Regeneration of silica-supported silicotungstic acid as a catalyst for the dehydration of glycerol

Article abstract of DOI:10.1002/cssc.201100635

The dehydration reaction of glycerol to acrolein is catalyzed by acid catalysts. These catalysts tend to suffer from the formation of carbonaceous species on their surface (coking), which leads to substantial degradation of their performances (deactivation). To regenerate the as-deactivated catalysts, various techniques have been proposed so far, such as the co-feeding of oxygen, continuous regeneration by using a moving catalytic bed, or alternating between reaction and regeneration. Herein, we study the regeneration of supported heteropolyacid catalysts. We show that the support has a strong impact on the thermal stability of the active phase. In particular, zirconia has been found to stabilize silicotungstic acid, thus enabling the nondestructive regeneration of the catalyst. Furthermore, the addition of steam to the regeneration feed has a positive impact by hindering the degradation reaction by equilibrium displacement. The catalysts are further used in a periodic reaction/regeneration process, whereby the possibility of maintaining long-term catalytic performances is evidenced. Copyright

Full text of DOI:10.1002/cssc.201100635

DOI: 10.1002/cssc.201100635
Regeneration of Silica-Supported Silicotungstic Acid as
a Catalyst for the Dehydration of Glycerol
[
a]
[a]
[b]
[c]
Benjamin Katryniok,* Sꢀbastien Paul, Mickaꢁl Capron, Virginie Belliꢂre-Baca,
[d]
[b, e]
Patrick Rey, and Franck Dumeignil
The dehydration reaction of glycerol to acrolein is catalyzed by
acid catalysts. These catalysts tend to suffer from the formation
of carbonaceous species on their surface (coking), which leads
to substantial degradation of their performances (deactivation).
To regenerate the as-deactivated catalysts, various techniques
have been proposed so far, such as the co-feeding of oxygen,
continuous regeneration by using a moving catalytic bed, or
alternating between reaction and regeneration. Herein, we
study the regeneration of supported heteropolyacid catalysts.
We show that the support has a strong impact on the thermal
stability of the active phase. In particular, zirconia has been
found to stabilize silicotungstic acid, thus enabling the nondes-
tructive regeneration of the catalyst. Furthermore, the addition
of steam to the regeneration feed has a positive impact by
hindering the degradation reaction by equilibrium displace-
ment. The catalysts are further used in a periodic reaction/re-
generation process, whereby the possibility of maintaining
long-term catalytic performances is evidenced.
Introduction
Limited petrol resources have inspired researchers to propose
biomass-based substitutes for gasoline, diesel, and kero-
acrolein in the product flow, leading to the formation of unde-
sired highly oxidized molecules, such as acrylic acid, acetic
acid, or even carbon oxides, whereby the yield in acrolein gen-
erally decreases. Furthermore, the oxygen concentration in the
flow must not surpass 7% to stay out of the explosive range
of compositions. This implies that most of the reaction mixture
consists of inert gas (nitrogen or steam), which decreases the
overall productivity of the process.
[
1–3]
sene.
Although the environmental impact of biofuels in
terms of carbon dioxide emissions is rather controversial, many
nations have decided to promote the blending of commercial
fuels with the aforementioned biomass-based compounds,
[
4]
such as biodiesel. As a consequence, the production of the
latter has significantly increased—especially in Europe. Biodie-
sel is produced by the transesterification of vegetable oils,
which involves the production of large quantities of crude
glycerol. Various valorization reactions have been described for
this highly functionalized molecule (triol). Among them, the
dehydration reaction to yield acrolein has attracted consider-
able attention because this compound is an intermediate
widely used for the production of amino acids and superab-
The use of a circulating catalyst bed was first studied by
Corma et al. in a fluid catalytic cracking (FCC)-type reactor with
zeolite-based catalysts, whereby a constant and high yield of
[18]
acrolein (60%) was obtained.
Furthermore, it was shown
that the process could proceed autothermally because the
heat released by coke combustion was recovered to evaporate
[
5,6]
sorbent polymers.
The dehydration reaction of glycerol is
[a] Dr. B. Katryniok, Prof. S. Paul
[
7,8]
catalyzed by acid catalysts such as zeolites,
supported inor-
Nevertheless, all of
Univ. Lille Nord de France, Ecole Centrale de Lille
Unitꢀ de Catalyse et de Chimie du Solide (UMR8181)
Citꢀ Scientifique, F-59655 Villeneuve d’Ascq (France)
Fax: (+33)320-33-54-99
[
9–11]
[12,13]
ganic acids,
or mixed metal oxides.
these catalytic systems suffer from a common drawback pre-
venting stakeholders from envisioning any robust commercial
applications because they are always subject to unavoidable
E-mail: benjamin.katryniok@ec-lille.fr
[b] Dr. M. Capron, Prof. F. Dumeignil
[
10,11,14]
deactivation by coke deposition.
As a consequence, and
Univ. Lille Nord de France, Universitꢀ Lille 1
Unitꢀ de Catalyse et de Chimie du Solide (UMR8181)
Citꢀ Scientifique, F-59655 Villeneuve d’Ascq (France)
to compensate for the low stability of the catalytic systems,
several processes have been proposed, based on the concept
of a continuous regeneration, such as the co-feeding of
[
c] Dr. V. Belliꢁre-Baca
RHODIA France
52 rue de la Haie Coq, 93308 Aubervilliers (France)
[
13,15,16]
oxygen,
tor and regenerator,
ing the reactant and regeneration feeds.
the use of a circulating bed with separated reac-
[
17,18]
or periodic regeneration by alternat-
[d] Dr. P. Rey
ADISSEO France SAS
[
19,20]
Antony Parc 2-10, 92160 Antony (France)
The first concept, based on co-feeding of oxygen, was initial-
ly proposed by Dubois et al. for zeolite catalysts loaded in
a fixed-bed reactor and was further adapted for other types of
[
e] Prof. F. Dumeignil
Institut Universitaire de France
1
0, Boulevard Saint-Michel, 75005 Paris (France)
[
15]
catalysts. Nevertheless, the main disadvantage of this type
of process lies in the concomitant presence of oxygen and
Supporting Information for this article is available on the WWW under
http://dx.doi.org/10.1002/cssc.201100635.
