Gas phase oxidation of furfural to maleic anhydride on V2O5/Γ-Al2O3 catalysts: Reaction conditions to slow down the deactivation

Article abstract of DOI:10.1016/j.jcat.2016.12.005

An alumina-supported vanadium oxide catalyst (13.9?wt.% vanadium oxide) has been characterized by different techniques and tested in the gas phase oxidation of furfural. These studies have shown that the catalyst unavoidably deactivates by deposition of maleates and resins over the surface. Full regeneration is accomplished by burning off these deposits at 773?K. The studies have also demonstrated that if the primary contact occurs at temperatures at which furfural conversion is low and then the temperature is increased in a low- to high-temperature mode, intense deposition of maleates and resins takes place and the catalyst is rapidly deactivated. The increase of the temperature does not result in removal of deposits but accelerates the deposition. Under this protocol, the yield of maleic anhydride never exceeded 30%, irrespective of the reaction conditions (temperature and O₂/furfural mole ratio). In contrast, if the catalyst first contacts the reaction mixture at high oxidizing potential, then the rate of maleate and resin deposition is much slower, and so is the deactivation rate, and the catalyst can display a higher yield of maleic anhydride for a longer period of time. A high oxidizing potential can be attained at a high reaction temperature (close to full conversion). A higher oxidizing potential at a given high temperature can be accomplished by increasing the O₂/furfural mole ratio. Thus, for example, first contacting the catalyst at 593?K (full conversion), 1?vol.% of furfural, and O₂/furfural mole ratio?=?10, obtained an initial maleic anhydride yield of 68%, and the yield was still greater than 50% after 15?h on stream. On contacting at 573?K with 1?vol.% furfural and 20?vol.% O₂, the maleic anhydride yield was initially close to 75% and was above 60% after 15?h.

Full text of DOI:10.1016/j.jcat.2016.12.005

Journal of Catalysis 348 (2017) 265–275
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Gas phase oxidation of furfural to maleic anhydride on V O /c-Al O3
2
5
2
catalysts: Reaction conditions to slow down the deactivation
⇑
N. Alonso-Fagúndez, M. Ojeda, R. Mariscal, J.L.G. Fierro, M. López Granados
Institute of Catalysis and Petrochemistry (CSIC), C/Marie Curie, 2, Campus de Cantoblanco, 28049 Madrid, Spain
a r t i c l e i n f o
a b s t r a c t
Article history:
An alumina-supported vanadium oxide catalyst (13.9 wt.% vanadium oxide) has been characterized by
different techniques and tested in the gas phase oxidation of furfural. These studies have shown that
the catalyst unavoidably deactivates by deposition of maleates and resins over the surface. Full regener-
ation is accomplished by burning off these deposits at 773 K. The studies have also demonstrated that if
the primary contact occurs at temperatures at which furfural conversion is low and then the temperature
is increased in a low- to high-temperature mode, intense deposition of maleates and resins takes place
and the catalyst is rapidly deactivated. The increase of the temperature does not result in removal of
deposits but accelerates the deposition. Under this protocol, the yield of maleic anhydride never
Received 6 October 2016
Revised 7 December 2016
Accepted 8 December 2016
Keywords:
Furfural
Gas phase oxidation
Maleic anhydride
Vanadium oxide
2
exceeded 30%, irrespective of the reaction conditions (temperature and O /furfural mole ratio). In con-
trast, if the catalyst first contacts the reaction mixture at high oxidizing potential, then the rate of maleate
and resin deposition is much slower, and so is the deactivation rate, and the catalyst can display a higher
yield of maleic anhydride for a longer period of time. A high oxidizing potential can be attained at a high
reaction temperature (close to full conversion). A higher oxidizing potential at a given high temperature
can be accomplished by increasing the O
alyst at 593 K (full conversion), 1 vol.% of furfural, and O
maleic anhydride yield of 68%, and the yield was still greater than 50% after 15 h on stream. On contacting
2
/furfural mole ratio. Thus, for example, first contacting the cat-
2
/furfural mole ratio = 10, obtained an initial
at 573 K with 1 vol.% furfural and 20 vol.% O
was above 60% after 15 h.
2
, the maleic anhydride yield was initially close to 75% and
Ó 2016 Elsevier Inc. All rights reserved.
1
. Introduction
Recently, other renewable chemical platforms have also been
reported to yield MA via oxidation, such as gas phase oxidation
of biobutanol, 5-hydromethylfurfural (HMF), and levulinic acid.
Thus far, the yield of MA from biobutanol is lower than that
obtained from furfural [10]. From HMF an overall yield of MA
and maleic acid of 80% has been reported for liquid phase oxidation
and at relatively high O pressure (1 MPa) [11]. A high yield of MA
2
(ca. 70%) has also been demonstrated for gas phase oxidation of
levulinic acid conducted at atmospheric pressure [12].
However, biobutanol, HMF, and levulinic acid are not yet com-
mercially available on the scale required for replacement of MA
production from oil. In contrast, FUR is one of the few commodities
currently produced on a large scale from biomass and is perfectly
coupled with the production of human food and animal feed,
because it is produced from agricultural residues [13]. FUR has
been identified as one of the top value-added chemicals derivable
from biomass [14] because a number of chemicals are already pro-
duced from furfural, and a wide variety of other biofuels, com-
modities, and fine chemicals can also be derived from it [15,16].
This makes the gas phase oxidation of furfural highly attractive.
Maleic anhydride (MA) is a commodity chemical with multiple
applications: production of unsaturated polyester resins, 1,4-
butanediol, vinyl copolymers, curing agents for epoxy resins, lubri-
cant additives, pharmaceuticals, and agrochemicals [1,2]. MA is a
petrochemical with market volume currently above 1800 kton/
year and is produced primarily by selective oxidation of n-butane
(
[
the benzene route became obsolete for environmental reasons)
1,2].
Renewable routes based on the oxidation of furfural (FUR) in
either gas [3–5] or liquid phase [6–8] have been demonstrated
technically by different research groups. Gas phase oxidation offers
the advantage of using low-cost air at atmospheric pressure
instead of either high-pressure O
2 2 2
[6,7] or more expensive H O
[
8,9] required for liquid phase oxidation of furfural.
⇑
Corresponding author.
E-mail address: mlgranados@icp.csic.es (M. López Granados).
http://dx.doi.org/10.1016/j.jcat.2016.12.005
021-9517/Ó 2016 Elsevier Inc. All rights reserved.
0

2
66
N. Alonso-Fagúndez et al. / Journal of Catalysis 348 (2017) 265–275
Gas phase oxidation of furfural is conducted using catalytic
fixed-bed reactors at temperatures ranging from 473 to 673 K
and using dilute furfural fed in O –N mixtures (below 5 vol.%)
value, and the furfural/O
alytic bed (approximately 43 mL/min was always fed, which is
equivalent to an overall contact time of 2.1 s or GHSV = 1720 h ).
The concentration of furfural in the gas phase was controlled using
a double saturator submerged in a thermostated silicone oil bath at
2 2
/N mixture was passed through the cat-
ꢀ
1
2
2
and notably short contact times (a few seconds). In addition, differ-
ent vanadium oxide-based catalysts (V oxide and V–Mo and V–Bi
mixed oxides) on a variety of supports (pumice, asbestos, and alu-
mina) have been tested [3–5,17–20]. Unfortunately, most of the
reports lack experimental and technical details, and a few of them
date back to the first half of the 20th century. Fundamental insight
into the reaction mechanism, the effect of different operational
variables, and the deactivation processes is still needed.
2
the required temperature. N was used as a carrier gas by bubbling
it through the furfural present in the saturators. Heavy oligomers
(brownish-colored) are unavoidably present dissolved in furfural
due to oligomerization reactions [15]. The feeding of these much
less volatile heavy products may cause plugging and deactivation
issues, which are prevented by using this bubbling feeding system.
This work shows that the catalytic properties of V
2
O
5
/
c
-Al
2
O
3
The O
rated into the feed line downstream of the saturator and prior to
the reactor. The N and O flows were controlled by mass flow con-
trollers. Furfural was purchased from Aldrich (99%). Concentra-
tions of organics in the inlet and effluent streams were measured
online via gas chromatography (Agilent 6820). A capillary column
2
required to set the gas phase concentration was incorpo-
depend on the thermal protocol to which the catalyst has been
subjected. Moreover, it has been shown that the effect can be used
efficiently to obtain a greater yield of MA. A frequent practice in
conducting a reaction is to contact the catalyst with the reaction
mixture at an initial temperature at which conversion is low, and
eventually increase the temperature to achieve greater conversion
and higher yields of the targeted products. The results reported in
this work demonstrate that the catalytic properties depend on the
thermal history. Substantially higher yields of MA can be attained
at full conversion if the catalyst initially contacts the gas reaction
mixture at the temperature required for full furfural conversion.
