Rubidium- and caesium-doped silicotungstic acid catalysts supported on alumina for the catalytic dehydration of glycerol to acrolein

Article abstract of DOI:10.1016/j.jcat.2011.11.004

Dehydration of glycerol was carried out using rubidium- and caesium-doped silicotungstic acid catalysts. These catalysts were prepared by varying concentration of the dopant metal cations while keeping the concentration of heteropoly acid unchanged. High acrolein selectivity (94-96%) was observed with unsupported caesium-doped silicotungstic acid and rubidium-doped silicotungstic acid with a dilute glycerol feed (0.5 wt.% in water). These catalysts were then supported on alpha-alumina and an alumina comprising a theta-delta mixture. Caesium-doped silicotungstic acid supported on theta-delta alumina gave a maximum selectivity of ca. 90% at 100% glycerol conversion for 90-h time online, with a 10 wt.% glycerol solution. With a more concentrated glycerol feed (20 wt.%), this catalyst achieved a space time yield of 210 g _(acrolein)kg(cat)-1h^-1. The catalyst was investigated further to determine the origin of the long-term stability. The binding strength of the partially doped silicotungstic acid on the alumina was found to be crucial to sustain the supported Keggin structure and hence the acidity of the active sites resulting in a high acrolein yield.

Full text of DOI:10.1016/j.jcat.2011.11.004

Journal of Catalysis 286 (2012) 206–213
Contents lists available at SciVerse ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Rubidium- and caesium-doped silicotungstic acid catalysts supported on
alumina for the catalytic dehydration of glycerol to acrolein
a
a
b
a
a
Muhammad H. Haider , Nicholas F. Dummer , Dazhi Zhang , Peter Miedziak , Thomas E. Davies ,
a
a
a
b
a,
⇑
Stuart H. Taylor , David J. Willock , David W. Knight , David Chadwick , Graham J. Hutchings
a
Cardiff Catalysis Institute, School of Chemistry, Cardiff University, Cardiff CF10 3AT, UK
Department of Chemical Engineering, Imperial College London, South Kensington Campus, London SW7 2AZ, UK
b
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 17 July 2011
Revised 18 October 2011
Accepted 6 November 2011
Available online 15 December 2011
Dehydration of glycerol was carried out using rubidium- and caesium-doped silicotungstic acid catalysts.
These catalysts were prepared by varying concentration of the dopant metal cations while keeping the
concentration of heteropoly acid unchanged. High acrolein selectivity (94–96%) was observed with
unsupported caesium-doped silicotungstic acid and rubidium-doped silicotungstic acid with a dilute
glycerol feed (0.5 wt.% in water). These catalysts were then supported on alpha-alumina and an alumina
comprising a theta-delta mixture. Caesium-doped silicotungstic acid supported on theta-delta alumina
gave a maximum selectivity of ca. 90% at 100% glycerol conversion for 90-h time online, with a
Keywords:
Glycerol dehydration
Acrolein
Silicotungstic acid
Caesium-doped catalysts
Supported heteropoly acids
1
0 wt.% glycerol solution. With a more concentrated glycerol feed (20 wt.%), this catalyst achieved a space
ꢀ1
ꢀ1
time yield of 210 g
kg
h . The catalyst was investigated further to determine the origin of the
ðacroleinÞ
ðcatÞ
long-term stability. The binding strength of the partially doped silicotungstic acid on the alumina was
found to be crucial to sustain the supported Keggin structure and hence the acidity of the active sites
resulting in a high acrolein yield.
Ó 2011 Elsevier Inc. All rights reserved.
1
. Introduction
The conversion of biomass and biomass-derived feed stocks into
products. Typically, in order to dehydrate glycerol, high tempera-
tures along with acidic conditions are required. Acidity plays a vital
role in glycerol dehydration controlling the selectivity to acrolein,
and acidic catalysts having a Hammett acidity (HA) [9] below 2
are considered desirable in this regard [10]. Tungstated zirconia
catalysts have been evaluated for glycerol conversion to acrolein
at a temperature of 300 °C, and they were found to be active
[11]. Tungsten was found to increase the acidity of the catalyst;
however, the stability of such catalysts at extended reaction times
remains a problem. This may be due to the high acidity of the cat-
alysts, which is thought to be responsible for coke formation and
ultimately lack of stability observed [12]. The effect of acidity on
the stability of catalysts for the dehydration of glycerol has been
studied by grafting zirconia onto silica [12]. Comparison of the
grafted and ungrafted catalysts appeared to differ with respect to
stability, with the zirconia-grafted catalyst being more stable. This
effect was attributed to a minor distortion in the structure of the
heteropoly acid (HPA), and its binding strength to the support,
leading to a decrease in acidity and in turn an increase in the sta-
bility of the catalysts [4]. Chai and co-workers [13,14] conducted
studies of glycerol dehydration using different niobium oxide-
based catalysts, which resulted in an acrolein selectivity of 50%
at a glycerol conversion of 90%. Use of basic materials, e.g. MgO,
did not result in acrolein formation, whereas it was observed with
valuable chemicals and fuels is emerging as an important aspect of
modern chemical processes [1]. Presently, transesterification of
plant oils is increasing as this process yields biodiesel and also
glycerol as a by-product [2,3]. Increasingly, utilisation of glycerol
as a versatile starting a material is concurrent with both its
increasing availability and chemical properties. The structure of
glycerol contains three hydroxyl groups, which makes it readily
soluble in water and other alcohols, and it can undergo chemical
and biochemical changes because of its ability to be functionalised
[
4]. The efficient chemical transformation of glycerol to acrolein
forms an appreciable contribution to current efforts to obtain
value-added products derived from biomass.
Acrolein is an important chemical intermediate in the manufac-
ture of polymers, and it is presently produced industrially in the
order of several million tonnes per annum from propene [5–7].
However, its preparation from glycerol is advantageous as this
pathway can be viewed as being based on a renewable feedstock
[
3,8]. When glycerol is subjected to acidic conditions, it quickly
undergoes decomposition with the release of water and other
⇑
Corresponding author.
E-mail address: hutch@cardiff.ac.uk (G.J. Hutchings).