ChemSusChem 0000, 00, 1 – 10
ꢃ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
1
&
ÞÞ
These are not the final page numbers!

B. Katryniok et al.
the glycerol solution. Nevertheless, the important investment
costs of this method are surely one of the main drawbacks to
be overcome before being able to scale this process up. Note
that there are other drawbacks, such as the mechanical stress
on the catalyst or the quantity of catalysts to be loaded into
such a reactor.
generation of this type of catalyst using an alternating feed
process and proved that combining the increased robustness
of our novel catalytic system with a well-designed regenera-
tion technology gives a powerful integrated solution for the
sustainable production of acrolein by glycerol dehydration.
Finally, the regeneration by alternating the reactant and re-
generation feeds was used by Arita et al. for zeolite-type cata-
Results and Discussion
[
19]
lysts loaded in a fixed-bed reactor. They proved that the ini-
tial catalytic performance was recovered, even if regeneration
resulted in hot spots due to the exothermicity of coke com-
bustion. Comparable results were also reported by Strohm
Characterization of fresh and spent catalysts
Nitrogen physisorption experiments
[
20]
et al.
Due to the straightforwardness and low investment
The catalysts containing 20 wt% STA on bare and zirconia-
grafted SBA-15 were analyzed by nitrogen physisorption to de-
termine their specific surface areas and their pore volume/size,
and thereby, to assess the accessibility of the channels after
impregnation and grafting. From the results given in Table 1,
costs needed to set this process up, it seems more favorable
than co-feeding or circulating the catalyst bed, as long as the
catalyst employed can resist the high regeneration tempera-
ture. The process can notably be performed by using two par-
allel reactors, whereby one works under reaction conditions
and the other one under regeneration conditions (spare reac-
tor); a technique already used for the catalytic reforming of
Table 1. Textural properties of supports and catalysts used.
[
21]
naphtha (Figure 1).
[
a]
2
ꢀ1
[c]
[d]
Sample
S
BET [m g
]
V
p
D
p
[
b]
2
ꢀ1
3
ꢀ1
(
S
theo. [m g ])
[cm g
]
[nm]
bare SBA-15
STA on SBA-15
541
1.20
1.05
1.15
0.68
8
8
8
7
431 (433)
465 (433)
378 (372)
ZrO
STA on ZrO
2
on SBA-15
/SBA-15
2
[
a] SBET =BET specific surface area. [b] Stheo. =theoretical specific surface
area in reference to the initial support. [c] V =porous volume. [d] D
average pore diameter obtained by the BJH model.
p
p
=
one can see that the impregnation of SBA-15 leads to a de-
Figure 1. Process with a spare reactor similar to that used for the catalytic
cracking of naphtha. (Adapted from Reference [20].)
2
ꢀ1
crease in the specific surface area (431 vs. 541 m g ) and pore
3
ꢀ1
volume (1.05 vs. 1.20 cm g ), whereas the average pore diam-
eter (8 nm) remains unchanged. The decrease in the specific
surface area is in the range of 20%, which suggests that this
decrease is only based on the increase in weight due to the in-
troduction of the active phase. Due to the nonporous charac-
ter of STA, the active phase introduced does not add any addi-
tional surface to the catalyst, whereby the specific surface area
decreases proportionally to the amount of deposited active
phase.
Among the aforementioned three classes of catalysts—zeo-
lites, mixed metal oxides and supported inorganic acids—the
latter is widely studied, especially due to the facile control of
the acidity and textural properties, which were both identified
as the most important parameters to enable high catalytic per-
[
11,22]
formances in the dehydration of glycerol.
Whereas the tex-
tural properties are mainly a function of the catalyst support,
the acid strength and the number of acid sites can directly be
influenced through the quantity (density) and type of the
active phase. Hereby, in particular, Keggin-type heteropolyacids
Conversely, when SBA-15 is grafted with 20 wt% of zirconia,
2
ꢀ1
the decrease in specific surface area (465 vs. 541 m g ) and
3
ꢀ1
pore volume (1.15 vs. 1.20 cm g ) is less pronounced than in
the aforementioned case of impregnation with the same
amount of STA. This result can easily be explained by the
porous nature of the grafted zirconia, whereby additional sur-
face and porosity are generated on the SBA-15 host support.
When the zirconia-grafted silica is further impregnated with
20 wt% of STA, the specific surface area decreases again by
(
HPAs) were used because their acid strength is easily tunable
[
23,24]
by the choice of central and addenda atoms.
However,
when considering the regeneration of these compounds by
means of oxidative coke combustion, they largely suffer from
their low thermal stability resulting in irreversible decom-
[
24,25]
2
ꢀ1
position.
roughly 20% (378 vs. 465 m g ). Furthermore, we note a de-
3
ꢀ1
We have recently demonstrated that the long-term catalytic
performance of silicotungstic acid (STA) could already be sub-
stantially increased when supported on zirconia-grafted meso-
crease in the pore volume (0.68 vs. 1.15 cm g ) and also in
the pore diameter (7 vs. 8 nm). Nevertheless, the decrease in
the latter is negligible and one can assume that the porous
network remains essentially fully accessible to the reactant.
[
26]
porous silica. Herein, we have focused on the oxidative re-
&2
&
www.chemsuschem.org
ꢃ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 10
ÝÝ
These are not the final page numbers!

Silica-Supported Silicotungstic Acid
Acidity of fresh catalysts
The aforementioned catalysts and their corresponding sup-
ports were further analyzed by performing temperature-pro-
grammed desorption of NH3 to determine the amount and
strength of the acid sites. From the results gathered in Table 2,
one can see that the bare SBA-15 support exhibits nearly no
ꢀ
). When the support is
Cat.
1
acid sites (0.004 mmol NH uptakeg
Scheme 1. Electronic interaction between a Keggin-type heteropolyanion
and metal oxide support.