In contrast, if the more frequent protocol is adopted, namely, the
catalyst first contacts the gas reaction mixture at temperatures giv-
ing rise to low conversion (less than 40%) and then the tempera-
ture is increased to achieve full conversion, a much lower yield
of MA is obtained. A characterization study of the fresh and used
catalyst by thermogravimetric analysis (TGA), diffuse reflectance
infrared Fourier transform (DRIFT), and X-ray photoelectron spec-
tra (XPS) techniques was conducted, aiming at finding an explana-
tion for this effect.
2
2
(HP-5, 30 m ꢁ 0.32 mm ꢁ 0.25
tion detector was used to analyze furfural and maleic anhydride,
and HP-PLOT Q (30 m ꢁ 0.32 mm ꢁ 0.20 m) and molecular sieve
(30 m ꢁ 0.32 mm ꢁ 0.25 m) capillary columns connected to a
thermal conductivity detector were used to determine the O , N
(internal standard), CO, and CO concentrations. All transfer lines
lm) connected to a flame ioniza-
l
l
2
2
2
were held at 393 K to prevent condensation of water and MA. Fur-
fural conversion is calculated as
ðFÞin ꢀ ðFÞout
F
X ¼
ð1Þ
ðFÞ
in
where (F)in and (F)out are the inlet and outlet furfural molar flow
rates, respectively, and MA yield as
MAout
Y
MA
¼
ð2Þ
ðFÞin
where MAout is the outlet molar flow rate of MA.
2
. Experimental methods
The CO yield formed by total oxidation of furfural was calcu-
lated on a furfural molar basis as
2
2.1. Catalyst preparation
ðCO
2
Þ
ꢀ MAout
out
In vanadium oxide catalysts supported on Al
2
O
3
, surface V
2
O
5
YCO2
¼
ð3Þ
5
ꢁ ðFÞ
in
crystallites have been identified as intrinsically less active for this
reaction than well-dispersed VO species [5]. The best alumina-
x
where (CO
accounts for the CO
cule implies release of a CO
)out is the outlet molar flow rate of CO
and MAout
2
2
supported vanadium oxide catalyst is that with the largest amount
of well-dispersed vanadium oxide species. The maximum surface
concentration of well-dispersed vanadium oxide species is attained
2
released to form MA (formation of a MA mole-
2
molecule by decarboxylation of furoic
acid)
2
in the range of 8–10 atoms of V per nm of alumina support.
Al
V atoms nm
3.9 wt.%) was prepared via incipient wetness impregnation of c-
2 3
O -supported vanadium oxide with a surface V loading of
2
.3. Catalyst characterization techniques
ꢀ2
8
1
2 5
(equivalent to a V O loading equivalent of
TGA under a controlled atmosphere (synthetic air) was per-
2
ꢀ1
2
Al O
3
(134 m g ) at 298 K using an aqueous solution of ammo-
VO , Alfa Aesar, 99.0%). Oxalic acid
, Alfa Aesar, 99.5%) was added (NH VO /oxalic acid molar
ratio = 0.5) to ensure the complete dissolution of NH VO in H O.
The impregnated solids were dried in ambient air at 393 K for
2 h and subsequently treated in ambient air at 773 K (heating rate
formed on a Mettler Toledo TGA/SDTA 851e unit at a rate of
0
nium metavanadate (NH
4
3
ꢀ1
.167 K s in the range 300–1000 K.
(
C H
2 2
O
4
4
3
XPS were acquired with a VG Escalab 200 R spectrometer
4
3
2
equipped with a hemispherical electron analyzer and a MgK
a
(
1253.6 eV) X-ray source. The solids were outgassed at 393 K for
1
ꢀ5
1
h at 10 mbar to remove water before transfer to the ion-
ꢀ1
of 0.167 K s ). This V loading is representative of the range at
which the surface concentration of well-dispersed V oxide species
is maximized [5]. The catalyst is referred to as 8VA.
pumped analysis chamber. The O1s, V2p3/2, C1s, and Al2p core
levels were scanned a sufficient number of times to obtain high
signal-to-noise ratios. The static charge of the samples was cor-
rected by referencing the binding energies (BEs) to that of the
2.2. Measurement of catalytic activity
2 3
Al2p peak in Al O (74.4 eV). The accuracy of the BE was ±0.1 eV.
The areas of the peaks were computed by fitting the experimental
spectra to Gaussian/Lorentzian curves after background removal
(using the Shirley function). Surface atom ratios were calculated
from peak area ratios normalized by the corresponding atomic sen-
sitivity factors [21].
Furfural oxidation was performed in a tubular fixed-bed stain-
less steel reactor. The catalytic bed (1.5 mL) was prepared by mix-
ing the catalyst (0.2 g, pellet size 300–425 m) with low surface
l
2
ꢀ1
area SiO
2
(2 g, <1 m g , pellet size 300–425
l
m). First, the cata-
ꢀ1
lysts were pretreated in a 21 vol.% O
2
/He flow (250 mL min
g
c
at
DRIFT spectra were obtained with a Nicolet 5700 FT spec-
trophotometer equipped with an in situ reaction chamber, a dif-
ꢀ1
)
at 623 K for 1 h. Then the temperature was set to the desired

N. Alonso-Fagúndez et al. / Journal of Catalysis 348 (2017) 265–275
267
fuse mirror system (Praying Mantis), and a high-sensitivity Hg–
Cd–Te cryodetector. Spectra were obtained at a resolution of
version. A hypothesis as to the chemical nature of these heavy prod-
ucts will be given later, based on the DRIFT results discussed below.
When the temperature was increased, the conversion increased
as expected. No deactivation could be observed at the different
temperatures. Between 528 and 588 K, the rate of formation of
ꢀ1
4
cm with an accumulation of 128 scans. Typically, the finely
ground samples (ca. 50 mg) were placed in the cup of the in situ
DRIFT reaction chamber, and different gas flows were passed
through the catalytic bed in an attempt to mimic the situation in
the real fixed-bed catalytic reactor. Furfural (Aldrich, 99%) was
MA and CO
2
increased faster than that of the resins, and conse-
accounts for a
quently, the increase in the yields of MA and CO
2
fed by flowing N
required to set the gas phase concentration was incorporated
into the feed line upstream of the saturator and prior to the DRIFT
cell. Mass flow controllers were used to set the N and O flows.
The DRIFT spectrum at room temperature for KBr loaded into the
cell was used as a background spectrum.
2
through a saturator at room temperature. The
large fraction of the increase in furfural conversion. Beyond
588 K, the yield of resins still increases, but the yield of MA
O
2
decreases remarkably, because MA becomes overoxidized to CO
at 633 K, it is almost fully oxidized to CO . Summarizing, there
are basically two routes of furfural transformation: furfural resini-
fication and furfural oxidation to form either MA or CO . The high-
2
;
2
2
2
2
est yield of MA (close to 30%) is attained between 588 and 603 K;
beyond these temperatures MA is overoxidized.
3
. Results and discussion
3.1. Catalytic activity
3.1.1. Effect of O
2
/furfural mole ratio
The same procedure was followed with different O
2
/furfural
Fig. 1 represents the catalytic behavior of the 8VA catalyst in the
temperature range 513–633 K. The catalyst is first contacted with
molar ratios for two different furfural concentrations (1 and
0.5 vol.% furfural); see Fig. 2 for the 1 vol.% results. (Those from
of 0.5 vol.% experiments are in Fig. S1 of the Supplementary Con-
tent). Temperature was increased from low to high temperature
and kept constant for 10 h. Each datum represents the average
for 10 h on stream. Initial transient data were not considered for
calculating the average value because the catalyst deactivated
rapidly during the first hours. No treatment was carried out
1
5
2 2 2
% furfural/5% O /N mixtures (O /furfural molar ratio 5:1) at
13 K, and the reaction temperature is increased stepwise (15 K
at each step). At the initial temperature (513 K), the early MA yield
is close to 10%, but in less than 2 h, the yield decreases to nearly
zero. The furfural conversion is initially close to 50%, but also
rapidly decreases to ca. 20%. The carbon dioxide yield remains close
to nil throughout the period in which the catalyst is held at 513 K.
The lack of carbon balance is ascribed to the formation of heavy
organics that cannot be analyzed by GC. It is well known that fur-
fural is highly prone to polymerization, forming furanic resins
between temperature steps, but for each new O
the catalyst was calcined under air flow at 773 K for 1 h prior to
measurement of the catalytic behavior at the new O /furfural mole
2
/furfural ratio
2
ratio. It has previously been verified that this calcination step com-
pletely regenerated the catalyst.