2 3 4 2 3
slightly acidic, e.g. ZrO , and highly acidic, e.g. H PO /Al O ,
0
021-9517/$ - see front matter Ó 2011 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2011.11.004

M.H. Haider et al. / Journal of Catalysis 286 (2012) 206–213
207
catalysts. Notably, acrolein selectivity increased with increasing
acidity from slightly acidic (30%) with Hammet acidity between
included the mixing of aqueous solution (5 mL H
lyst) of heteropoly acid with the support and stirring it for 20 h fol-
2
O per g of cata-
ꢀ
3 and +7 to highly acidic catalysts (70%) with Hammet acidity be-
tween ꢀ8 and ꢀ3, showing that acidity was definitely exerting a
lowed by drying (110 °C, 16 h). The same method was followed to
prepare 30% silicotungstic acid supported on theta-delta Al
These catalysts are referred to as follows: untreated Al
0STA-1, 20STA-1, 30STA-1, 40STA-1, 50STA-1 and 30STA-2.
2
O
3
-2.
role in controlling the acrolein selectivity.
2 3
O -1 as
Supported heteropoly acid catalysts are very acidic, typically
with a Hammet acidity of ca. ꢀ9, and they have been used in the
catalytic dehydration of glycerol [15]. Different loadings of silico-
tungstic acid (STA), supported on silica or carbon, have been inves-
tigated; however, these catalysts were also found to be unstable
2.2.2. Caesium- and rubidium-doped silicotungstic acid (STA) catalysts
Different caesium-doped STA catalysts were prepared using
aqueous solutions of caesium carbonate (0.05, 0.15, 0.25 and
.35 mol) in deionised water (10 mL), respectively. Separately, an
[
16]. Zirconia-supported HPA catalysts were found to be more
0
thermally stable, with an improved dispersion of the HPA active
phase, and this was found to be a key factor in tuning the activity
and selectivity towards acrolein [15,17]. HPAs supported on Al O
2 3
have been studied [4] notably by Martin et al. [18]. While high
acrolein selectivities (70–80%) could be obtained, rapid deactiva-
tion was observed. This was thought to be due to structural
changes in the HPA active phase at high temperatures or the block-
ing of the acid sites on the HPA due to coking. Phosphotungstic acid
aqueous solution of STA (0.06 M/20 mL) was prepared. The cae-
sium solution was added dropwise to the STA solution with stirring
over 30 min and then dried (110 °C, 16 h). Similarly, rubidium-
doped STA catalysts were prepared with rubidium acetate. These
catalysts are referred to as caesium (0.05 M)-doped silicotungstic
acid (0.06 M): Cs0.05/STA, Cs0.15/STA, Cs0.25/STA, Cs0.35/STA,
Rb0.05/STA, Rb0.15/STA and Rb0.25/STA.
(
PTA) and phosphomolybidic acid (PMA) HPAs were also evaluated
for glycerol dehydration [18,19], and high selectivities were ob-
served. These catalysts were very active and selective towards
acrolein achieving a selectivity of up to 98% at 100% glycerol con-
version but were found to have short lifetimes due to coking
2.2.3. Cs-doped HPAs (silicotungstic acid, silicomolybidic acid and
phosphomolybidic acid) Catalysts
Further caesium (0.25 M)-doped HPAs were also prepared from
an aqueous solution of caesium carbonate (0.25 M/10 mL) and
aqueous solutions (0.06 M/20 mL) of HPAs (silicotungstic acid, sil-
icomolybidic acid: SMA and phosphomolybidic acid: PMA). A cae-
sium solution was added dropwise to the aqueous solution of
HPAs, respectively, while stirring. This process turned the clear
solution of HPAs opaque, indicating precipitation of the catalyst.
Stirring was continued for 30 min followed by drying at 110 °C
overnight. These catalysts are denoted as Cs/STA, Cs/SMA and Cs/
PMA.
[
20]. Addition of alkaline and alkaline earth cations to these cata-
lysts improved the performance, but they were still found to deac-
tivate over a short reaction time [21].
It is therefore clear that for the application of this class of cata-
lyst to be viable on a larger scale, catalyst deactivation must be
overcome. One strategy to reduce deactivation has been to co-feed
oxygen in an effort to reduce the accumulation of coke on the cat-
alyst [22]. One approach adopted by Dubois and co-workers in-
+
volved the pre-treatment of the titania support with Cs followed
by impregnation by the phosphotungstic acid with co-fed oxygen
in the reaction stream [23]. These catalysts maintained glycerol
conversion of ca. 98% for 21 h with 78% acrolein yield [23].
However, the inclusion of oxygen increased the concentration of
CO and other oxygenated products and consequently decreased
2
the acrolein selectivity.
Against this background, we have developed caesium and
rubidium exchanged silicotungstic acids as catalysts for glycerol
dehydration to acrolein, and we have also investigated the effect
of supporting them on two different types of alumina. A major
aim of the work was to determine whether stable catalysts could
be prepared and operated in the absence of co-fed oxygen for the
transformation of relatively high concentrations of glycerol in an
aqueous feed.
2
.2.4. Cs- and Rb-doped STA supported on Al
catalysts
.05 M solutions of caesium and rubidium salts were prepared
2 3 2 3
O -1 and Al O -2
0
from their carbonate and acetates, respectively, in 10 mL of deion-
ised water. Metal solutions were added dropwise into the STA
(
0.06 M) solution, respectively, followed by the addition of the
Al -1 support while stirring. Stirring was continued for 20 h fol-
lowed by drying at 110 °C for 16 h. The same procedure was
followed for preparing Al -2-supported catalysts. These cata-
lysts are referred to as caesium-doped silicotungstic acid
supported on Al ; Cs/STA-1, Rb/STA-1, Cs/STA-2 and Rb/STA-2.
2 3
O
2 3
O
2 3
O
2.3. Catalyst characterisation
BET surface area analysis was carried out by means of nitrogen
adsorption using a Micromeritics Gemini instrument. Prior to anal-
ysis, all the samples were degassed (120 °C, 1 h) using a Micromer-
itics Flow prep 060 instrument. Powder XRD analysis was
conducted with an XpertPro Panalytical instrument using a Cu
2
. Experimental
2.1. Materials
Glycerol (99% Sigma Aldrich) was used as received. For the cata-
radiation source producing Ka monochromatic X-rays. The pattern
lyst preparation, silicotungstic acid {H
HSiW, 99.9% Sigma Aldrich) and rubidium acetate and caesium car-
bonate (Sigma Aldrich) were used. Two types of alumina supports
were used: -Al (Al -1) and a mixture of delta and theta
Al (Al -2) both supplied by Vertellus Speciality Chemicals.