3
Thermogravimetric analysis of fresh catalyst
[
a]
Table 2. Amount of acid sites of supports and catalysts used.
The thermal stability of STA supported on bare SBA-15 and zir-
conia-grafted SBA-15 was determined by means of thermogra-
vimetric analysis (TGA) under air. From the curves shown in
Figure 2, one can see that both samples exhibit a characteristic
3
uptake [mmolg
weak
Cat.ꢀ
1
] (relative amount [%])
Sample
NH
total
medium
strong
bare SBA-15
0.004 0.0005 (10)
0.178 0.068 (38)
0.339 0.018 (5)
0.442 0.086 (19)
0.0035 (90)
0.051 (29)
0.214 (63)
0.199 (45)
0.0000 (0)
0.059 (33)
0.107 (32)
0.157 (36)
ZrO
STA on SBA-15
STA on ZrO /SBA-15
2
on SBA-15
2
[a] Desorption temp.: weak 100–3008C; medium 300–4508C; strong
>
4508C.
grafted with zirconia, the quantity of acid sites increases
ꢀ
1
(
0.178 mmol NH uptakeg
) because zirconia is known to
Cat.
3
be a Lewis acid. By deconvolution of the thermograms, one
can see that these sites are rather homogeneously distributed
over the whole range of strengths (38% weak sites, 29%
medium sites and 33% strong sites). Conversely, impregnation
of the supports with STA leads to a significant increase in the
number of acid sites because the HPA is a well-known Brøn-
sted acid. Hereby, the catalyst based on SBA-15 exhibits NH
3
ꢀ
1
uptake of 0.339 mmolg
and the ZrO /SBA-15-based one of
2
Cat.
ꢀ
1
Figure 2. Weight loss of STA supported on a) SBA-15 and b) zirconia-grafted
0
.442 mmolgCat. The stronger increase of the latter is attri-
.
SBA-15.
buted to only partial coverage of zirconia, whereby some
Lewis acid sites of the zirconia contribute to the total acidity of
the catalyst. When comparing the distribution of the sites, one
can see that the catalyst based on bare SBA-15 exhibits a ma-
loss of crystalline water between 100 and 1508C. When STA
was supported on bare SBA-15, a second loss of 1.1%—with
reference to the initial weight—was observed starting from
3008C. Such behavior can be ascribed to the loss of constitu-
tional water, according to Scheme 2. However, STA supported
on zirconia-grafted silica exhibits no loss at high temperature,
implying that HPA remained intact. This stabilization of STA
when supported on zirconia-modified silica was ascribed to
electronic interactions between HPA and the support
(Scheme 1). As initially stated for the acidity, the electronic in-
teractions between STA and zirconia were stronger than those
with silica, whereby the thermal stability increased because
cleavage of constitutional water necessitated breaking of these
ꢀ
, cor-
Cat.
1
jority of medium acid sites (0.214 mmol NH uptakeg
3
responding to 63%). Conversely, the catalyst based on zirco-
nia-grafted SBA-15 shows a decreased number of medium acid
ꢀ
, corresponding to 45%),
Cat.
1
sites (0.199 mmol NH uptakeg
3
thus giving rise to an increased number of weak sites
ꢀ
1
(
0.086 mmol NH3 uptakegCat. , corresponding to 19%). This
shift in the acid strength is explained by electronic interactions
between the support and the HPA (Scheme 1). During impreg-
nation with STA, the OH groups at the surface of the support
are protonated, leading to a positively charged surface. The
addition of HPA then results in an electronic interaction,
whereby the strength of the latter depends on the character
[
14,26]
[28]
of the metal oxide support.
Alsalme et al. recently stated
electronic bonds.
that metal oxide supports (e.g.,
niobia, zirconia, and titania) sig-
nificantly
decreased
acid
strength of supported HPAs with
[
27,28]
regard to a silica support.
Scheme 2. Sequence of reactions leading to the thermal decomposition of STA.
ChemSusChem 0000, 00, 1 – 10
ꢃ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemsuschem.org
&
3
&
ÞÞ
These are not the final page numbers!

B. Katryniok et al.
Thermogravimetric analysis of spent catalyst
and the liquid phase was analyzed. The resulting chromato-
grams exhibited a large number of small peaks, but only one
major signal. From the resulting mass spectrum (Figure 4), we
attempted to identify the product using the NIST spectral data-
base. The proposed molecules (Figure 4) generally exhibit the
presence of aromatic cycles and oxygen-containing functional
groups (hydroxyl, acetate, methoxyl, and carboxyl groups).
Nevertheless, none of the proposed molecules were plausibly
explained with regard to established reaction schemes, al-
though the formation of aromatic molecules from acetone and
The spent catalyst containing STA on bare silica was analyzed
by means of TGA under air after 25 h of reaction to determine
the optimum temperature for oxidative regeneration. From the
weight profile, one can see that the catalyst loses significant
mass starting from 3008C (Figure 3). This loss cannot only be
explained by the cleavage of constitutional water because the
[18]
allylic alcohol was already postulated by Corma et al. Fur-
thermore, as the technique was limited to THF-soluble carbo-
naceous species, some types of coke, for example, graphite-
like carbon, remained unidentified; a fact that was even visibly
detected because the THF-washed catalyst remained dark.
Following GC–MS, both types of spent catalysts (ZrO -graft-
2
1
3
ed and ZrO -free catalyst) were analyzed by means of C NMR
2
spectroscopy. The spectra exhibit three massive bands at d=
3
0, 65, and 130 ppm. Similar spectra were reported in the liter-
[29,30]
ature for spent zeolite catalysts used in catalytic cracking.
The corresponding bands were ascribed to aliphatic (d=20–
0 ppm) and aromatic molecules (d=100–160 ppm). Hereby,
8
the aliphatic signal was more pronounced in the region of
d=50–90 ppm, which indicated the presence of heteroatomic
carbon supposedly originating from oxygenated species
Figure 3. a) TGA and b) differential scanning calorimetry of spent catalyst
(STA on bare SBA-15).
relative change in mass is quan-
tified at 18.0% (vs. 1.1% initially
observed for the fresh catalyst).