[
15,16]. Most of these products are condensed and stack in/down-
stream the catalytic bed as a consequence of its high boiling point.
As a matter of fact, when the catalyst is under stream for an
extended period, both the volume and the weight of the discharged
catalytic bed are visibly greater than those initially loaded, indicat-
ing that certain products have been formed and retained in the cat-
alytic bed during the course of the reaction. A minor fraction of
these heavy products end up retained by the solid catalyst depos-
ited at its surface (as will be shown later by TGA of the used cata-
lyst). The lack of carbon balance is hereafter accounted for by the
term ‘‘resins.” At 513 K, initially the resin yield is close to 35%, but
once steady state is reached, it accounts for nearly the entire con-
For experiments conducted with 1 vol.% furfural and for the sto-
ichiometric O
2
/furfural mole ratio required to obtain MA (2 mol
oxygen per 1 mol furfural), the highest MA yield is 17.3% at 618 K,
ꢀ
3 ꢀ1
equivalent to a TOF number of 0.19 ꢁ 10
2
s . When the O con-
centration becomes higher, the maximum achieved MA yield is
increased. Thus, for the experiment conducted at 1 vol.% furfural
2
and 10 vol.% O , the maximum yield of MA was 30.8% at 583 K,
ꢀ
3 ꢀ1
equivalent to a TOF number of 0.33 ꢁ 10
s . On the other hand,
for all the experiments conducted at 0.5 vol.% furfural (see Fig. S1),
the yield of MA is around 22–24%. For this series of experiments, the
maximum MA yield was 24% (equivalent to a TOF number of
100
100
618 K
633 K
603 K
80
60
40
20
0
80
60
40
20
0
5
88 K
5
73 K
5
58 K
5
43 K
5
28 K
513 K
0
10
20
30
40
50
60
70
80
time (h)
Fig. 1. Catalytic behavior of the 8VA catalyst with increase of the reaction temperature. Reaction conditions: 1 vol.% furfural and 5 vol.% O
balance N yield; ( ) resin yield.
. Symbols: (j) furfural conversion; (s) maleic anhydride yield; (4) CO
2 2
(O /furfural molar ratio = 5),
2
2
r

2
68
N. Alonso-Fagúndez et al. / Journal of Catalysis 348 (2017) 265–275
1
% furfural
O /F=10
1
00
100
80
60
40
20
0
O /F=2
O /F=5
O /F=20
2
2
2
2
8
0
0
0
0
0
6
4
2
5
20
560
600
640 520
560
600
640 520
560
600
640 520
560
600
640
Temperature (K)
Fig. 2. Catalytic behavior of 8VA catalyst under different O
2
/furfural mole ratios and with increasing reaction temperature. Furfural concentration: 1 vol.%. Symbols: (j)
) resin yield.
furfural conversion; (s) maleic anhydride yield; (4) CO yield; (
2
r
ꢀ
3 ꢀ1
0
.13 ꢁ 10
s
) obtained at 588 K and 10 vol.% O
2
. A simple inspec-
higher MA yields are obtained. Finally, it must also be borne in
mind that when the protocol of Figs. 1 and 2 is used, intense deac-
tivation takes place during the first hour. In contrast, when this
high-temperature protocol is used, the deactivation is less severe.
tion of the complete set of results also indicates that the tempera-
tures at which the maximum yields are achieved become smaller
with increasing the O
2
/furfural mol ratio (618, 588, 583, and
83 K for 1 vol.% furfural and O /furfural mol ratio of 2, 5, 10, and
0, respectively) but no remarkable differences in the MA yield
5
2
2
The highest yield of MA is obtained at higher O
and consequently, because the CO yield is not greatly affected by
the O /furfural ratio, the yield of resins is lower. At an O /furfural
2
concentrations
2
exist between the different reaction conditions. The same behavior
can be observed for experiments conducted with 0.5 vol.% furfural.
2
2
mol ratio of 20, the MA yield is always greater than 60% throughout
the time on stream investigated (the yield of resins is less than
20%). For the last time on stream investigated (14 h), the MA yield
3
2
.1.2. Exploring the effect of O /furfural mol ratio and of furfural
concentration when the reaction is started at temperatures with
complete furfural conversion
ꢀ
3 ꢀ1
was 63.7%, equivalent to a TOF number of 0.69 ꢁ 10
s . The lat-
ter values were double those obtained in a low- to high-
ꢀ3 ꢀ1
A new protocol was adopted to conduct the reaction. Rather
than starting at 513 K, the catalyst was directly contacted with
the gas reaction mixture at the lowest temperature required to
be practically at full conversion (the latter was selected based on
the results of the previous protocol used in Figs. 1 and 2). This pro-
cedure was applied in practice by flowing only oxygen through the
reactor for 1 h at 573 K, and the temperature was subsequently set
to the selected value and then the reaction mixture was let through
the reactor.
temperature protocol (30.8% and 0.33 ꢁ 10
is likely that even higher MA yields could have been obtained at
an O /furfural ratio higher than 20, although in a hypothetical
industrial application, this would have implied the use of air
enriched in O . It is quite remarkable that increasing the O /fur-
fural ratios results in depletion of the resin formation. It is well
known that furfural is slowly oligomerized in the presence of O
at room temperature [15], and extensive formation of resins would
be expected at higher O /furfural ratios. This observation indicates
s , respectively). It
2
2
2
2
2
Fig. 3 shows the time evolution of the product yield for three
that under these conditions, the formation of resins is not ther-
mally driven but catalyzed by surface sites. Formation of resins
cannot be prevented, but it can be remarkably minimized at com-
different O
of furfural (1 vol.%). The catalyst was regenerated by calcination
at 773 K in air for 1 h before proceeding with each new O /furfural
2
/furfural molar ratios with a constant concentration
plete conversion using an O
used for this set of experiments.
2
/furfural ratio ꢂ20 and the protocol
2
mol ratio. It was previously verified that the catalyst is completely
regenerated by this reactivation protocol.
In a new group of experiments (Fig. 4), we explored the effect of
Throughout the period of time investigated and for all the
experiments, the conversion was always above 99.5% (the conver-
changing the O
tion constant and changing the furfural concentration. In this case,
an O /furfural mol ratio as high as 40 could be explored without
using an O concentration >20 vol.%. The temperatures were again
initially set to reach (close to) 100% furfural conversion. The cata-
lyst was calcined at 773 K in air for 1 h prior to every new reaction
condition. Similarly to the experiments described in Fig. 3, very
high MA yields were obtained for all the experiments and higher
2 2
/furfural mol ratio but holding the O concentra-
sion is therefore not included in the figure) and the CO
2
yield was
2
in the range 10–15%. In contrast, the yield of MA smoothly but con-
tinuously diminished, and consequently, the lack of carbon balance
indicates that the yield of resins increased. Initially, the MA yield
2
was 60, 70, and 75% for O
respectively, and became 45, 50, and 60% after 15 h on stream.
Yield of CO did not change substantially with time on stream,
2
/furfural mol ratios of 5, 10, and 20,
O
2
/furfural ratios resulted in notably high MA yield and low resin
yield. Remarkably, negligible formation of resins was observed
for several hours when /furfural = 40 (furfural concentra-
2
and consequently the yield of resins behaves in the opposite man-
ner to MA and increases with time on stream. It is worth noticing
that, no matter what the reaction conditions and the time of reac-
tion, the MA yield was higher than the best MA yield found for the
low- to high-temperature protocol experiments shown in Fig. 2
O
2
tion = 0.25 vol.%). The possibility that the lower formation of resins
is due to a dilution effect of furfural when going from high to low
furfural conversion can be discarded because this dilution effect
was not observed in Fig. 2 or Fig. S1 (no significant differences in
the resin yield was observed in the figure when changing from 1
to 0.5 vol.%).
(
initiating the reaction at 513 K and then raising the temperature
close to full conversion). This demonstrates that when the gas mix-
ture initially contacts the catalyst at full conversion, substantially

N. Alonso-Fagúndez et al. / Journal of Catalysis 348 (2017) 265–275
269
Furfural concentration = 1 v/v%
1
00
O /F=5
O /F=10
O /F=20
2
2
2
8
6
4
2
0
0
0
0
0
0
5
10
15
0
5
10
15
0
5
10
15
time (h)
Fig. 3. Time evolution of the catalytic properties when the 8VA catalyst first contacts the reaction mixture at temperatures at which furfural conversion is complete. Constant
reaction conditions: furfural concentration = 1 vol.%. Variable reaction conditions: O concentration = 5, 10, and 20 vol.% and reaction temperatures set at 603, 593, and 573 K,
respectively. Symbols: (s) maleic anhydride yield; (4) CO yield; ( ) resin yield.