4 2
[SiW12O40]-nH O,HSiW}
was collected over the 2h range of 10–80°. Raman spectroscopy
was carried out with a Renishaw inVia Raman Microscope using
a 514 nm laser source. Thermogravimetric analysis of the catalysts
was carried out using a Setaram Labsys instrument enabling con-
current weight loss with its differential and heat flow changes
(
a
2
O
3
2 3
O
2
O
3
2 3
O
(
2
TG-DTA/DSC). The experiments were carried out in an N atmo-
2
2
.2. Catalyst preparation
sphere from 30 °C to 600 °C with a ramp rate of 5 °C/min. Ammonia
temperature-programmed desorption analysis was carried out
using a TPDRO 1100 series instrument (Thermo Electron Corpora-
tion). All of the samples were pre-treated in helium (1 h, 100 °C)
followed by ammonia treatment (1 mL/min) and then ammonia
desorption up to 700 °C in a flow of helium.
.2.1. Silicotungstic acid on alumina catalysts
A series of catalysts with different wt.% loadings of silicotung-
stic acid (STA) (10%, 20%, 30%, 40% and 50%) on Al
pared by means of an incipient wetness method. The method
2 3
O -1 were pre-

2
08
M.H. Haider et al. / Journal of Catalysis 286 (2012) 206–213
2.4. Glycerol dehydration
Evaluation of catalyst performance was carried out using a lab-
oratory-scale microreactor (Fig. S1). The catalyst test equipment
comprised of four main parts: syringe pump, pre-heater, catalytic
reactor and cold traps for product collection. In addition, some
experiments to probe longer-term catalyst performance were car-
ried out using an HPLC pump in place of the syringe pump. The
aqueous glycerol feed, 0.5–20 wt.%, was fed into the pre-heater
a
b
Fig. 1. Acrolein b formation from glycerol a.
(
200 °C) at a liquid feed rate of 1 mL/h. The vapourised feed was
then swept through the system and through the catalyst bed in a
flow of inert nitrogen carrier gas (100 mL/min). All of the catalysts
were pressed and sieved to a uniform particle size distribution of
a
c
d
f
2
50–425
lm before use and were packed into the 0.8 mm i.d.
stainless steel reactor between plugs of silica wool. The catalysts
3
Fig. 2. Hydroxyacetone d formation.
(
density of ca. 1 g/mL) were packed to a volume of 0.25–3 cm ;
ꢀ1
the GHSV of the inert carrier was therefore 2000–24,000 h . After
exiting the reactor, the reaction products were collected in a series
of cold traps. Three traps were used as this was found to be
the most efficient method and ensured that any carry-over from
the first trap was collected in subsequent traps. The contents of
the cold traps were combined for analysis, which was performed
offline using a Varian CP 3800 gas chromatograph equipped with
capillary column (ZBWAX plus: i.d. 0.53 ꢁ 30 m). Gas samples were
also collected and analysed offline by means of a Varian CP 3800
a
e
00
GC with a Porapack Q: 1/8 ꢁ 2 m column. Mass balances were cal-
culated to be 99% +5% for all reactions.
3
. Results and discussion
Initially, catalysts were investigated using a low concentration
of glycerol (0.5 wt.%) to evaluate the effect of the various prepara-
tion techniques on acrolein selectivity. Different loadings of silico-
g
h
Fig. 3. Ethanal h formation.
2 3
tungstic acid on Al O -1 catalysts were tested at 315 °C for 3-h
time-on-stream (Table 1). The alpha-alumina-supported catalysts
have been reported previously as being suitable materials for the
preparation of acrolein [17]. As described previously, acid-
catalysed double dehydration of glycerol a leads predominantly
to acrolein b (Fig. 1). Inevitably, however, especially given that
hot acidic conditions are necessary to effect this transformation,
by-products are formed. Those identified in the present studies
are hydroxyacetone, ethanal, propanal, allyl alcohol, propene and
acetone; these were measured using gas chromatography and
identified by comparison with authentic samples. While purely
speculative, it is possible to deduce reasonable mechanisms to ex-
plain the formation of the two major by-products, hydroxyacetone
d and ethanal h. The former compound is most likely formed by an
initial dehydration involving a primary hydroxy group of glycerol.
This would produce the enol c, tautomerisation of which would
then lead to the observed product d (Fig. 2).
merisation of which would then give the hydroxy-aldehyde f,
3-hydroxypropanal, which is well set up, especially under acidic
conditions, to undergo fragmentation, as shown in Fig. 3. Although
ethanal h was detected, formaldehyde g, or obvious products aris-
ing from formaldehyde, such as formic acid, was not. However,
decomposition is also likely or simply loss of this highly volatile
product may explain why it was not detected. Perhaps signifi-
cantly, only a trace quantity of carbon dioxide was detected
amongst the various products.
The origins of the remaining and less abundant by-products are
not so clear. One obvious pattern is that all of these compounds,
propanal, allyl alcohol, propene and acetone, can be viewed as
deriving from the reduction of other products. Thus, propanal
could be formed by hydrogenation of acrolein b, allyl alcohol from
C@O reduction of the same major product, propene from hydrog-
enolysis of allyl alcohol; a similar reaction would give acetone from
hydroxyacetone d. Of course, other hydrogenolyses of CAO bonds
in earlier precursors or even in glycerol itself could also lead to
these by-products. One feature in favour of this deduction is that
additional experiments in this area using related catalysts have
clearly shown that a reducing atmosphere is generated during such
dehydrations, for reasons that are not entirely clear at present.