One can thereby ascribe this de-
crease to the decomposition of
the deposited carbonaceous
species issued from the dehydra-
tion reaction. This is further
confirmed by the strong exo-
thermic peak at 4208C observed
by means of differential scan-
ning calorimetry (Figure 3). As
a second intermediate conclu-
sion, one can state that the opti-
mum temperature for the oxida-
tive regeneration of the catalyst
is higher than 3008C. This also
implies that oxidative regenera-
Figure 4. Mass spectrum of the major peak obtained in GC–MS analysis of the spent catalyst.
tion will inevitably result in the
loss of constitutional water from
STA when supported on bare SBA-15, which is considerably
less the case when supported on zirconia-grafted SBA-15 due
to its aforementioned increased thermal stability (cf. Figure 3).
(C ꢀO). Furthermore, the small band at around d=20–30 ppm
ali
was identified as ꢀCH and ꢀCH groups. Altogether, this was
2
3
in agreement with the proposed molecules from GC–MS analy-
sis (Figure 5).
Characterization of the coke deposit
FTIR spectroscopy
The type of coke deposit on the spent catalyst was further de-
13
termined by using C NMR spectroscopy and GC–MS measure-
ments. For the latter technique, pure THF was added to the
spent catalyst (ZrO -grafted and ZrO -free supported catalysts),
The presence of STA before and after catalytic tests was further
studied by means of IR spectroscopy. From the spectra
(Figure 6), one can see that all catalysts exhibit characteristic
2
2
&4
&
www.chemsuschem.org
ꢃ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 10
ÝÝ
These are not the final page numbers!

Silica-Supported Silicotungstic Acid
Catalytic tests
Initial catalytic performance and regeneration in dry and wet
air
The initial catalytic performance and deactivation of STA on
SBA-15 and STA on ZrO -grafted SBA-15 catalysts were studied
2
during 25 h of reaction. From Table 3, one can see that both
catalysts give 71% yield of acrolein during the first 5 h of reac-
tion. After 25 h of reaction, the catalytic activity of both cata-
lysts is significantly decreased, which is attributed to the depo-
sition of coke on the surface of the catalyst (cf. TGA of spent
catalyst in Figure 3). The conversion of STA on bare silica is
more than halved (41% vs. 84% initially), whereas the zirconia-
containing catalyst has lost only one fifth of its initial activity
1
3
Figure 5. C NMR spectra of spent catalyst based on a) bare silica and
b) ZrO -grafted silica.
(
78 vs. 96% initially). Comparable results were also reported by
2
[32]
Kim et al. using the same type of catalyst on zirconia. This
phenomenon is ascribed to the reduced acid strength of STA
when using strong Lewis acids as supports (cf. acidity of fresh
[26]
catalysts, Scheme 1).
The reduced acid strength results in
less carbon deposit, and therefore, increased long-term catalyt-
ic performance. In the next step, both catalysts were regener-
ated in situ, meaning that the catalysts were left inside the re-
actor after the reaction to directly perform the regeneration
process. Therefore, the inlet feed was replaced with air
ꢀ
1
(
20 mLmin ), and the reactor temperature was kept at 2758C
to enable regeneration while avoiding the decomposition of
STA (cf. TGA of fresh catalyst in Figure 2). After 6 h of regenera-
tion under air, the catalytic performance was determined
again. From Table 3, one can see that only the catalyst contain-
ing zirconia nearly recovered its initial performance, exhibiting
89% conversion (vs. 96% initially) and 78% of yield in acrolein
(
vs. 71% initially). The catalyst based on bare silica also
Figure 6. FTIR spectra for bulk STA (a), STA on fresh (c) and spent (b) SBA-15,
STA on fresh (e) and spent (d) ZrO -grafted SBA-15.
showed a higher catalytic activity than before regeneration
with 60% conversion versus 41% after 25 h, but its initial activ-
ity (84% of conversion) was not recovered in this case. Further-
more, the yield of acrolein after the regeneration cycle was
2
absorption bands of the STA
Keggin unit at n˜ =981 and
ꢀ
1
[a]
928 cm , which were attribut-
Table 3. Catalytic performance tests (blank test results are given in Table S1 in the Supporting Information).
ed to the symmetric stretching
of the addenda atom–terminal
oxygen bond (W=O) and the
asymmetric stretching of the
[
d]
Reaction time
–5 h (0–1 h)
Regeneration
Periodic regeneration
24–25 h
[b]
[c]
0
24–25 h
dry
wet
96–97 h
2
C
0 wt% STA on SBA-15
[e]
[%]
84 (86)
83 (79)
5 (5)
71 (68)
91 (86)
41
57
5
24
85
60
50
4
30
72
64
66
7
42
83
90
84
3
76
88
87
85
5
74
91
[
f]
central
atom–oxygen
bond
SAC [%]
[
31]
[g]
(SiꢀO), respectively.
Because
SHA [%]
[
h]
Y
[%]
i]
these signals were even detect-
ed for the spent catalysts (after
[
CB [%]
[
j]
2
5 h of reaction), one can sup-
2
20 wt% STA on ZrO /SBA-15
[
e]
pose that the active phase re-
mains essentially intact during
the reaction. Nevertheless, par-
tial decomposition cannot be
excluded because the amounts
of the corresponding products
would be low, and therefore,
remain below the limit of detec-
tion.
C
[%]
96 (98)
74 (38)
12 (17)
71 (37)
96 (56)
78
88
12
69
100
89
88
11
78
99
86
89
11
77
100
78
45
12
35
67
76
46
16
35
71
[
f]
SAC [%]
[
g]
S
Y
HA [%]
[h]
[%]
i]
[
CB [%]
ꢀ
1
[
a] Operating conditions: 2758C reaction temperature; 0.3 g catalyst, 220 h gas hourly space velocity of glyc-
erol (gly); STA=silicotungstic acid (H 40). [b] 6 h in dry air at 2758C. [c] 6 h in wet air at 2758C; molar
ratio N /O /H O=0.78:0.17:0.05. [d] Isochronical reaction and regeneration cycle of 10 min. [e] C=conversion
of glycerol. [f] SAC =selectivity to acrolein. [g] SHA =selectivity to acetol. [h] Y=yield in acrolein. [i] CB=carbon
balance. [j] ZrO amount 20 wt%.