2
2
r
O concentration = 10 v/v%
2
1
00
O /F= 10
O /F= 20
O /F= 40
2
2
2
8
6
4
2
0
0
0
0
0
0
5
10
15
0
5
10
15
20
25
0
5
10
15
20
time (h)
Fig. 4. Time evolution of the catalytic properties when the 8VA catalyst first contacts the reaction mixture at temperatures at which furfural conversion is complete. Constant
reaction conditions: O concentration = 10 vol.%. Variable reaction conditions: furfural concentration = 1, 0.5, and 0.25 vol.% and reaction temperatures set at 593, 593, and
83 K, respectively. Symbols: (s) maleic anhydride yield; (4) CO yield; ( ) resin yield.
2
5
2
r
Deactivation was again evident in all experiments, but the MA
yield remained well above that observed in Figs. 1 and 2. Thus,
for the experiment conducted at 0.5 vol.% of FUR, 10 vol.% of O
and 593 K, the MA yield and TOF number were, respectively,
a thermal effect and due to the lower temperatures needed to con-
duct the reaction at higher O /furfural ratios cannot be fully ruled
out. But it seems that this is a secondary effect and that other
mechanistic aspects must be involved. Thus, in Fig. 4, it is clearly
2
2
,
ꢀ3 ꢀ1
6
3.2% and 0.34 ꢁ 10
maximum MA yield achieved with the same O
tion but under the low- to high-temperature protocol (which were
s
, almost three times higher than the
observed that the same temperature (593 K) was used for O
fural of both 10 and 20 and the highest yield of MA was obtained
with O /furfural of 20 (for the whole time on stream investigated).
2
/fur-
2
and FUR concentra-
2
ꢀ
3 ꢀ1
2
4% and 0.13 ꢁ 10
s
, respectively). And for the experiment
We will come back to this point at the end of the Discussion
section.
Summarizing, a notably higher yield of MA can be achieved at
full furfural conversion if the fresh catalyst is initially contacted
with the gas phase reaction mixture at temperatures required for
complete furfural conversion. In contrast, a much lower MA yield
conducted at 0.25 vol.% of FUR, 10 vol.% of O
2
, and 583 K, the MA
ꢀ3
ꢀ1
yield was initially close to 86.0% (TOF = 0.23 ꢁ 10
s
) and after
ꢀ3
ꢀ1
2
0 h on stream it was 70.2% (TOF = 0.19 ꢁ 10
s
).
The reason for the improvement of the MA yield when a higher
2
/furfural ratio was used is not clear. The possibility that it is just
O

2
70
N. Alonso-Fagúndez et al. / Journal of Catalysis 348 (2017) 265–275
is obtained if the low- to high-temperature protocol of increasing
the temperature from low to high conversion is followed.
the signals are so weak that a single peak at an average position
of around 287.5 eV was used for a correct simulation of C1s in
the fresh sample. The presence of these oxygenated hydrocarbons
in the used catalyst is also evident in the O1s peak at 533.1 eV
accompanying the contribution at 531.2 ± 0.1 eV from oxygen in
alumina. The former O1s core level was not present in the fresh
sample and confirms the presence of oxygenated hydrocarbon spe-
cies [23]. Moreover, the XPS O/V at. ratio of the used catalyst is
higher than that of the fresh catalyst (2.39 and 2.10, respectively)
as a consequence of the deposition of such oxygenated hydrocar-
bons. The calcination of the used catalyst at 773 K results in a
remarkable decrease in the XPS C/Al and O/Al at. ratios (from
3.2. Characterization of the used catalyst
Fig. 5 shows the TGA and DTG (differential thermal analysis)
analysis obtained after the catalyst was in contact with a reaction
mixture consisting of 0.5 vol.% furfural and 2.5 vol.% O (balance
) at 633 K for 10 h. The catalyst was initially in contact with
2
N
2
the reaction mixture at 513 K and temperature increased in the
stepwise manner described in Fig. 2. The catalytic bed was down-
loaded and the grains of the used catalysts (grayish) were easily
picked out from the inert silica (white) and from resin particles
1
.09 to 0.17 and from 2.39 to 2.10, respectively) and the disappear-
ance of the O1s signal at 533.1 eV. All previous results prove the
substantial removal of these resinlike deposits.
For the V analysis, the V2p3/2 core level was selected to monitor
the vanadium species because the highly intense O1s peak over-
shadows the V2p1/2 contribution. In this region of the spectrum,
(
black). Two weight loss processes can be observed. The first one,
identified in the DTG as peaking at 370 K, is assigned to the
removal of physisorbed water. Another weight loss process peak-
ing at 720 K, representing near 5 wt.% of the weight of the used cat-
alyst, corresponds to the burning off of C-containing products.
These products are compatible with the formation of heavy hydro-
carbons (resins) described when the activity data are discussed.
TGA results evidence that part of these resin-like products remain
deposited over the catalyst and, moreover, that they can be
removed by calcination at 773 K.
3 4
two O1s satellite lines from MgKa and MgKa emission lines
appear (because a monochromator was not used), which interferes
with straightforward analysis of the V2p1/2 line [24]. These two
satellite lines occurred at 8.4 and 10.2 eV lower BE than the main
O1s core level (from the MgKa1,2 line) and possessed relative inten-
sities of 0.08 and 0.041, respectively. The deconvolution of this
region was accomplished by considering all of the above signals
and considering that the area and BE of the V2p1/2 core level are
Fig. 6 shows the relevant core levels obtained by XPS analysis of
fresh and used 8VA catalyst. The used sample corresponds to that
studied by TGA. XPS studies were also extended to the effect that
calcination at 773 K (regenerated catalyst) has on the chemical sur-
face properties. Table 1 summarizes the XPS data obtained for the
different core levels. The Al2p core level for Al O was chosen as a
2 3
reference (74.4 eV) to compensate for the charge effects and accu-
rately determine the BEs of the different XPS signals.
Comparison of the intensities of the C1s signals of fresh and
used catalysts (the adventitious carbon signal in the fresh catalyst
arises from ubiquitous organic chemicals contamination present in
ambient air) indicates that a remarkable deposition of hydrocar-
bonaceous species took place over the surface of the catalyst as a
consequence of the reaction environment (the XPS C/Al at. ratio
went from 0.12 to 1.09, ca. 10 times greater). This observation is
in agreement with the conclusions of the TGA results. Besides a
very intense peak at 284.8 eV from hydrocarbon-like species, the
presence of oxygenated species in the used catalyst is clear in
the C1s signals at 286.2 and 288.5 eV (CAO and C@O species)
one-half and at 7.4 eV higher BE than those of V2p3/2
.
The BE of the V2p3/2 core level for all the samples investigated
here is at 517.6 ± 0.1 eV, indicating that V(V) species predominate.
The presence of V(IV) or V(III) species cannot be rejected, especially
in the used catalyst, but the presence of such reduced V species
could not be confirmed. When the overall deconvolution of the
O1s and V2p core levels in this region was conducted, the presence
of those species was not required for a statistically good deconvo-
lution, indicating that if they are present, they must be at much
lower concentrations than V(V) species. What is more relevant is
that the V/Al at. ratios in all samples are above the nominal value,
in agreement with the presence of well-dispersed polyvanadate
species covering the surface of the support [5]. These ratios for
the fresh and used catalysts are very much alike, but the reaction
environment and the further calcination of the catalyst at 773 K
might affect this dispersion. because the XPS V/Al at. ratio of the
calcined used catalyst is slightly less than that of the fresh catalyst
[
22]. The latter peaks are almost absent in the fresh catalyst, and
(
0.17 vs. 0.13).
1
00
In situ DRIFT studies conducted under relevant reaction condi-
tions are quite helpful in identifying the hydrocarbonaceous spe-
cies present over the surface of the catalyst under those
conditions [25] and, as shown below, will offer clues to under-
standing why the thermal protocol is critical for defining the
potential yield of MA. Two different experiments were conducted
9
8
6
4
2
0
8
9
9
9
9
8
by flowing a reaction mixture containing furfural and O
2
at of
0
.25 and 10 vol.%, respectively (O /furfural mol ratio = 40), through
2
the catalyst bed placed in the in situ DRIFT cell (these conditions
are those used in the last graph of Fig. 4). In a first set of experi-
ments, the catalyst was contacted directly with the gas mixture
at 573 K, aiming at simulating a very high furfural conversion.
