure of 5 t, pellet diameter
8 mm) that was mounted inside a reaction cell onto a sapphire
sample holder approximately 1200 mm in front of the first aperture
of a differentially pumped electrostatic lens system. The home-
built electron lens serves as the input system for a (modified) com-
mercial hemispherical electron analyzer (PHOIBOS 150, Specs-
GmbH). Gases were introduced to the cell through calibrated mass
flow controllers, heating was provided by a near-IR continuous-
wave laser (808 nm) at the rear of the sample, and the temperature
was monitored by a thermocouple attached directly to the sample
surface.
[11] H. Bluhm, M. Hꢀvecker, E. Kleimenov, A. Knop-Gericke, A. Liskowski, R.
Schlçgl, D. S. Su, Top. Catal. 2003, 23, 99–107.
[12] G. J. Hutchings, J. Mater. Chem. 2004, 14, 3385–3395.
[13] N. Ballarini, F. Cavani, C. Cortelli, S. Ligi, F. Pierelli, F. Trifiro, C. Fumagalli,
G. Mazzoni, T. Monti, Top. Catal. 2006, 38, 147–156.
[14] G. W. Coulston, S. R. Bare, H. Kung, K. Birkeland, G. K. Bethke, R. Harlow,
N. Herron, P. L. Lee, Science 1997, 275, 191–193.
[15] P. Mars, D. W. van Krevelen, Chem. Eng. Sci. Spec. Suppl. 1954, 3, 41–57.
[16] S. R. Morrison, The chemical physics of surfaces, Plenum Press, New York,
1977.
[17] D. X. Wang, H. H. Kung, M. A. Barteau, Appl. Catal. A 2000, 201, 203–
213.
[18] M. J. Lorences, G. S. Patience, F. V. Diez, J. Coca, Appl. Catal. A 2004, 263,
193–202.
[19] M. Abon, K. E. Bere, P. Delichere, Catal. Today 1997, 33, 15–23.
[20] R. Schlçgl, A. Knop-Gericke, M. Hꢀvecker, U. Wild, D. Frickel, T. Ressler,
R. E. Jentoft, J. Wienold, G. Mestl, A. Blume, O. Timpe, I. Uchida, Top.
Catal. 2001, 15, 219–228.
For the in situ Raman experiments, the catalyst sample was placed
in a temperature-controlled Raman cell (Linkam CCR1000), which
was mounted on a microscope stage. A 1 mW beam of an Ar ion
laser (Spectra-Physics Stabilite 2018) operating at 488 nm was fo-
cused on the sample and the scattered light was analyzed in a gra-
ting spectrometer coupled to a charge-coupled device camera
(Princeton Instruments, 500 groovesmmꢀ1, 750 mm).
Temperature-programmed oxidation and reduction experiments
were performed in a tube furnace (Carbolite). The catalyst was
placed above a frit in a fixed-bed reactor. The temperature was
measured with a thermocouple (K-type) in the sample and the O2/
H2 gas concentration was analyzed by TCD detectors (Rosemount
[21] R. Moos, M. Wedemann, M. Spçrl, S. Reiss, G. Fischerauer, Top. Catal.
2009, 52, 2035–2040.
[22] M. Eichelbaum, R. Stçßer, A. Karpov, C.-K. Dobner, F. Rosowski, A.
Trunschke, R. Schlçgl, Phys. Chem. Chem. Phys. 2012, 14, 1302–1312.
[23] M. Eichelbaum, M. Hꢀvecker, C. Heine, A. Karpov, C. K. Dobner, F. Rosow-
ski, A. Trunschke, R. Schlçgl, Angew. Chem. 2012, 124, 6350–6354;
Angew. Chem. Int. Ed. 2012, 51, 6246–6250.
Analytical). The applied temperature ramp was always 6 Kminꢀ1
.
Before the measurements the sample was preheated in He at
4008C to remove residual water and adsorbates.
N2 adsorption studies were conducted at 77 K on a Quantachrome
Autosorb-6B analyzer to measure the specific surface area. Prior to
the measurement, the sample was outgassed in vacuum at 2508C
for 3 h. The specific surface area SBET was calculated according to
the 11-point BET method over a p/p0 =0.05–0.3 pressure range.
[24] R. K. Grasselli, Top. Catal. 2002, 21, 79–88.
[25] M. Abon, J. M. Herrmann, J. C. Volta, Catal. Today 2001, 71, 121–128.
[26] F. Rouvet, J. M. Herrmann, J. C. Volta, J. Chem. Soc. Faraday Trans. 1994,
90, 1441–1448.
[27] L. Sartoni, A. Delimitis, J. K. Bartley, A. Burrows, H. Roussel, J. M. Herr-
mann, J. C. Volta, C. J. Kiely, G. J. Hutchings, J. Mater. Chem. 2006, 16,
4348–4360.
[28] J. M. Herrmann, P. Vernoux, K. E. Bere, M. Abon, J. Catal. 1997, 167,
106–117.
Acknowledgements
[29] K. Ait-Lachgar, A. Tuel, M. Brun, J. M. Herrmann, J. M. Krafft, J. R. Martin,
J. C. Volta, M. Abon, J. Catal. 1998, 177, 224–230.
[30] J. Sinkkonen, Phys. Status Solidi B 1981, 103, 231–237.
[31] J. Haemers, E. Baetens, J. Vennik, Phys. Status Solidi A 1973, 20, 381–
386.