4
SiW12O
2
2
2
2
ChemSusChem 0000, 00, 1 – 10
ꢃ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemsuschem.org
&
5
&
ÞÞ
These are not the final page numbers!

B. Katryniok et al.
only slightly higher than before (30 vs. 24%), which implies
that the catalyst lost some selectivity to acrolein. This result is
ascribed to thermal decomposition of STA in hot spots result-
ing from exothermal oxidative combustion of the deposited
carbon. As the thermal stability of STA decreased when sup-
ported on bare silica, this phenomenon was only observed for
the zirconia-free catalyst. Nevertheless, because no decomposi-
tion of the active phase in the spent catalyst was detected by
means of FTIR spectroscopy (cf. Figure 6), we decided to
modify the regeneration conditions to hinder the thermal de-
composition reaction of the catalyst (Scheme 1) by introducing
steam into the regeneration feed (equilibrium displacing).
Therefore, the two fresh catalysts were again used in the dehy-
dration reaction for 25 h. The regeneration was then per-
formed under steam-enriched air by replacing the glycerol
feed with water (molar ratio N /O /H O=0.78:0.17:0.05),
after 96 h. From the results reported in Table 3, one can see
that the catalyst based on zirconia-free silica shows a stable
performance with an acrolein yield in the range of 74–76% at
a glycerol conversion of 87–90%. At a first glance, these results
seem to be in contradiction with the former ones obtained
after regeneration in dry air, for which thermal decomposition
was observed. Nevertheless, one has to bear in mind that, in
contrast to the former series of tests, the catalyst was only re-
acted for 5 h (vs. 24 h previously) before starting regeneration.
Thereby, the rather limited quantity of deposited carbon re-
sults in fewer hot spots, which explains the decreased thermal
decomposition of this catalyst.
The catalyst based on zirconia-grafted silica exhibits rather
low catalytic performance with no more than 35% yield of
acrolein and a glycerol conversion in the range of 76–78%.
Once again, this result seems to be in contradiction with the
former performance of this catalyst after regeneration. This be-
havior was ascribed to the important activation period of the
catalyst. Whereas the catalyst based on nongrafted silica
showed initially high performance even during the first hour
under stream (Table 3; 68% yield in acrolein), the catalyst
based on zirconia-grafted silica showed rather low perfor-
mance during start up, which was owied to decreased selectiv-
ity to acrolein (38%). Comparable observations were also re-
ported by Chai et al. for zirconia-supported phosphotungstic
2
2
2
whereas the reactor temperature remained unchanged. Similar
to the case for dry air, the catalyst based on zirconia-grafted
SBA-15 recovered its initial performances. Conversely, the zirco-
nia-free catalyst again did not recover its initial catalytic activity
(64% conversion vs. 84% initially). Nevertheless, one can state
in this case that the yield of acrolein is significantly higher
when regeneration is performed in steam-enriched air (42 vs.
24% in a dry atmosphere). This result suggests that, as expect-
ed, the thermal decomposition of STA was at least partially in-
hibited, which may be explained by 1) the increased heat ca-
pacity of steam, leading to lower hot-spots, and 2) the shift of
the equilibrium of the decomposition reaction (Scheme 2) to
the left side due to the presence of water, as proposed by sev-
[10,12,35]
acid and also for niobium oxide.
More recently, Lauriol-
Garbay et al. stated an important activation period when using
[36]
zirconium–niobium mixed oxides. Even though there is still
work required to shed light on a mechanistic explanation for
this effect, one can presume that the presence of Lewis acid
sites from zirconia is responsible for this behavior. Alhanash
et al. proposed an activation mechanism for Lewis acid sites
that explained the increased selectivity towards acetol general-
[
33,34]
eral researchers.
As far as STA on ZrO -grafted SBA-15 catalyst is concerned,
2
one can see that there is no effect at all from the presence of
water in the regenerative atmosphere, which is consistent with
the aforementioned explanation. Indeed, as shown by TGA (cf.
Figure 3), STA exhibits increased thermal stability when sup-
ported on zirconia-grafted SBA-15, whereby no decomposition
takes place under the conditions applied during the regenera-
tion process. Accordingly, the addition of water during regen-
eration has no impact at all.
[9]
ly observed for Lewis acid catalysts (Scheme 3). It is then this
pronounced activation effect that explains the low catalytic
performance of the zirconia-containing catalyst when using
short reaction cycles (10 min) in contrast to the previous re-
ported results obtained when using a long reaction cycle of
24 h.
Periodic regeneration experiments
Conclusions
The experiments using periodic regeneration of the catalysts
are based on the aforementioned idea of using a spare reactor,
where the spent catalyst is regenerated in parallel to the run-
ning reactor. Therefore, the catalysts were left under stream for
The use of zirconia-grafted silica as a support for STA signifi-
cantly increases the thermal stability of the active phase (re-
duction of the constitution water loss), whereby the regenera-
tion of the spent catalyst by means of oxidative combustion of
coke was made possible. The initial catalytic performance was
almost fully recovered, implying that no thermal destruction of
the HPA compound took place. Conversely, when using bare
silica as a support for STA, the latter exhibits rather low ther-
mal stability. Thereby, the regeneration of the spent catalyst
did not result in complete recovery of the initial catalytic per-
formance. Nevertheless, in this case, the presence of water in
the regeneration flow showed a positive impact on the catalyt-
ic performance of the regenerated catalyst. This is ascribed to
partial inhibition of the thermal decomposition of the HPA by
equilibrium displacement. Furthermore, the catalysts were
5
h before starting to alternate the reactant and regeneration
flows with an interval time of 10 min for each flow. As seen
from the previous experiments, the presence of steam during
regeneration had no effect on the subsequent catalytic perfor-
mance of the zirconia-containing catalyst, and even the posi-
tive impact in the case of the zirconia-free catalyst was rather
low. As the latter did not recover its initial performance even
under a steam-rich atmosphere, the experiments of periodic
regeneration were performed with a regeneration feed consist-
ing of dry air. The performance of the catalysts during the peri-
odic regeneration cycles was determined after 24 h and again
&6
&
www.chemsuschem.org
ꢃ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 10
ÝÝ
These are not the final page numbers!