DRIFT spectra were recorded after different times on stream
(Fig. 7). Although the contact mode used in the DRIFT cell (flowing
through the catalyst bed) is not exactly that taking place in the
fixed bed reactor, it is still valid to identify species at the catalyst
surface and to draw relevant conclusions. In a second set of exper-
iments, the catalyst initially contacted the gas mixture at 523 K
3
00 400
500 600
700 800 900 1000 1100 1200
(
representative of a situation in which the conversion is low and
Temperature (K)
after 120 min on stream, the temperature is increased to 573 K).
All these latter results are collected in Fig. 8. These two experi-
ments are representative of two different reaction mixtures. The
Fig. 5. TGA of the used 8VA catalyst under a 20 vol.% O
2
2
/N flow and a heating ramp
ꢀ
1
of 10 K min. (see text for a description of the aging conditions).

N. Alonso-Fagúndez et al. / Journal of Catalysis 348 (2017) 265–275
271
Al 2p
O1s
20000
2000
V2p1/2
V2p3/2
C 1s
fresh
fresh
fresh
2000
used
used
used
used and
calcined at 773 K
used and
calcined at 773 K
used and
calcined at 773 K
68
70 72 74 76 78 80 82
BE (eV)
294 291 288 285 282 279 276
BE (eV)
540 535 530 525 520 515 510
BE (eV)
Fig. 6. Al2p, C1s, O1s, and V2p core levels of fresh 8VA catalyst, used 8VA catalyst, and used 8VA catalyst after calcination under air at 773 K (regenerated).
Table 1
Summary of XPS results obtained with fresh, used, and fresh and calcined 8VA catalyst.
BE (eV)a
XPS at. ratiosa
V/Al
V2p3/2
517.6
Al2p
74.4
O1s
C1s
O/Al
2.10
C/Al
Fresh
Used
531.2
284.8 (61%)
0.17 (0.09)b
0.12
2
87.5 (39%)
517.5
517.7
74.4
74.4
531.3 (86%)
284.8 (37%)
286.2 (18%)
2
0.16 (0.09)b
0.13 (0.09)b
2.39
2.10
1.09
533.1 (15%)
88.5 (45%)
Used and calcined at 773 K
531.3
284.8 (72%)
0.17
287.8 (28%)
a
Value between brackets corresponds to area percentage.
Nominal V/Al at. ratio deduced from the V loading incorporated during the synthesis.
b
Mal
F
MAc
MA
0.5 a.u
0
.1 a.u
F
F
MA
5
1
min.
F
0 min.
1
3
5 min.
0 min.
5
min.
1
0 min.
5 min.
6
1
0 min.
1
20 min.
3
0 min.
0 min.
6
5
73 K
573 K
120 min.
4
000
3600
3200
2800
2000
1800
1600
1400
1200
-1
-1
wavenumber (cm )
wavenumber (cm )
Fig.7. DRIFT spectra obtained under a flow of the furfural/O
2 2 2
/N mixture (0.25 vol.% furfural, 10 vol.% oxygen, and balance N ) at 573 K. Fresh 8VA catalyst is contacted
directly at 573 K for different times on stream (from top to bottom, t = 5, 10, 15, 30, 60, and 120 min). F, Mal, and MAc stand for furfural, maleates, and malic acid, respectively.

2
72
N. Alonso-Fagúndez et al. / Journal of Catalysis 348 (2017) 265–275
Mal
0
.5 (a.u.)
MA
0
.1 (a.u.)
F
MAc
Mal
5
1
min.
MA
0 min.
1
3
5 min.
0 min.
5
min.
6
1
0 min.
20 min.
1
1
0 min.
5 min.
5
23 K
3
6
1
0 min.
0 min.
523 K
573 K
x 0.2
20 min.
5
73 K
1
50 min.
1
50 min.
x 0.2
2000
4
000
3600
3200
2800
1800
1600
1400
1200
-1
-1
wavenumber (cm )
wavenumber (cm )
2 2 2
Fig. 8. DRIFT spectra obtained under flow of the furfural/O /N mixture with 0.25 vol.% furfural and 10 vol.% oxygen (balance N ) at 523 and 573 K. Fresh 8VA catalyst is first
contacted with the reaction mixture at 523 K for different times on stream, the temperature is subsequently raised at 573 K, and a new spectrum is recorded after 150 min on
stream. F, Mal, and MAc stand for furfural, maleates, and malic acid, respectively.
Table 2
first experiments obviously correspond to those that produce the
maximum MA yield, whereas the second series illustrate those
used under the protocol of initially contacting the catalyst with
reaction conditions at low temperature and then increasing the
temperature to get higher conversion. Finally, in both cases, after
remaining on stream for a given period of time, the feed was
switched to pure N (Figs. S2 and S3). The spectra collected at dif-
2
ferent times after the switching convey information on the surface
species involved in the reaction.
Assignments of the DRIFT bands of spectra represented in Figs. 7 and 8.
Wavenumbers
Assignment
Reference
ꢀ
1
(
cm
)
3133
CAH stretching in C@C in furfural, MA, MAc
and maleate
[26–29]
3125 and 3080 CAH stretching in C@C in furfural, MA, MAc
This work
This work
and maleate
CAH stretching in strongly chemisorbed CAC
containing species
2980, 2942,
and 2882
2
1
1
1
843 and 2811 CAH stretching in aldehyde group of furfural
853
786
[27,28,35]
[30,34]
Fig. 7 represents the spectra collected by contacting the fresh
catalyst with the reaction mixture directly at 573 K. Each spectrum
represented in Fig. 7 is the average obtained during the period of
spectrum collection (approximately 4 min) and is the result of sub-
tracting the spectrum of the catalyst before contact with the reac-
tion mixture from that obtained at a given time on stream.
Fortunately, IR techniques have been extensively used for the
study of the chemisorption of furfural on different metal oxides
and for MA production from butane on different vanadium cata-
lysts [26–34]. Those studies are very helpful in identifying the dif-
ferent species present in our DRIFT studies. Table 2 summarizes the
assignments of the DRIFT bands and the corresponding references
on which the assignment of the band has been previously done.
C@O symmetric str. in MA
C@O asymmetric stretching in MA
724 and 1714 C@O str. in maleic acid
[30]
[26–28]
[31]
1693 and 1671 C@O stretching in furfural
1640–1650 C@C stretching in maleic acid or in maleate
1568
Furfural ring stretching mode
[26–
2
8,32,33]
1
1
540 or 1570
430 or 1473
C@O asymmetric stretching in maleate
C@O symmetric stretching in maleate
[30,31,34]
1473 and 1464 Furfural ring stretching modes
[26–
2
8,32,33]
1
1
393 and 1367 Furfural ring stretching modes
[26–
2
8,32,33]
340 and 1258 Stretching CAC and/or bending CAH
This work
vibrations of strongly chemisorbed species
Bending CAH in maleate
ꢀ1
The DRIFT features in the region 2000–1200 cm are first dis-
cussed. In this region, relevant C@O, CAH, C@C, and CAC function-
alities are IR active. If none of the bands prevail over the rest, the
overall spectrum is quite featureless. Indeed, in the first spectrum
recorded after 5 min on stream, bands at 1786, 1714, 1473 and
1303
[31]
[26–
28,32,33]
1280 and 1241 In-plane CAH deformation modes in furfural
ring
ꢀ
1
1
430 cm are observed and superimposed over a notably wide
which suggests a chelating bidentate carboxylate, although the
bridging configuration cannot be ruled out. The bands at 1853
and 1786 cm have been assigned to the presence of maleic anhy-
ꢀ1
feature that peaks near 1540 cm . As time on stream increases,
other bands at 1853, 1671, 1393, 1340, 1280- and 1258 cm are
revealed. The bands at 1540 and 1430 cm have been assigned
previously to the formation of carboxylate (maleate) species in a
bidentate configuration [30,31,34]. The consensus is that the sepa-
ꢀ
1
ꢀ1
ꢀ1
dride species [30,34]. The remaining bands at 1714, 1671, 1473,
ꢀ1
1
393, 1340, 1280, and 1258 cm can be better assigned by inves-
tigating the spectra that appear when furfural and O
and only N flushes the catalyst at 573 K. These results are
explained and discussed in more details in Supplementary Content
2
are removed
ration between these bands (
D
m
) (asymmetric and symmetric CAO
of
50–500 cm is expected for unidentate carboxylate and 150–
2
stretching bands) depends on the type of carboxylate. Thus
D
m
ꢀ
1
3
(
Fig. S2), but summarizing this experiment, we conclude that the
ꢀ1
1
80 and 60–100 cm for bridging and chelating bidentate car-
ꢀ1
bands not assigned in Fig. 7 at 1671, 1473, 1393, and 1280 cm
arise from furfural and that at 1714 cm from maleic acid.