The supported 30% silicotungstic acid catalyst (30STA-1), the
best performing catalyst from the initial investigations, was tested
at different temperatures to determine the effect on acrolein selec-
tivity and glycerol conversion (Table 2). As the glycerol feed con-
centration was relatively low, relatively low reaction
temperatures were possible compared to 300–400 °C, which are
typically required [15,17]. Consequently, complete conversion
A likely pathway to ethanal h features a retro-aldol reaction as
shown in Fig. 3. In this case, an initial dehydration of the secondary
hydroxy group in glycerol would lead to the isomeric enol e, tauto-
Table 1
Product selectivities (in mol.%) obtained from glycerol (0.5 wt.%) dehydration over
ꢀ
1
various catalysts at 315 °C, GHSV 24,000 h , 3 h reaction duration.
Catalyst
Product selectivities (mol%)
Acrolein Acetol Ethanol Propanal Acetone Allyl alcohol
0
2
3
4
5
STA-1
21
21
4
2
1
–
11
13
12
12
14
1
1
2
1
1
44
8
7
6
6
3
0
0
0
0
0STA-1 74
0STA-1 80
0STA-1 80
0STA-1 78

M.H. Haider et al. / Journal of Catalysis 286 (2012) 206–213
209
Table 2
Product selectivities (in mol.%) obtained from glycerol (0.5 wt.%) dehydration over 30STA-1 catalyst over different reaction temperatures. GHSV 24,000 hꢀ1, 3-h reaction duration.
All at 100% glycerol conversion.
Temperature (°C)
Acrolein
Acetol
Ethanal
Acetone
Propanal
Acrylic acid
STYa
1
2
2
2
2
3
4
75
00
30
50
90
15
00
85
86
91
92
86
79
35
7
10
8
6
3
4
–
–
–
5
12
55
1
–
–
–
4
7
6
1
–
–
–
1
2
3
–
–
2
2
–
–
–
3.03
3.04
3.3
3.31
3.04
2.82
1.39
2
–
a
ꢀ1
ꢀ1
Space time yield (g
kg
h
).
ðacroleinÞ
ðcatÞ
was possible at lower temperatures. However, it was found that
the maximum acrolein selectivity was favoured over a temperature
range of 200–290 °C (Table 2). We consider that the acrolein selec-
tivity decreased above this temperature range due to the preferen-
tial formation of ethanal via the retro-aldol fragmentation pathway
illustrated in Fig. 3. This could be due to a change in the acidity of
the active sites or a loss of the supported Keggin structure. For the
reaction carried out at 250 °C, the acrolein selectivity was high at
7 h, the selectivity decreased to around 80% with conversion
decreasing to ca. 30% (Fig. 4). Carbonaceous deposits and the loss
of activity are common with acidic catalysts used for glycerol
dehydration [16,22]. Co-fed oxygen is one technique to reduce
coking; however, this is associated with an increase in carbon
oxide formation [24].
In an effort to improve catalyst selectivity and stability for reac-
tions without co-fed oxygen, a range of catalyst formulations were
again tested with a lower glycerol concentration. From this second-
ary screening, rubidium- and caesium-doped silicotungstic acid
catalysts were found to achieve high selectivity to acrolein at high
glycerol conversion (Table 3). Over the lowest concentration of
caesium and rubidium, i.e. 0.05 M-doped STA catalysts, a selectiv-
ity of 96% and 94%, respectively, at 100% glycerol conversion was
obtained. However, silicomolybidic and phosphomolybidic acid
doped with caesium were not as selective or as active in compari-
son. This loss of catalytic activity may be due to loss of structure of
the molybdenum compounds and complete decomposition of the
acid to metal oxide at the reaction temperature studied [19].
The best performing catalysts from this secondary screening
(Cs0.05/STA and Rb0.05/STA) were then selected to be supported
9
2% and at 100% glycerol conversion, and it was sustained for
3
-h time-on-stream.
Further testing of the 30STA-1 catalyst was conducted by
increasing the glycerol feed concentration to 20 wt.%. Increasing
the glycerol feed concentration affected the catalyst stability,
resulting in a decrease of activity, particularly at lower tempera-
tures (270–300 °C) (Fig. 4). Condensation of glycerol (b.p. 290 °C)
on the catalyst surface at these temperatures is considered as the
cause of the rapid loss of conversion. Therefore, reactions were per-
formed at higher temperatures, and the effect on acrolein selectiv-
ity was observed. Reactions carried out at 270 °C to 330 °C resulted
in an acrolein selectivity of ca. 90%. However, the acrolein selectiv-
ity decreased to ca. 85% following an increase in the reaction tem-
perature to 360 °C. The investigation of reaction temperature was
carried out over 4-h time-on-stream. The temperature at which
the acrolein yield was found to be highest and the glycerol conver-
sion sustained (330 °C) was applied to a longer reaction time
2 3 2 3
on Al O -1 and Al O -2 (XRD patterns shown in Fig. S2) and
investigated with a higher glycerol feed concentration (10 wt.%)
(Table 4). Initially, catalysts Cs/STA-1 and Rb/STA-1 both achieved
a high acrolein selectivity of 91% at 100% conversion; however, the
conversion decreased to ca. 37% and 44%, respectively, after 10-h
reaction time.
(
24 h). Product analysis over 24 h online revealed that an acrolein
selectivity of ca. 90% was maintained for the first 7-h on-stream
and 100% glycerol conversion was maintained. However, beyond
Metal-modified STA catalysts were supported on Al
analogously to those prepared with the Al -1 support, and the
effect on glycerol dehydration was studied (Table 5). With the
Al -2-supported catalysts, the maximum selectivity achieved
2 3
O -2
2 3
O
2 3
O
was ca. 90% at 100% conversion in the case of the Cs/STA-2 catalyst
at a temperature of 300 °C. In order to assess the catalyst stability
further much longer time-on-stream studies were performed. The
catalyst was found to be stable up to 90 h of reaction time with a
1
7
0 wt.% glycerol feed (Fig. 5). Conversion then decreased to ca.