Silica-Supported Silicotungstic Acid
[
9]
Scheme 3. Mechanistic explanation for the activation period observed for Lewis acid catalysts.
tested under alternating feed cycles of glycerol and air. Due to
the short cycle time of 10 min, the catalyst based on zirconia-
free silica exhibited a higher performance; a fact explained by
the absence of an activation period in this case.
drous, Riedel). Zirconium(IV) n-propoxide (0.76 g, 70 wt% solution
in n-propanol; Aldrich) was added to the mixture, corresponding
to an equivalent of 0.2 g of ZrO after hydrolysis. The slurry was
2
left under stirring for another 8 h to hydrolyze the organic zirconia
precursor. Then, the solid was filtered off, washed with dry ethanol,
One can, therefore, divide the catalysts into two classes:
and dried in air at 808C. Calcination was performed in static air at
1) Brønsted acid catalysts with initially high selectivity to acro-
ꢀ1
6
508C for 3 h (heating ramp of 1 Kmin ) to remove the remaining
lein, but suffering from a short catalyst lifetime; and 2) Lewis
acid catalysts with a significant activation period, but increased
long-term performance. With regard to these results, the re-
generation processes are significantly different for each class
of catalyst. Due to the short activation period, Brønsted acid
catalysts can be employed in fluidized catalytic bed reactors,
similar to the fluidized catalytic cracking process, with a short
contact time in the reaction zone. Conversely, the optimal re-
generation process for Lewis acid catalysts is similar to
a moving bed process, such as that used in catalytic reforming.
Hereby, the residence time of the catalyst in the reaction sec-
tion is in the range of hours or days, whereby the activation
period of the catalyst becomes negligible.
organic compounds. The corresponding support exhibited an ex-
perimental zirconia content of 18.1 wt% versus 20 wt% calculated
theoretically, which is explained by the washing of the support
prior to drying and calcination.
Impregnation with STA: The impregnation with STA was performed
as follows. STA (0.2 g; Sigma) was added to a slurry of support
(0.8 g) in water (20 mL). The mixture was subsequently stirred for
another 2 h before the water was evaporated under vacuum. The
obtained catalyst was dried in air for 24 h at 708C before use. The
final catalyst particle size was <100 mesh.
Characterization techniques
Nitrogen adsorption/desorption isotherms were obtained at
ꢀ
1908C by using a Micromeritics ASAP 2010 analyzer. The specific
surface area (SBET) was determined by using the Brunauer–Emmett–
Teller (BET) method, and the pore size distribution was calculated
according to the Barret–Joyner–Halenda (BJH) formula. The total
Experimental Section
Catalyst preparation
pore volume (V ) was determined by using the value measured at
p
Two types of catalysts were prepared for this study. Both catalysts
contained 20 wt% silicotungstic acid (STA, H SiW O ) supported
a relative pressure (P/P ) of 0.995. Before measurement, the sam-
ples were outgassed under vacuum at 1408C for 3 h.
0
4
12 40
either on bare SBA-15 or on zirconia-grafted SBA-15 (final composi-
tion: 20 wt% STA, 16 wt% ZrO , 64 wt% SiO )
Temperature programmed desorption of NH (NH -TPD) was used
3
3
2
2
to characterize the acidity in terms of amount and strength of the
acid sites. The experiments were performed by using a Micromerit-
ics AutoChem II 2920 apparatus connected to a mass spectrometer
(Varian). The samples were first degassed at 3008C before being
flushed three times with NH at 1008C. Afterwards, NH desorption
Preparation of SBA-15 mesoporous silica: The SBA-15 mesoporous
silica support was prepared according to the procedure described
[37]
by Roggenbuck et al.
Accordingly, triblock copolymer P123
(
3.25 g; EO PO EO , Pluronic P123; Aldrich) were dissolved in
2
0
70
20
3
3
ꢀ1
a mixture of distilled water (101 mL) and hydrochloric acid (8.7 mL,
7 wt%; Fluka) at 408C. Then, tetraethyl-ortho-silicate (TEOS; 6.5 g,
was performed with a temperature ramp of 10 Kmin up to
5508C and detected by using a thermal conductivity detector
(TCD) and a mass spectrometer.
3
purity ꢁ99%; Aldrich) was added (resulting in a molar ratio of
TEOS/P123/HCl/H O=1.0:0.018:3.3:191). The reaction mixture was
2
The thermal stability of the supported STA and the efficiency of
the oxidative regeneration of the spent catalyst were evaluated by
interpreting thermogravimetric measurements performed on a Mi-
cromeritics AutoChem II 2920 apparatus. The samples (about
20 mg) were initially treated with a helium flow for 1 h at 1008C to
remove physisorbed water. Then, the samples were heated to
stirred for another 24 h at 408C, then transferred to a Teflon-
coated autoclave and heated for hydrothermal treatment over 24 h
at 1408C. After filtration, the silica was washed with distilled water
and dried in air at 808C. Finally, calcination was performed in static
ꢀ1
air at 6508C for 3 h (heating ramp of 1 Kmin ).
ꢀ
1
Zirconia grafting: The preparation of the zirconia-grafted SBA-15
was performed by using zirconium(IV) propoxide as an organic
precursor for zirconia. The experimental protocol is as follows:
SBA-15 support (0.8 g) was slurried in dry ethanol (10 mL; anhy-
5008C under air (5 Kmin ) while recording the weight loss. In the
case of the spent catalyst, the thermogravimetry was coupled with
differential scanning calorimetry to monitor the heat flow from the
oxidative combustion of coke.