[32] P. J. Gellings, H. J. M. Bouwmeester, Catal. Today 2000, 58, 1–53.
[33] J. M. Herrmann, Catal. Today 2006, 112, 73–77.
[34] M. Hꢀvecker, A. Knop-Gericke, R. W. Mayer, M. Fait, H. Bluhm, R. Schlçgl,
J. Electron Spectrosc. Relat. Phenom. 2002, 125, 79–87.
[35] F. Ben Abdelouahab, R. Olier, N. Guilhaume, F. Lefebvre, J. C. Volta, J.
Catal. 1992, 134, 151–167.
[36] Z. Y. Xue, G. L. Schrader, J. Phys. Chem. B 1999, 103, 9459–9467.
[37] M. Conte, G. Budroni, J. K. Bartley, S. H. Taylor, A. F. Carley, A. Schmidt,
D. M. Murphy, F. Girgsdies, T. Ressler, R. Schlçgl, G. J. Hutchings, Science
2006, 313, 1270–1273.
The present work has been supported by the German Federal
Ministry of Education and Research (BMBF) as part of the ReAlSe-
lOx project, grant number 033R028. The authors thank Dr. Frank
Girgsdies for XRPD, Florian-Fabien Rybicki for BET measurements
and Dr. Raimund Horn (all FHI) for providing the Raman setup.
The HZB staff is acknowledged for their continual support of the
high-pressure electron spectroscopy activities of the FHI at BESSY
II. We are very grateful to Prof. R. Stçßer (Humboldt-Universitꢀt
zu Berlin) for his steady support regarding the development of
the microwave cavity perturbation technique.
[38] S. M. Bennici, B. M. Vogelaar, T. A. Nijhuis, B. M. Weckhuysen, Angew.
Chem. 2007, 119, 5508–5512; Angew. Chem. Int. Ed. 2007, 46, 5412–
5416.
Keywords: electron transport
microwave chemistry · oxidation · vanadium
· heterogeneous catalysis ·
[39] A. C. Ferrari, J. Robertson, Phys. Rev. B 2000, 61, 14095–14107.
[40] M. Hotta, M. Hayashi, M. T. Lanagan, D. K. Agrawal, K. Nagata, Isij Int.
2011, 51, 1766–1772.
[1] C. H. Bartholomew, R. J. Farrauto, Fundamentals of industrial catalytic
processes, Wiley-Interscience, Hoboken, 2006.
[2] F. Cavani, Catal. Today 2010, 157, 8–15.
[41] A. Karpov, C. K. Dobner, R. Glaum, S. A. Schunk, F. Rosowski, Chem. Ing.
Tech. 2011, 83, 1697–1704.
[42] A. Grandin, J. Chardon, M. M. Borel, A. Leclaire, B. Raveau, J. Solid State
Chem. 1993, 104, 226–231.
[43] W. Bauhofer, J. Phys. E 1981, 14, 934–938.
[44] M. G. Willinger, D. S. Su, R. Schlçgl, Phys. Rev. B 2005, 71, 155118–
155125.
[3] F. Cavani, J. H. Teles, ChemSusChem 2009, 2, 508–534.
[4] M. A. Pepera, J. L. Callahan, M. J. Desmond, E. C. Milberger, P. R. Blum,
N. J. Bremer, J. Am. Chem. Soc. 1985, 107, 4883–4892.
[5] G. Centi, F. Trifiro, J. R. Ebner, V. M. Franchetti, Chem. Rev. 1988, 88, 55–
80.
[6] G. Centi, Catal. Today 1993, 16, 5–26.
ꢁ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 2318 – 2329 2328

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
[45] S. Laubach, P. C. Schmidt, A. Thissen, F. J. Fernandez-Madrigal, Q. H. Wu,
W. Jaegermann, M. Klemm, S. Horn, Phys. Chem. Chem. Phys. 2007, 9,
2564–2576.
[46] J. Haber, M. Witko, J. Catal. 2003, 216, 416–424.
[47] P. Ayyappan, A. Ramanan, C. C. Torardi, Inorg. Chem. 1998, 37, 3628–
3634.
[51] M. Binnewies, R. Glaum, M. Schmidt, P. Schmidt, Chemical Transport Re-
actions, Walter de Gruyter, Berlin, 2012.
[52] R. Glaum, R. Gruehn, Z. Kristallogr. 1992, 198, 41–47.
[53] A. Knop-Gericke, E. Kleimenov, M. Hꢀvecker, R. Blume, D. Teschner, S. Za-
feiratos, R. Schlçgl, V. I. Bukhtiyarov, V. V. Kaichev, I. P. Prosvirin, A. I. Ni-
zovskii, H. Bluhm, A. Barinov, P. Dudin, M. Kiskinova, Adv. Catal. 2009,
52, 213–272.
[48] H. Y. Kang, S. L. Wang, P. P. Tsai, K. H. Lii, J. Chem. Soc. Dalton Trans.
1993, 1525–1528.
[49] R. Glaum, C. Welker-Nieuwoudt, C.-K. Dobner, M. Eichelbaum, F. Gru-
chow, C. Heine, A. Karpov, R. Kniep, F. Rosowski, R. Schlçgl, S. A. Schunk,
S. Titlbach, A. Trunschke, Chem. Ing. Tech. 2012, 84, 1766–1779.
[50] H. R. Tietze, Aust. J. Chem. 1981, 34, 2035–2038.
Received: December 21, 2012
Published online on May 14, 2013
ꢁ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 2318 – 2329 2329