ꢀ1
boxylates, respectively [31]. In this case,
D
m
is around 110 cm
,
ꢀ1

N. Alonso-Fagúndez et al. / Journal of Catalysis 348 (2017) 265–275
273
Fig. S2 also demonstrates that furfural, MA, and maleic acid,
which were present at the surface of the catalyst when contacted
with the furfural–O –N stream, are desorbed by switching to N .
2 2 2
and angle defined by the CAOAmetal bonds) [31,36]. We tenta-
tively assigned this band to the symmetric stretching band of
another carboxylate species; the antisymmetric counterpart
should be now overshadowed by the most intense peak at
This indicates that these adsorbed species are involved in the reac-
tion mechanism. In contrast, the bands arising from maleate species
ꢀ1
1570 cm . Again, this carboxylate species is very likely a chelating
ꢀ1
ꢀ1
(
1540 and 1430 cm ) were absent in the spectra of Fig. S2, and this
means that maleate species remain adsorbed at the surface of the
catalyst after switching to N . Therefore, they are not directly par-
bidentate carboxylate because
D
m
is around 100 cm . Another
ꢀ1
assignment is also possible for this 1473 cm band, but it will
be proposed later when the 4000–2600 cm region is discussed.
ꢀ
1
2
ꢀ
1
ticipating in the reaction mechanism. As we will see below, these
species are very likely involved in the deactivation of the catalyst.
Finally, the weak band at 1360 cm that appears only at shorter
times on stream (from 0 to 15 min) and is subsequently overshad-
owed by the most intense bands is assigned to furfural; the
remaining bands of furfural were overshadowed by the bands from
maleates, MA, and maleic acid.
ꢀ1
The bands at 1340 and 1258 cm remain elusive. Stretching
CAC and bending CAH vibrations frequently appear in this region
of the spectrum, and they must arise from other species than MA,
maleic acid and furfural, because they remain deposited over sur-
After 120 min on stream at 523 K, the temperature was
increased to 573 K, and the spectrum after 150 min at this temper-
ature is included in Fig. 8. This spectrum is similar to those previ-
ously described at 523 K, except that the bands are now much
more intense (the intensity was divided by 5 for inclusion in the
figure). This result indicates that MA and maleate species are also
present at the surface of the catalyst, but with a concentration
nearly five times higher than those at 523 K. At this point it is also
very important to notice by a simple comparison of the scales of
the figures that when the fresh catalyst was directly contacted
2
face after switching to N .
ꢀ
1
Returning to Fig. 7, the 4000–2600 cm region displays CAH
stretching vibrations at 3125, 3080, 2980, 2942, 2843, and
ꢀ1
ꢀ1
2
811 cm . The negative peaks at 3719, 3672, and 3605 cm indi-
cate that the OH groups of the alumina support are involved in the
ꢀ1
adsorption of different species. The bands at 3125 and 3080 cm
arise from HAC stretching vibrations in C@C bonds, and therefore
they are compatible with the presence of furfural, MA, and MAc
ꢀ1
and maleate species. The bands at 2843 and 2811 cm are CAH
stretching vibrations of the aldehyde group of furfural [27,28,35].
2 2
with furfural/O /N mixture at 573 K (Fig. 7), the bands from mal-
ꢀ1
The bands at 2980 and 2942 cm are typical of HAC stretching
vibrations in saturated CAC bonds, and therefore they indicate
the presence of other species different from any of those men-
tioned thus far, because none of these species possess single CAC
bonds. We revisit these two bands later when discussing Fig. 8.
Summarizing the DRIFT studies conducted by contacting the
fresh catalyst with the reaction mixture at 573 K, it can be said that
maleate species, furfural, MA, and maleic acid are present at the
surface of the catalyst. The latter three species are desorbed once
eate and MA were much less intense than when it was contacted
previously at 523 K and then temperature was raised to 573 K. This
indicates that maleates do not decompose on reaching higher tem-
peratures but, on the contrary, the rate of deposition increases.
Examining the region of CAH stretching vibrations, bands at
ꢀ1
3125, 3080, 2980, 2942, 2882, 2843, and 2811 cm are observed,
and their intensity increases with time on stream. The features at
ꢀ1
3125 and 3080 cm are characteristic of CAH stretching vibra-
tions in C@C bonds and are in principle assigned to furfural. Maleic
anhydride or maleic acid cannot be ruled out, either. The presence
furfural and O
species and other spectator species do not desorb and remain
extensively over the surface sites after switching to N
2
are removed from the feed. In contrast, the maleate
ꢀ1
of furfural also explains the bands at 2843 and 2811 cm assigned
2
.
to the CAH bond of the aldehyde group of furfural [27,28,35]. The
ꢀ1
Fig. 8 represents the spectra recorded by contacting the fresh
catalyst with the reaction mixture first at 523 K. After 120 min
on stream, the temperature was increased to 573 K. The presence
of maleic anhydride is evident from the bands at 1853 and
negative peaks at 3719, 3672, and 3605 cm denote the perturba-
tion of OH groups at the surface of the alumina by the chemisorp-
tion of different species.
ꢀ1
The features at 2980, 2942, and 2882 cm arise from stretching
ꢀ
1
1
1
786 cm [30,34] and that of maleic acid from the shoulder at
CAH vibrations in saturated CAC bonds. The first two bands were
ꢀ1
714 cm [30]. A simple comparison of the intensity of the bands
previously observed in Fig. 7, but now they are much more intense
ꢀ1
ꢀ1
arising from maleate species (now at 1570 and 1430 cm ) in
Figs. 7 and 8 (both at quite the same scale) reveals that these bands
in Fig. 8 are much more intense than those in Fig. 7. The different
position of the bands with respect to the carboxylate species of
Fig. 7 suggests that the new reaction conditions in Fig. 8 have chan-
ged the exact nature of the carboxylate species. The maleate bands
experience a continuous and intense growth with time on stream,
which means that these species are rapidly built up over the sur-
face of the catalyst at 523 K. In contrast, those at 1853 and
(it is likely that the band at 2880 cm
was too weak to be
observed in Fig. 7). As indicated in the discussion of Fig. 7, none
of the species so far detected by DRIFT (furfural, maleic anhy-
dride/acid, or maleate) have saturated CAC bonds, and conse-
quently, new species with saturated CAC bonds must be present.
It is reasonable to assign these bands to the resin-like products that
are produced in the course of the reaction. These bands are more
intense in the spectra of Fig. 8 collected at 523 K because, as
demonstrated in the catalytic experiments, at this temperature,
the production of resins is much faster than under the conditions
of Fig. 7, and therefore, the bands from these resin species must
also be more intense. In principle, saturated CAC bonds are not
present in furfural resins, but in Supplementary Content, a route
that explains its formation is proposed. In this context, the pres-
ꢀ1
1
786 cm from MA are less intense. The C@C stretching vibration
ꢀ1
at 1640 cm and the CAH bending mode of maleate and maleic
anhydride at 1303 cm
ꢀ
1
[31], which are difficult to detect in
Fig. 7, are now more clearly observed.
ꢀ1
Another intense band at 1473 cm is clearly detected. Assign-
ment of this band to one of the stretching modes of the furfural
ring, although reasonable (in fact earlier assigned in Fig. 7), must
be ruled out in this case, because the most intense band from fur-
ence of CH
assignment of the band at 1473 cm to the CAH scissoring bend-
ing mode of CH [37]. The band was attributed above to chelating
bidentate carboxylate species. Irrespective of what the real assign-
ment is (both assignments can be correct), the outcome is always
the same: maleate and resin deposition.
2
units in the resin deposits is compatible with the
ꢀ1
2
ꢀ1
fural at 1673 cm could not be observed. Therefore, this band at
ꢀ1
1
473 cm must be due now to a different species than furfural.
As mentioned above, the position of carboxylate bands derived
from dicarboxylic acids depends strongly on the type of carboxy-
late (unidentate, bridging bidentate, and chelating bidentate), but
also on the metal oxide on which the carboxylate is adsorbed
2
The spectra after switching to N flow at 573 K were also col-
lected (Fig. S3). The results indicate (a more detailed discussion
is given in Supplementary Content) that MA, maleic acid, and fur-
fural are desorbed. No negative band indicating maleate desorption
(
for instance on V oxide or alumina) and on the geometry (length

2
74
N. Alonso-Fagúndez et al. / Journal of Catalysis 348 (2017) 265–275
100
was detected, even if a much higher concentration of surface
maleates had been built up during the time on furfural/O
stream.