ChemSusChem 0000, 00, 1 – 10
ꢃ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemsuschem.org
&
7
&
ÞÞ
These are not the final page numbers!

B. Katryniok et al.
ꢀ
1
The IR spectra were recorded for samples in pressed KBr tablets by
using a spectrometer equipped with a mercury cadmium telluride
controller (Brooks). The airflow (1.2 mLh ) was either introduced
by passing through the evaporator or just behind the latter
through valve 1, which is a two-position, four-way sampling valve
(ViciValco). The first path was used to enrich the regeneration flow
with steam, whereas the second path provided the possibility of
a periodic permutation between air and reactant flow, for which
the valve was actuated by a motor commanded by a PC to control
the time of each cycle.
ꢀ1
(
2
MCT) detector. The nominal spectral resolution was 4 cm and
56 scans (acquisition time 52 s) were accumulated for each spec-
trum. A KBr spectrum was used as a background sample for post-
processing.
The nature of the carbon species deposited on the spent catalyst
13
was determined by performing C NMR spectroscopy using
a Bruker spectrometer at an operating frequency of 100.6 MHz.
Furthermore, the deposited carbon was further analyzed by means
of GC–MS measurements: THF was added to the spent catalyst,
and the suspension was stirred for 1 h before the liquid phase was
analyzed by means of GC–MS using a Varian 3800 gas chromato-
graph associated to a Varian Saturn 2000 mass spectrometer.
The detailed procedure of regeneration was as follows: prior to the
catalytic test, the glycerol feed was stabilized while bypassing the
reactor. After 12 h of stabilization, glycerol was fed to the catalyst
for 25 h to determine the initial performance. Afterwards, the glyc-
erol feed was stopped while maintaining helium flow in the reac-
tor. After 1 h, the reactor was again bypassed, whereby the catalyst
remained under an inert-gas atmosphere (helium). Then, water was
introduced by means of an HPLC pump and diluted in air. The cor-
responding water/air feed was stabilized for 12 h before being fed
to the reactor. After 6 h of regeneration, the HPLC pump was
stopped, and the regeneration feed was replaced by helium,
whereby the catalyst became flushed with inert gas. Then, the re-
actor was again bypassed and glycerol was introduced by means
of the HPLC pump. After 12 h of stabilization of the glycerol feed,
the dehydration reaction was restarted to determine the catalytic
performance after regeneration. Due to the complexity of the pro-
cess, it was not adaptable for pulsed regeneration.
Catalytic reaction
The catalytic performance was determined at atmospheric pressure
for 300 mg of catalyst in a tubular fixed-bed reactor (8.1 mm diam-
eter, 100 mm length) at a reaction temperature of 2758C. The reac-
tion setup is shown in Figure 7. The glycerol solution (10 wt%) was
fed by means of an HPLC pump (Gilson) at a flow rate of
ꢀ1
1
.5 mLh . The solution was evaporated in a homemade evapora-
tor placed in a heated box (2108C). The glycerol vapors were fur-
The content of the traps was analyzed by performing HPLC using
a THERMO SpectraSystem with a Phenomenex Rezex ROA organic
+
acid H column (250 mm length, 7 mm particle size) as a stationary
ꢀ
1
phase and an aqueous solution of sulfuric acid (5 mmolL
0
;
ꢀ1
.5 mLmin ; isocratic) as an eluting agent. For detection, the
system was equipped with a refractometer (THERMO Surveyor Plus
RI). The catalytic performances (conversion of glycerol C , selectivi-
gly
ty S , yield Y ) were calculated by using Equations (1)–(3):
N
N
molgly, reactant feedꢀmolgly, outlet feed
Cgly ð%Þ ¼
ꢂ 100 %
ꢂ 100 %
ð1Þ
mol
gly, reactant feed
molC in prod:, reactant feed
SN ð%Þ ¼ molC in gly, reactant feedꢀmol
ð2Þ
ð3Þ
C in gly, outlet feed
YN ð%Þ ¼ Cgly ð%Þ ꢂ SN ð%Þ
The selectivity was reported for acrolein and acetol. Other identi-
fied products were acetaldehyde, propionaldehyde, acetone, and
allylic alcohol, of which the cumulated selectivity did not exceed
1
%. The carbon balance was determined by taking into account
unconverted glycerol in the outlet and the yields of all identified
products, which were calibrated by injection of standard solutions.
Figure 7. Reaction setup for alternating the glycerol and air feed by a period-
ical switch of valve 1.
Acknowledgements
ꢀ1
ther diluted in a flow of 30 mLmin inert carrier gas (nitrogen)
regulated by a mass flow controller (Brooks). The reaction mixture
was then fed to the reactor, which was located in a separate com-
partment. During the stabilization period of the glycerol feed
We thank Gerard Cambien for the nitrogen physisorption experi-
ments and Olivier Gardoll for NH -TPD and TGA measurements.
3
Furthermore, we thank Adisseo for financial support.
(
12 h), the reaction mixture bypassed the reactor. Downstream of
the reactor, the condensable products were collected for 1 h in
one of the two parallel ice traps (initially filled with 10 g of water),
whereby an uninterrupted sampling of the products was possible
by swapping. The mass balance of the traps was always greater
than 97%.
Keywords: acrolein · dehydration · glycerol · heterogeneous
catalysis · supported catalysts
[1] A. J. Ragauskas, C. K. Williams, B. H. Davison, G. Britovsek, J. Cairney, A.
Eckert, W. J. Frederick, Jr., J. P. Hallet, D. J. Leak, C. H. Liotta, J. R. Mielenz,
R. Murphy, R. Templer, T. Tschaplinski, Science 2006, 311, 484.
For the regeneration experiments, the setup was further equipped
with a second feeding line for air, also regulated by a mass flow
[2] A. Corma, S. Iborra, A. Velty, Chem. Rev. 2007, 107, 2411.
&8
&
www.chemsuschem.org
ꢃ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 10
ÝÝ
These are not the final page numbers!