5% beyond 90-h time-on-stream, decreasing further to ca. 55%
after 198-h on-stream. The catalyst was able to be partially regen-
erated at 350 °C by flowing 5% oxygen (bal. nitrogen) for 24 h,
regaining 90% conversion and 80% acrolein selectivity for a further
2
4-h on-stream at reaction temperature (Fig. 5). Furthermore, this
catalyst was also tested at a higher feed concentration (20 wt.%),
and the reaction was carried out for 10 h (Table 6). The catalyst
was stable, maintaining glycerol conversion and acrolein selectiv-
ity; 100% and ca. 90%, respectively, resulting in a space time yield
ꢀ
1
ꢀ1
(
STY) of 210 gacrolein kg
h . The space time yield achieved over
cat
1
ꢀ
ꢀ1
the Rb/STA-2 catalyst was determined to be 195 gacrolein kg
h ,
cat
and this was maintained over the 10-h reaction period. The high
selectivity achieved with the supported metal-doped STA catalysts
is considered to be related to the acid strength of the active sites.
We have observed that generally with STA-based catalysts, the ini-
tial acrolein selectivity is high. The stability of these active sites
Fig. 4. Time-on-stream glycerol conversion (closed symbols) and acrolein selec-
tivity (open symbols) over 30STA-1 catalyst at different temperatures. Conditions
ꢀ
1
2
0 wt.% glycerol feed, 1.2 mL/h, GHSV 2000 h . T; j/h = 270 °C, d/s = 300 °C, ꢀ/
e = 330 °C and N/4 = 360 °C.

2
10
M.H. Haider et al. / Journal of Catalysis 286 (2012) 206–213
Table 3
Product selectivities (in mol.%) obtained from glycerol (0.5 wt.%) dehydration over various catalysts at 250 °C, GHSV 24,000 hꢀ1, 3-h reaction duration.
Catalyst
Conversion (%)
Acrolein
Acetol
Ethanal
Propanal
Acetone
Acrylic acid
STYa
Cs0.05/STA
Cs0.15/STA
Cs0.25/STA
Cs0.35/STA
Cs0.25/SMA
Cs0.25/PMA
Rb0.05/STA
Rb0.15/STA
Rb0.25/STA
100
10
97
2
31
96
27
93
0
26
33
94
92
90
4
27
5
44
5
7
4
7
8
–
–
–
–
24
11
–
–
–
–
–
–
–
2
–
–
–
–
–
–
–
–
6
–
–
–
–
–
22
3
56
7
11
2
3.49
0.01
3.1
0
0.08
0.06
3.32
3.26
2.14
23
100
100
82
2
2
a
ꢀ1
ꢀ1
Space time yield (g
kg
h
).
ðacroleinÞ
ðcatÞ
Table 4
Product selectivities (in mol.%) obtained from glycerol (10 wt.%) dehydration over Al
-1-supported Cs- and Rb-doped STA catalysts. GHSV 12,000 h 1
ꢀ
.
2
O
3
Catalyst
Temperature (°C)
Time (h)
Conversion (%)
Acrolein
Acetol
Ethanal
Propanal
Acetone
STYa
Cs/STA-1
300
1.5
5
100
66
37
100
60
31
91
87
85
89
84
83
9
10
13
6
10
11
–
1
–
2
4
4
–
1
1
2
1
1
–
–
2
1
1
99
41
12.4
95.5
32
1
0
330
1.5
5
1
0
8.7
Rb/STA-1
300
1.5
5
87
54
35
100
69
44
91
88
88
91
85
84
6
10
9
4
10
12
1
1
2
3
3
3
1
1
1
1
1
1
1
1
1
1
1
1
74
28
12
100
43.3
17.7
1
0
330
1.5
5
1
0
a
ꢀ1
ꢀ1
Space time yield (g
kg
h
).
ðacroleinÞ
ðcatÞ
Table 5
Product selectivities (in mol.%) obtained from glycerol (10 wt.%) dehydration over Al
2 3
O .
-2-supported STA and Cs- and Rb-doped STA catalysts. GHSV 12,000 h 1
ꢀ
Catalyst
0STA-2
Temperature (°C)
Time (h)
Conversion (%)
Acrolein
Acetol
Ethanal
Propanal
Acetone
STYa
3
300
1.5
5
83
100
100
75
73
71
6
5
8
8
9
9
6
8
7
1
1
2
41
79
77
1
0
Cs/STA-2
Rb/STA-2
300
1.5
5
97
100
100
100
100
88
87
84
78
77
6
9
11
4
4
3
3
13
11
2
1
1
3
2
1
–
1
2
1
94
98
96
89
84
1
0
3
30
300
30
1.5
5
9
1.5
5
1.5
5
100
100
100
100
83
79
75
73
4
8
4
8
1
4
15
10
2
2
3
2
1
1
1
1
91
87
85
83
3
a
ꢀ1
ꢀ1
Space time yield (g
kg
h
).
ðacroleinÞ
ðcatÞ
based on the Keggin structure has been improved by partially dop-
ing the STA with Cs or Rb and supporting on Al -2. Therefore, the
octahedral sites for delta and they are distributed equally amongst
octahedral and tetrahedral sites for the theta phase, which have up
to 480 vibrational modes, making it highly surface active [25–28].
2 3
O
catalyst acidity is maintained at the desired strength, typically
thought of over a HA range of ꢀ2 to ꢀ8 [10,13,14], and this is cru-
cial for high acrolein selectivity [15,16]. We will discuss the stabil-
ity of the active site with respect to the high acrolein yield further
with the aid of various characterisation techniques.
2 3
The catalysts with this stable, low surface area Al O -1 support
(Table S1) resulted in very weak interactions with the deposited
acid species [13]. For long-term catalytic activity of silicotungstic
acid-based catalysts, a strong interaction of the silicotungstic acid
with the support is desirable [29]. This strong interaction provides
stability under the high temperature conditions required for glyc-
The stability of the alumina-supported catalysts over reaction
times of 10 h was investigated. Al
found to be less stable for time-on-stream studies when compared
to theta-delta Al -2-supported catalysts. Alpha-alumina is
2 3
O -1-supported catalysts were
2 3
erol dehydration [17]. Thermal analysis of the Al O -1-supported
2
O
3
catalysts illustrates the instability associated with this weaker
interaction. The lower stability can be related to the presence of
an exothermic peak at around 500 °C in the heat flow analysis pro-
file (Fig. 6). This feature is related to undispersed or weakly bound
silicotungstate species on the support [15,17]. These species are
considered to decompose during the glycerol dehydration reaction
resulting in a loss of catalytic activity. However, this thermal
thought to be the most stable phase of alumina, which is related
to its low surface area obtained upon heating boehmite (AlOOH).