2 2
/N
Conv. decr. T
Y MA decr. T
Conv. inc.T
Y MA inc.T
Summarizing all the results of the DRIFT figures, it can be con-
cluded that when the fresh catalyst is initially contacted with the
reaction mixture at 523 K (a representative temperature for very
low conversion), surface maleates and resin-like products are
deposited at a considerably higher rate than if contacted directly
at 573 K (a representative temperature for close to full conversion).
If after being at 523 K the temperature is increased to 573 K, this
effect is even more remarkable. In contrast, if the fresh catalyst ini-
tially contacts the reaction mixture at a high temperature at which
the conversion is close to completion, the deposition rates of mal-
eates and resins significantly decreased.
Maleates and resins must be involved in the deactivation of the
route of formation of MA, either by poisoning the active sites or by
restricting the access of the reactants (fouling). If the reaction is ini-
tiated at low temperature, these species are rapidly deposited and
they quickly deactivate the catalyst. Actually, less than 1 h is
needed to deeply deactivate the catalyst, as stressed earlier when
Fig. 1 was discussed: the conversion at 513 K was initially 50%
and the MA yield was 10%, but they decrease rapidly, and after
8
6
4
2
0
0
0
0
0
460
480
500
520
540
560
580
600
Temperature (K)
Fig. 9. Hysteresis behavior of the 8VA catalyst evidenced by cycling the temper-
ature from high to low temperatures and back to high temperatures in the range
93–493 K. Reaction conditions: 1 vol.% furfural and 10 vol.% O (balance N ).
2 2
Symbols: j and h, furfural conversion; d and s, maleic anhydride yield; open
5
1
h these values were 20% and 3%, respectively. The increase in
symbols: decreasing temperature; closed symbols: increasing temperature.
the temperature reinforces the deactivation because, as shown by
DRIFT, the deposition rate of those species accelerates. Once these
products have been deposited, the only way to regain a fresh sur-
face and remove these species is to burn them off under air at 773 K.
An estimation of the activation energy for MA and resin forma-
Although the values for conversion of furfural are quite similar
for the cooling and heating ramps, the yield of MA shows clear hys-
teresis behavior and depends on the manner in which the temper-
ature was set. Thus, an average MA yield close to 54% was reached
when the catalyst was first contacted with the reaction mixture at
tion conducted using 1 vol.% furfural and 10 vol.% O
2
shows that
the value for MA formation is significantly higher than that for
ꢀ1
ꢀ1
5
93 K (more precisely, as reported in Fig. 3, an initial yield of 67%
resins (98.3 ± 9.9 kJ mol vs. 16.2 ± 0.5 kJ mol ). Similar values
were found for other reaction conditions. The difference in the acti-
vation energy for MA and resin formation implies that as the tem-
perature increases, the rate of MA formation is enhanced over that
of resins. Very likely, this is also the situation for maleates. Conse-
quently, the rapid accumulation of heavy resins and/or maleates
over the surface is relatively retarded if the catalyst is contacted
with the reaction mixture at high temperatures (at which conver-
sion is full). Then the deactivation rate is less intense and the cat-
alyst can display its full potential to selectively oxidize furfural to
MA for a longer time.
MA was obtained at the beginning, but as stated above, the catalyst
suffered from progressive deactivation and was near 50% yield
after 10 h on stream; the value represented in the figure was an
average for the last 5 h). In contrast, when 593 K was reached in
the low- to high-temperature mode once the catalyst had been at
low temperatures (closed symbol), the MA yield was only ca.
4
2%. This value is still higher than that presented in Fig. 2, close
to 30%, but this again stresses the importance of the selection of
the initial temperature of contact with the reaction mixture to
define the effective MA yield. The same can be concluded for the
yield of MA at 573 K, although the differences in MA yield between
the down and up ramps are smaller at this temperature. At 553 K
and below, the cooling and heating MA curves overlap. Therefore,
the thermal protocol is of paramount importance for higher perfor-
mance, especially for MA yield and catalyst stability: contacting
the fresh catalyst directly with the reaction mixture at tempera-
tures high enough to have furfural conversion close to 100% results
in promotion of selective conversion of furfural to MA. Besides, it is
worth noticing that once the catalyst contacts the reaction mixture
at lower temperatures, it gets deactivated in terms of MA yield.
Under the conditions of Fig. 9, this temperature threshold is that
In an attempt to gain more evidence of the importance of ini-
tially contacting the catalyst with the reaction mixture at either
high temperature (close to full conversion) or low temperature
(
below 40%), an additional experiment was conducted. Instead of
using the low- to high-temperature protocol depicted in Figs. 1
and 2 (increasing the temperature from low to high conversion),
the catalyst was directly contacted with the gas reaction mixture
initially at 593 K. Under these conditions the conversion was close
to 100%. Then the temperature was decreased in 20 K steps down
to 473 K (low conversion) and then back again to 593 K (full
conversion).
giving rise to conversions below 80%, but at other furfural and O
concentrations, this threshold may change.
2
In practice, only oxygen was flowed through the reactor for 1 h
at 573 K, and then temperature was increased to 593 K. Subse-
It seems reasonable to assume that contacting at high temper-
ature (full conversion) is not the key parameter for reaching high
MA yields, but that the catalyst is subjected to a reaction mixture
with high oxidizing potential. A high oxidizing potential can be
accomplished at high temperatures, but also by working at high
quently, the furfural/N
stream to have 1 vol.% furfural and 10 vol.% O
0 h on stream, the temperature was decreased to 493 K in 20 K
2
stream was incorporated into the O
2
2
(balance N ). After
2
1
steps and increased again back to 593 K. The catalyst was on
stream 10 h at each temperature. The points represented in the fig-
ure are the average catalytic activity for 10 h (steady state), except
for the first measurement, conducted at 593 K, during which con-
tinuous deactivation was observed. For this result, the value repre-
sented was the average for the last 5 h. In the remainder of the
results, no sign of deactivation was noted and the values were
averaged for the complete 10 h period. Fig. 9 represents the cat-
alytic properties obtained.
2
O /furfural ratios. A Mars–van Krevelen mechanism is operative
in this reaction [19,20]. Consequently, O atoms from the surface
vanadium oxide species are involved in the selective oxidation.
In the course of the reaction, the V(V) sites are reduced, providing
the O atoms to furfural and/or to an intermediate. The capacity of
oxidizing the reduced vanadium back to V(V) is essential to heal
the O vacancy. The oxidation of the reduced catalyst by gaseous

N. Alonso-Fagúndez et al. / Journal of Catalysis 348 (2017) 265–275
275
O
2
is the rate-determining step. In this context, a high oxidizing
from polystyrene waste for reactions of interest in biomass valorization, Catal.
Today 34 (2014) 285–294.
9] N. Alonso-Fagundez, I. Agirrezabal-Telleria, P.L. Arias, J.L.G. Fierro, R. Mariscal,
potential of the reaction mixture means a rapid and efficient
restoration of the V(V) sites, the promotion of the selective route
to MA, and, consequently, the lessening of the unselective forma-
tion of resins and maleates. Deactivation by resin and/or maleate
deposition, although unavoidable, is substantially slowed down
at high oxidizing potential of the reaction mixture.
[
M. López, Granados, Aqueous-phase catalytic oxidation of furfural with H
high yield of maleic acid by using titanium silicalite-1, RSC, Adv. 4 (2014)
4960–54972.
2 2
O :
5
[
10] G. Pavarelli, J. Velasquez Ochoa, A. Caldarelli, F. Puzzo, F. Cavani, J.L. Dubois, A
new process for maleic anhydride synthesis from a renewable building block:
the gas-phase oxidehydration of bio-1-butanol, ChemSusChem 8 (2015) 2250–
2259.
[
[
11] X. Li, Y. Zhang, The conversion of 5-hydroxymethyl furfural (HMF) to maleic
anhydride with vanadium-based heterogeneous catalysts, Green Chem. 18
4
. Conclusions
2 5 2 3
When V O /c-Al O catalysts are used in the gas phase oxida-
(
2016) 643–647.
12] A. Chatzidimitriou, J.Q. Bond, Oxidation of levulinic acid for the production of
maleic anhydride: breathing new life into biochemicals, Green Chem. 17
tion of furfural to MA, the catalyst is unavoidably deactivated by
deposition of maleates and/or resinlike products. The regeneration
of the catalyst is entirely achieved by burning off these deposits at
(
2015) 4367–4376.
[13] A.S. Mamman, J.M. Lee, Y.C. Kim, I.T. Hwang, N.J. Park, Y.K. Hwang, J.S. Chang, J.
S. Hwang, Furfural: hemicellulose/xylose-derived biochemical, Biofuels,
Bioprod. Biorefin. 2 (2008) 438–454.