Silica-Supported Silicotungstic Acid
[
3] J. N. Chheda, G. W. Huber, J. A. Dumesic, Angew. Chem. 2007, 119, 7298;
Angew. Chem. Int. Ed. 2007, 46, 7164.
[23] T. Okuhara, N. Mizuno, M. Misono, Adv. Catal. 1996, 41, 113.
[24] I. V. Kozhevnikov, Catalysts for Fine Chemicals, Vol. 2, 2002, Wiley, Chi-
chester.
[4] European directive 2003/30/EC of May 8th, 2003.
[
5] B. Katryniok, S. Paul, M. Capron, F. Dumeignil, ChemSusChem 2009, 2,
[25] I. V. Kozhevnikov, J. Mol. Catal. A: Chem. 2007, 262, 86.
[26] B. Katryniok, S. Paul, M. Capron, C. Lancelot, P. Rey, V. Belliꢂre-Baca, F.
Dumeignil, Green Chem. 2010, 12, 1922.
[27] A. M. Alsalme, P. V. Wiper, Y. Z. Khimyak, E. F. Kozhevnikova, I. V. Kozhev-
nikov, J. Catal. 2010, 276, 181.
[28] A. Alsalme, E. F. Kozhevnikova, I. V. Kozhevnikov, Appl. Catal. A: Gen.
2010, 390, 219.
[29] J. L. Bonardet, M. C. Barrage, J. Fraissard, J. Mol. Catal. A: Chem. 1995,
96, 123.
7
19.
6] B. Katryniok, S. Paul, P. Rey, V. Belliꢂre-Baca, F. Dumeignil, Green Chem.
010, 12, 2079.
[
[
2
7] A. K. Kinage, P. P. Upare, P. Kasinathan, Y. K. Hwang, J.-S. Chang, Catal.
Commun. 2010, 11, 620.
[
[
8] C. J. Jia, Y. Liu, W. Schmidt, A. H. Lu, F. Schꢄth, J. Catal. 2010, 269, 71.
9] A. Alhanash, E. F. Kozhevnikova, I. V. Kozhevnikov, Appl. Catal. A: Gen.
2010, 378, 11.
[
[
10] S.-H. Chai, H.-P. Wang, Y. Liang, B. Q. Xu, Green Chem. 2008, 10, 1087.
11] E. Tsukuda, S. Sato, R. Takahashi, T. Sodesawa, Catal. Commun. 2007, 8,
[30] P. CastaÇo, G. Elordi, M. Olazar, A. T. Aguayo, A. Pawelec, J. Bilbao, Appl.
Catal. B: Environ. 2011, 104, 91.
1
349.
[31] R. Thouvenot, M. Fournier, R. Franck, C. Rocchiccioli-Deltcheff, Inorg.
Chem. 1984, 23, 598.
[
[
[
[
12] S.-H. Chai, H.-P. Wang, Y. Liang, B. Q. Xu, J. Catal. 2007, 250, 342.
13] F. Wang, J.-L. Dubois, W. Ueda, J. Catal. 2009, 268, 260.
14] H. Atia, U. Armbruster, A. Martin, J. Catal. 2008, 258, 71.
15] J.-L. Dubois, C. Duquenne, W. Hçlderich (Arkema), WO2006087083,
[32] Y. T. Kim, K.-D. Jung, E. D. Park, Bull. Korean Chem. Soc. 2010, 31, 3283.
[33] Y. Konishi, K. Sakata, M. Misono, Y. Yoneda, J. Catal. 1982, 77, 169.
[34] I. V. Kozhevnikov, Handbook of Green Chemistry, Vol. 2, Wiley-VCH, Wein-
heim, 2009.
2006.
[
[
16] A. Ulgen, W. Hçlderich, Catal. Lett. 2009, 131, 122.
17] J.-L. Dubois, C. Duquenne, W. Hçlderich, J. Kervennal (Arkema),
WO2006087084, 2006.
[35] S.-H. Chai, H.-P. Wang, Y. Liang, B. Q. Xu, Appl. Catal. A: Gen. 2009, 353,
213.
[36] P. Lauriol-Garbay, J. M. M. Millet, S. Loridant, V. Belliꢂre-Baca, P. Rey, J.
Catal. 2011, 280, 68.
[
[
[
18] A. Corma, G. W. Huber, L. Sauvanaud, P. O’Connor, J. Catal. 2008, 257,
163.
[37] J. Roggenbuck, G. Koch, M. Tiemann, Chem. Mater. 2006, 18, 4151.
19] Y. Arita, H. Kasuga, M. Kirishiki (Nippon Catalytic Chem. Ind.),
JP2008110298, 2008.
20] J. J. Strohm, A. H. Zacher, J. E. White, M. J. Gray, V. Lebarbier,
US2011075743.
Received: October 11, 2011
[
[
21] R. Galiasso Tailleur, C. Albornoz, Catal. Today 2010, 150, 308.
22] S.-H. Chai, H.-P. Wang, Y. Liang, B. Q. Xu, Green Chem. 2007, 9, 1130.
Revised: November 12, 2011
Published online on && &&, 0000
ChemSusChem 0000, 00, 1 – 10
ꢃ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemsuschem.org
&
9
&
ÞÞ
These are not the final page numbers!

FULL PAPERS
B. Katryniok,* S. Paul, M. Capron,
V. Belliꢁre-Baca, P. Rey, F. Dumeignil
Catalytic support: The regeneration of
supported silicotungstic acid, widely
used in the dehydration of glycerol to
yield acrolein, is described. The nature
of the support has a strong impact on
the thermal stability of the active phase.
Zirconia stabilizes silicotungstic acid,
thus enabling efficient and nondestruc-
tive regeneration (see picture).
&& – &&
Regeneration of Silica-Supported
Silicotungstic Acid as a Catalyst for
the Dehydration of Glycerol
&10
&
www.chemsuschem.org
ꢃ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 0000, 00, 1 – 10
ÝÝ
These are not the final page numbers!