3
+
In this phase, Al cations occupy only 2/3 of the octahedral inter-
stitial sites corresponding to its low surface area when compared
to theta-delta Al
superstructure, in which Al ions occupy more than 13 out of 16
2 3
O -2. This phase contains a monoclinic spinel
3+

M.H. Haider et al. / Journal of Catalysis 286 (2012) 206–213
211
Fig. 7. Heat flow analysis of Cs-doped STA catalysts; a = Cs0.05/STA, b = Cs0.15/STA,
c = Cs0.25/STA and d = Cs0.35/STA, dotted line: zero heat flux.
Fig. 5. Time-on-stream glycerol conversion and acrolein selectivity over Cs/STA-2
catalyst at 300 °C with a period of regeneration. Conditions 10 wt.% glycerol feed,
ꢀ
1
1
mL/h, GHSV 6000 h . Catalyst regeneration at 200 h, 350 °C for 24 h with 5% O
2
in N (100 mL/min).
2
2
00 °C are considered to be due to the removal of physisorbed
and hydrated water from the heteropoly acid [30]. Those features
found in the range of 200 °C and 300 °C are considered to be due
to structural changes of the partially doped acid species which
are not visible with other fully doped STA catalysts [32]. The same
behaviour was observed with rubidium-doped silicotungstic acid
catalysts (Fig. S3). Therefore, over the reaction temperatures used
the structure of the metal-doped catalysts should be stable.
However, supporting the fully doped STA material on the alumina
Table 6
Product selectivities (in mol.%) obtained from glycerol (20 wt.%) dehydration over
various catalysts at 300 °C, GHSV 6000 h , 10-h reaction duration.
ꢀ
1
Catalyst Conversion Acrolein Acetol Ethanal Propanal Acetone STYa
(%)
Cs/STA- 100
2
88
8
2
1
1
210
leads to weak binding with the support due to the lack of available
þ
H
5
O
which interacts with surface species on the support. In the
2
case of the partially doped Cs/STA, supporting this material leads
þ
to improved binding due to the presence of the remaining H
5
O
2
moieties. These enable the stronger STA–support interaction. Char-
acterisation of the supported catalyst was conducted, and thermal
analysis indicated that the caesium silicotungstate moiety sup-
ported on Al O -1 was weakly bound with an exothermic peak
2 3
around 500 °C (Fig. 8a), and as mentioned previously, this is related
to the STA decomposition. In the case of the catalysts supported on
2 3
Al O -2, there was strong binding between the support and the
caesium-doped silicotungstate species showing no exothermic re-
sponse for unbound or weakly bound species (Fig. 8b). In addition,
the partially doped Cs/STA materials on both supports (Al
2 3
O -1 and
2
) no longer have endothermic heat transitions associated with the
instability of the Keggin structure. For the supported Rb/STA
materials, a similar trend emerges with respect to the Keggin struc-
Fig. 6. Heat flow analysis of 30STA catalysts supported on alumina (1 and 2), dotted
line: zero heat flux.
2 3
transition was not present with the Al O -2-supported catalysts,
which showed stronger binding of the silicotungstic acid with
the support, resulting in increased catalyst stability (Fig. 6).
In the case of caesium-doped silicotungstic acid catalysts,
where an increase in the Cs concentration led to a decrease in
activity, a similar trend prevails. Cs (0.05 M)-doped silicotungstic
þ
acid partially replaces the H
5
O
moiety of silicotungstic acid which
2
can be seen in the form of an exothermic peak around 500–550 °C
and is attributed to instability of the modified silicotungstic acid
þ
species. However, at loadings above 0.05 M, the H
5
O
+
moiety of sil-
2
icotungstic acid was completely replaced with Cs to form a stable
Cs-doped silicotungstate phase [30,31]. Therefore, at higher load-
ings, no exothermic response was observed in the heat flow anal-
ysis profile (Fig. 7). The heat flow features recorded at 100 and
Fig. 8. Heat flow analysis of Cs- and Rb-doped STA catalysts supported on alumina
(
1 and 2); a = Cs/STA-1, b = Rb/STA-2, c = Rb/STA-1, d = Cs/STA-2, dotted line: zero
heat flux.

2
12
M.H. Haider et al. / Journal of Catalysis 286 (2012) 206–213
and strong), and they were in greater density when compared to
the Al -1 support, which contained only small amounts of med-
ium acid sites (Fig. 10). Analysis of the CsSTA-2 catalyst without
NH pre-treatment indicated that the material was stable over
e
d
c
2 3
O
3
the temperature range investigated.
Laser Raman spectroscopy was used for the analysis of Keggin
structure for the unsupported Cs-doped STA catalysts in Fig. S4.
ꢀ
1
The transitions shown at ca. 998 cm are attributed to the W@O
ꢀ1
b
a
stretch, ca. 974 cm for WAOAW bending, ca. 850 for WAOAW
ꢀ1
stretching, ca. 555 cm for OASiAO bending, while those at 110,
ꢀ1
1
3
50 and 225 cm are associated with WO [34]. There is a loss
of several transitions related to the Keggin structure with the high-
est Cs loading, and we attribute this to the distortion of the STA
structure. Caesium addition induces a contraction effect on the
100
200
300
400
500
600
700
þ
STA structure as Cs replaces the H
2
O
moieties [35]. This corre-
5
Temperature (°C)
sponds to the X-ray analysis results in which the reflection related
to the Cs/STA phase found at 26° is completely lost in the case of
Fig. 9. Ammonia thermal-programmed desorption analysis of Cs-doped STA
catalysts, a = Cs0.05/STA (with no NH3 pre-treatment), b = Cs0.05/STA, c = Cs0.15/
STA, d = Cs0.25/STA and e = Cs0.35/STA.
0
.35 M catalyst (Fig. S5). As previously reported, silicotungstic acid
needs strong binding sites on the supporting material in order to
maintain the active silicotungstate phase over a long time period
at a higher temperature [17]. Intense stretching and bending vibra-
tional bands of tungsten-oxygen were seen in the Raman shift for
the weakly bound caesium silicotungstate species on Al O -1
ture, instability due to a transition at ca. 200 °C for the Rb/STA-1.