14] J.J. Bozell, G.R. Petersen, Technology development for the production of
biobased products from biorefinery carbohydrates – the US Department of
Energy’s ‘‘top 10” revisited, Green Chem. 12 (2010) 539–554.
15] K.J. Zeitsch, The Chemistry and Technology of Furfural and Its Many By-
Products, Sugar Series, vol. 13, Elsevier, The Netherlands, 2000.
[16] R. Mariscal, P. Maireles-Torres, M. Ojeda, I. Sádaba, M. López, Granados,
furfural: a renewable and versatile platform molecule for the synthesis of
chemicals and fuels, Energy Environ. Sci. 9 (2016) 1144–1189.
7
73 K.
When the fresh catalyst initially contacts the reaction mixture
[
at low temperature, an intense deposition of maleates and resins
takes place and the catalyst is rapidly and deeply deactivated.
The increase of the temperature from low to high (from low to full
conversion) does not result in removal of deposits, but accelerates
their deposition. Consequently, the catalyst remains in a deacti-
vated state, and in practice the yield of MA never exceeds 30%, irre-
spective of the reaction conditions.
In contrast, when the fresh catalyst initially contacts the reac-
tion mixture at high temperature (close to full conversion), the rate
of deposition of maleates and resins is severely slowed down. Con-
sequently, the yield of MA is remarkably enhanced and the deacti-
vation rate is greatly diminished. Under this high-temperature
contact protocol, the MA yield can also be further improved by
[
[
[
[
17] E.R. Nielsen, Vapor phase oxidation of furfural, Ind. Eng. Chem. 41 (1949) 365–
368.
18] N.A. Milas, W.L. Walsh, Catalytic oxidations. I. Oxidations in the furan series, J.
Am. Chem. Soc. 57 (1935) 1389–1393.
19] K. Rajamani, P. Subramanian, M.S. Murthy, Kinetics and mechanism of vapor
phase oxidation of furfural over tin vanadate catalyst, Ind. Eng. Chem. Prod.
Res. Dev. 15 (1976) 232–234.
20] M.S. Murthy, K. Rajamani, Kinetics of vapour phase oxidation of furfural on
vanadium catalyst, Chem. Eng. Sci. 29 (1974) 601–609.
21] C.D. Wagner, Sensitivity factors for XPS analysis of surface atoms, J. Electron
[
[
[
Spectrosc. Relat. Phenom. 32 (1983) 99–102.
22] S. Biniak, G. Szyma n´ ski, J. Siedlewski, A. S´ wiatkoski, The characterization of
2
increasing the O /furfural mole ratio. Therefore, the high oxidizing
activated carbons with oxygen and nitrogen surface groups, Carbon 35 (1997)
1
[23] J.L. Figueiredo, M.F.R. Pereira, M.M.A. Freitas, J.J.M. Órfão, Modification of the
surface chemistry of activated carbons, Carbon 37 (1999) 1379–1389.
24] C.L. Pieck, S. Del Val, M. López Granados, M.A. Bañares, J.L.G. Fierro, Bulk and
surface structures of V /ZrO systems and their relevance for o-xylene
potential of the reaction mixture seems in practice the key param-
eter to favor the selective transformation to MA and to prevent
resin and maleate formation and subsequent deactivation.
799–1810.
[
2
O
5
2
Acknowledgment
oxidation, Langmuir 18 (2002) 2642–2648.
[
25] J.V. Ochoa, C. Trevisanut, J.M.M. Millet, G. Busca, F. Cavani, In situ DRIFTS-MS
study of the anaerobic oxidation of ethanol over spinel mixed oxides, J. Phys.
Chem. C 117 (2013) 23908–23918.
[26] G. Allen, H.J. Bernstein, Internal rotation: VIII. The infrared and Raman spectra
of furfural, Can. J. Chem. 3 (1955) 1055–1061.
27] M. Rogojerov, G. Keresztury, B. Jordanov, Vibrational spectra of partially
oriented molecules having two conformers in nematic and isotropic solutions:
furfural and 2-chlorobenzaldehyde, Spectrochim. Acta Part A 61 (2005) 1661–
Financial support from the Spanish Ministry of Economy and
Competitiveness is gratefully acknowledged Project (CTQ2015-
6
4226-C3-1-R).
[
Appendix A. Supplementary material
1670.
[
28] P. Adámek, K. Volka, Z. Ksandr, I. Stibor, Vibrational spectra of 2-furaldehyde, J.
Mol. Spectrosc. 47 (1973) 252–267.
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.jcat.2016.12.005.
[29] P. Mirone, P. Chiorboli, Infrared and Raman spectra and vibrational assignment
of maleic anhydride, Spectrochim. Acta 18 (1962) 1425–1432.
[
30] Z.Y. Xue, G.L. Schrader, Transient FTIR studies of the reaction pathway for n-
butane selective oxidiation over vanadyl pyrophosphate, J. Catal. 184 (1999)
References
8
7–104.
[
[
[
[
[
[
[
[
1] H.H.K. Lohbeck, W. Fuhrmann, N. Fedtke, Maleic and fumaric acids, in:
Ullmann’s Encyclopedia of Industrial Chemistry, Weinheim, Germany, 2000,
pp. 463–473.
2] T.R. Felthouse, J.C. Burnett, B. Horrell, M.J. Mummey, Yeong-Jen Kuo, Maleic
anhydride, maleic acid and fumaric acid, in: Kirk-Othmer Encyclopedia of
Chemical Technology Online, 2001.
[31] K.D. Dobson, A.J. McQuillan, In situ infrared spectroscopic analysis of the
adsorption of aliphatic carboxylic acids to TiO , ZrO , Al O , and Ta O from
2
2
2
3
2 5
aqueous solutions, Spectrochim. Acta Part A 55 (1999) 1395–1405.
[32] A.R. Katritzky, J.M. Lagowski, Infrared absorption of heteroaromatic, five-
membered, monocyclic nuclei. Part I. 2-monosubstituted furans, J. Chem. Soc.
(1959) 657–660.
[33] R. Grigg, J.A. Knight, M.V. Sargent, 1131. Studies in furan chemistry. Part I. The
infrared spectra of 2,5-disubstituted furans, J. Chem. Soc. (1965) 6057–6060.
[34] G. Busca, G. Ramis, V. Lorenzelli, On the mechanism of the selective oxidation
of C4 linear hydrocarbons to maleic anhydride: An FT-IR study of the
adsorption and oxidation of 1,3-butadiene on vanadia-titania, J. Mol. Catal.
55 (1989) 1–11.
[35] L. Ballester, C. Carrió, J.F. Bertran, Fermi resonance in rotational isomers of
furfural, Spectrochim. Acta Part A 28 (1972) 2103–2112.
[36] S.J. Hug, D. Bahnemann, Infrared spectra of oxalate, malonate and succinate
adsorbed on the aqueous surface of rutile, anatase and lepidocrocite measured
with in situ ATR-FTIR, J. Electron Spectrosc. Relat. Phenom. 150 (2006) 208–
219.
3] D.R. Kreile, V.A. Slavinskaya, M.V. Shimanskaya, E.Y. Lukevits, The reactivity of
furan compounds in vapor-phase catalytic oxidation, Chem. Heterocyl. Compd.
5
(1972) 429–430.
4] V.A. Slavinskaya, D.R. Kreile, É.E. Dzilyuma, D.É. Sile, Incomplete catalytic
oxidation of furan compounds (review), Chem. Heterocyl. Compd. 13 (1978)
7
10–721.
5] N. Alonso-Fagúndez, M. López Granados, R. Mariscal, M. Ojeda, Selective
conversion of furfural to maleic anhydride and furan with VO
ChemSusChem 5 (2012) 1984–1990.
x 2 3
/Al O catalysts,
6] H. Guo, G. Yin, Catalytic aerobic oxidation of renewable furfural with
phosphomolybdic acid catalyst: an alternative route to maleic acid, J. Phys.
Chem. C 115 (2011) 17516–17522.
7] S. Shi, H. Guo, G. Yin, Synthesis of maleic acid from renewable resources:
Catalytic oxidation of furfural in liquid media with dioxygen, Catal. Commun.
[37] P. Taheri, T. Hauffman, J.M.C. Mol, J.R. Flores, F. Hannour, J.H.W. de Wit, H.
Terryn, Molecular interactions of electroadsorbed carboxylic acid and succinic
anhydride monomers on zinc surfaces, J. Phys. Chem. C 115 (2011) 17054–
17067.
1
2 (2011) 731–733.
8] N. Alonso-Fagúndez, V. Laserna, A.C. Alba-Rubio, M. Mengibar, A. Heras, R.
Mariscal, M. López, Granados, poly-(styrene sulphonic acid): an acid catalyst