The transition at 500 °C is related to weak binding to the support
which is not present for the Rb/STA-2 (Fig. 8c and d).
2
3
(Fig. 11c), as compared to the weak band found with Al O -2
2
3
The nature of acidic sites, i.e. weak, medium or strong, on the
catalyst surface was determined by ammonia temperature-
programmed desorption. The classification of acid site strength is
considered as follows: weak (150–300 °C), medium (300–500 °C)
and strong (P500 °C) [33]. Catalysts doped with different concen-
trations of caesium were analysed (Fig. 9). All materials possess
acid sites of medium strength. However, we observe decomposi-
tion of the STA structure at higher temperatures. This is confirmed
by the TPD analysis the Cs0.05/STA sample which was not pre-
(Fig. 11a). Moreover, in the case of Cs/STA-2, the main W@O
stretching, and the WAOAW bending shoulder band, was less de-
fined and along with the shifting of the WAOAW stretch towards
lower wavelength, which is considered to be due to the bond
weakening/distortion by strong interaction with the surface of
the alumina-2 support [15,36]. Analysis of the t(W@O) transition
after use indicates that the stability of the silicotungstate is lower
with the weakly bound species on Al O -1 (Fig. 11d) when com-
2
3
pared to the species on Al O -2 which retains this transition
2
3
treated with NH
3
, which showed significant desorption of decom-
(Fig. 11b).
position products at ca. 550 and 700 °C (Fig. 9a). The slow rise of
the TCD response over the temperature range is considered to be
due to thermal expansion of the carrier gas. The best performing
catalyst (Fig. 9b) has a low density of medium acid sites, whereas
the catalyst with the highest density of medium sites and a broader
feature present in the TPD profile has the highest Cs loading and
exhibits poor catalytic activity.
Considering the differences in the catalysts prepared on differ-
2 3
ent alumina supports, ammonia TPD indicated that the Al O -2
support contained a broad range of acidic sites (weak, medium
XRD patterns for the bulk silicotungstic acid show distinct
reflections from (220), (310), (222), (400), (411), (420), (332)
and (510) lattice planes reported by Pizzio et al. [30], which
can be seen in the lower concentration caesium catalyst along with
the new crystalline peaks of caesium-doped silicotungstate
species (Fig. S5) [31]. However, in the case of fully doped silico-
tungstic acid catalysts, the parent STA phases are replaced with
the Cs/STA phases showing complete doping. The caesium- and
rubidium-doped silicotungstic acid catalysts were found to be
selective for acrolein over 3-h on-stream, however, at a low
c
b
a
100
200
300
400
500
600
700
Temperature (°C)
Fig. 10. Ammonia thermal-programmed desorption analysis of Cs-doped STA
supported on alumina (1 = a and 2 = c) catalysts, b = Cs/STA-2 with no NH3 pre-
treatment.
Fig. 11. Raman spectroscopy of Cs-doped STA catalysts supported on alumina prior
to reaction and postreaction. a = Rb/STA-2; fresh, b = Cs/STA-2; fresh, c = Cs/STA-2;
used and d = Cs/STA-1; fresh and e = Cs/STA-1; used.

M.H. Haider et al. / Journal of Catalysis 286 (2012) 206–213
213
knowledge, to be the most stable silicotungstic acid-derived cata-
lyst and one that does not require oxygen in the feed gas to achieve
stable operation. Analysis of the catalysts suggests that the origin
of the long-term stability is related to the strength of the partially
doped silicotungstic acid on the alumina support. Doping with cae-
sium maintains the Keggin structure of the silicotungstic acid,
resulting in long-term stability and high acrolein yield observed.
Acknowledgments
This work formed part of the Glycerol Challenge. The Technol-
ogy Strategy Board is thanked for their financial support. This pro-
ject is co-funded by the Technology Strategy Board’s Collaborative
Research and Development programme, following an open compe-
tition. The Technology Strategy Board is an executive body estab-
lished by the Government to drive innovation. We thank
Vertellus Specialities Chemicals, USA, for the alumina supports
used in this study.
Appendix A. Supplementary material
Fig. 12. X-ray diffraction patterns of Cs-doped supported STA (1 and 2), a = Cs/STA-
fresh, b = Cs/STA-2 used (10 h reaction), c = Cs/STA-1 fresh, (circles = Cs/STA
species and blocks = respective Al support).
2
2
O
3
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.jcat.2011.11.004.
glycerol feed concentration. These catalysts are reported to have
insufficient stability over a longer reaction time to tolerate higher
glycerol feed concentrations [17,20]. Thermal analysis of the most
selective catalyst, the partially doped Cs/STA, indicated that it
would not be sufficiently stable over a longer reaction time with
higher feed concentrations. Therefore, for use under increased
glycerol feed conditions, the stability and long-term activity of
the Cs-based STA would require a support.
References
[1] A.J. Ragauskas, C.K. Williams, B.H. Davison, G. Britovsek, J. Cairney, C.A. Eckert,
W.J. Frederick Jr., J.P. Hallett, D.J. Leak, C.L. Liotta, J.R. Mielenz, R. Murphy, R.
Templer, T. Tschaplinski, Science (Washington, DC, US) 311 (2006) 484–489.
[
2] M. Snare, I. Kubickova, P. Maeki-Arvela, D. Chichova, K. Eraenen, D.Y. Murzin,
Fuel 87 (2008) 933–945.
[3] United Nations Framework Convention on Climate Change. <http://unfccc.int>.
[
4] B. Katryniok, S. Paul, V. Belliere-Baca, P. Rey, F. Dumeignil, Green Chem. 12
2010) 2079–2098.
5] G.W. Keulks, L.D. Krenzke, T.M. Notermann, Adv. Catal. 27 (1978) 183–225.
(
Partially caesium- and rubidium-doped STA catalysts were sup-
[
ported on alumina, and the delta-theta-Al
were stable for longer reaction times compared to the alpha-
Al catalysts. The XRD pattern for Al -1 support has the char-
acteristic diffraction reflections of the alpha phase (Fig. S2) (ICDD
ref. code 01-078-2427). These diffraction reflections further
indicate that the binding of the doped STA is weak as we observed
additional peaks in the case of catalyst (for caesium-doped
silicotungstic acid phase) (Fig. 12) [31]. However, in the case of
O
2 3
-supported catalysts
[6] L.D. Krenzke, G.W. Keulks, A.V. Sklyarov, A.A. Firsova, M.Y. Kutirev, L.Y.
Margolis, O.V. Krylov, J. Catal. 52 (1978) 418–424.
[
[
7] J. Hagen, Industrial Catalysis: A Practical Approach (1999) 440.
8] B. Katryniok, S. Paul, M. Capron, F. Dumeignil, ChemSusChem 2 (2009) 719–
730.
2
O
3
2 3
O
[
9] L.P. Hammett, A.J. Deyrup, J. Am. Chem. Soc. 54 (1932) 4239–4247.
[
[
10] A. Neher, T. Haas, D. Arntz, H. Klenk, W. Girke, US Patent 5387720, 1995.
11] S. Kuba, P. Concepcion Heydorn, R.K. Grasselli, B.C. Gates, M. Che, H. Knozinger,
Phys. Chem. Chem. Phys. 3 (2001) 146–154.
12] B. Katryniok, S. Paul, M. Capron, C. Lancelot, V. Belliere-Baca, P. Rey, F.
Dumeignil, Green Chem. 12 (2010) 1922–1925.
13] S.-H. Chai, H.-P. Wang, Y. Liang, B.-Q. Xu, Green Chem. 9 (2007) 1130–1136.
[14] S.-H. Chai, H.-P. Wang, Y. Liang, B.-Q. Xu, J. Catal. 250 (2007) 342–349.
15] S.-H. Chai, H.-P. Wang, Y. Liang, B.-Q. Xu, Appl. Catal. A 353 (2009) 213–222.
16] E. Tsukuda, S. Sato, R. Takahashi, T. Sodesawa, Catal. Commun. 8 (2007) 1349–
[
[
Al
2
O
3
-2 support, there is strong interaction between the support
-2 sup-
and the active material as distinct reflections for the Al
2
O
3
[
[
port were not observable (Fig. 12). The diffraction pattern for the
Cs/STA-2, which had undergone catalyst testing for 10 h, indicated
that there was no observable loss of the Cs/STA species (Fig. 12b).
1353.
[17] S.-H. Chai, H.-P. Wang, Y. Liang, B.-Q. Xu, Green Chem. 10 (2008) 1087–1093.
[
[
18] H. Atia, U. Armbruster, A. Martin, J. Catal. 258 (2008) 71–82.
19] S. Erfle, U. Armbruster, U. Bentrup, A. Martin, A. Brueckner, Appl. Catal. A 391
4
. Conclusions
(
2011) 102–109.
20] A. Alhanash, E.F. Kozhevnikova, I.V. Kozhevnikov, Appl. Catal. A 378 (2010) 11–
8.
[
1
Acidic heteropoly anion-based catalysts have been prepared
[
[
21] H. Atia, U. Armbruster, A. Martin, Appl. Catal. A 393 (2011) 331–339.
22] J.L. Dubois, C. Duquenne, W. Holderich, FR Patent 2005-1499-2882052, 2006.
and evaluated for the catalytic dehydration of glycerol. Initially,
catalysts were screened at various temperatures and with different
glycerol feed concentrations. Rubidium- and caesium-doped silico-
tungstic acid catalysts were found to be very selective for acrolein
formation. However, these unsupported catalysts were not stable
over longer reaction times even with a low glycerol feed concen-
tration (0.5 wt.%). Supporting these partially doped catalysts on
[23] Y. Magatani, K. Okumura, J.-L. Dubois, J.-F. Devaux, WO Patent 2009-JP67115-
011033689, 2011.
[
[
2
24] F. Wang, J.-L. Dubois, W. Ueda, Appl. Catal. A 376 (2010) 25–32.
25] A. Boumaza, L. Favaro, J. Ledion, G. Sattonnay, J.B. Brubach, P. Berthet, A.M.
Huntz, P. Roy, R. Tetot, J. Solid State Chem. 182 (2009) 1171–1176.
[26] P. Colomban, J. Mater. Sci. 24 (1989) 3002–3010.
[
[
27] N. Fremy, V. Maurice, P. Marcus, Surf. Interface Anal. 34 (2002) 519–523.
28] G. Paglia, A.L. Rohl, C.E. Buckley, J.D. Gale, Phys. Rev. B: Condens. Matter Mater.
Phys. 71 (2005) 224115.
Al
2
O
3
-1 (alpha) did not result in a stable catalyst. However, sup-
-2 (mixture of
porting caesium-doped silicotungstic acid on Al
theta and delta phases) was found to be stable for up to 90-h reac-
tion time and gave a maximum selectivity of ca. 90% acrolein along
2
O
3
[29] I.V. Kozhevnikov, Chem. Rev. (Washington, DC) 98 (1998) 171–198.
[
[
[
30] L.R. Pizzio, M.N. Blanco, Micropor. Mesopor. Mater. 103 (2007) 40–47.
31] J.A. Santos, Proc. Roy. Soc. London, Ser. A 150 (1935) 309–322.
32] A.A. Babad-Zakhryapin, Izv. Akad. Nauk SSSR, Ser. Khim. (1963) 215–220.
at 100% glycerol conversion with
a
space time yield of
(10 wt.% glycerol). When the glycerol feed
concentration was increased to 20 wt.% glycerol, a space time yield
[33] J.C. Yori, J.M. Grau, V.M. Benitez, J. Sepulveda, Appl. Catal. A 286 (2005) 71–78.
[34] P. Sabatier, G. Gaudion, Comptut. Rend. 166 (1918) 1033–1039.
ꢀ
1
ꢀ1
1
05 gacrolein kg
h
cat
[
35] L. Pesaresi, D.R. Brown, A.F. Lee, J.M. Montero, H. Williams, K. Wilson, Appl.
Catal. A 360 (2009) 50–58.
ꢀ
1
cat
ꢀ1
of 210 gacrolein kg
h
was achieved, although it was stable for a
[36] R. Thouvenot, M. Fournier, R. Franck, C. Rocchiccioli-Deltcheff, Inorg. Chem. 23
1984) 598–605.
(
shorter time-on-stream. This catalyst appears, to the best of our