Dehydration of glycerol in gas phase using heteropolyacid catalysts as active compounds

Article abstract of DOI:10.1016/j.jcat.2008.05.027

Different silica-, alumina-, and aluminosilicate-supported heteropolyacid catalysts were prepared using phosphomolybdic acid H₃PMo₁₂O₄₀sxH₂O, phosphotungstic acid H₃PW₁₂O₄₀sxH

Full text of DOI:10.1016/j.jcat.2008.05.027

Journal of Catalysis 258 (2008) 71–82
Contents lists available at ScienceDirect
Journal of Catalysis
www.elsevier.com/locate/jcat
Dehydration of glycerol in gas phase using heteropolyacid catalysts as active
compounds
Hanan Atia a, Udo Armbruster b, Andreas Martin b,∗
a
Faculty of Women for Science, Arts and Education, Ain Shams University, Cairo, Egypt
Leibniz Institut für Katalyse e.V. (branch Berlin), Richard-Willstätter-Str. 12, D-12489 Berlin, Germany
b
a r t i c l e i n f o
a b s t r a c t
Article history:
Different silica-, alumina-, and aluminosilicate-supported heteropolyacid catalysts were prepared using
phosphomolybdic acid H3PMo12O40·xH2O, phosphotungstic acid H3PW12O40·xH2O, silicotungstic acid
H4SiW12O40·xH2O, and ammonium phosphomolybdate (NH4)3PMo12O40·xH2O as precursor compounds.
The as-synthesised solids were characterised by nitrogen adsorption, XRD, TG/DTA, Raman spectroscopy,
and TPD of ammonia. Silica-supported heteropolyacids are rather well crystallised, whereas alumina-
supported samples are X-ray amorphous. Investigations using Raman spectroscopy of calcined samples
Received 19 March 2008
Revised 24 May 2008
Accepted 29 May 2008
Available online 30 June 2008
Keywords:
Glycerol
Dehydration
Acrolein
Heteropolyacids
Catalysis
Gas phase
and TG/DTA revealed that molybdenum-containing heteropolyacids tend to decompose partly close to
◦
400
C into molybdates and MoO3 whereas tungsten-containing samples are stable. This makes in
particular tungsten-based materials interesting acid catalysts for the dehydration of glycerol in the gas
phase. In particular, the influence of selected support materials, catalyst loading, and temperature on
acrolein formation was studied at standardised reaction conditions (10% by weight of glycerol in water,
◦
225–300 C, modified contact time 0.15 kg h mol−1). Surprisingly, alumina is found to be superior to
silica as support material with regard to catalyst activity and selectivity. Nevertheless, tungsten based
heteropolyacids showed outstanding performance and stability. Acrolein was always the predominant
product with maximum selectivity of 75% at complete conversion over silicotungstic acid supported over
alumina and aluminosilicate.
© 2008 Elsevier Inc. All rights reserved.
1. Introduction
erol [3]. An intrinsic problem in glycerol conversion is the presence
of primary and secondary hydroxyl groups with different reactiv-
The rapidly rising production of biodiesel from vegetable oils
has led to a drastic surplus of by-product glycerol in the chemical
markets. Glycerol production in the US already averages more than
350,000 tons per year and in Europe, its production has tripled
within the last ten years to approximately 600,000 tons per year.
This may increase further as European Commission implemented
EU directive 2003/30/EC which targets at the addition of 5.75% of
biofuel to car fuels by the year 2010 [1]. Unfortunately, the cur-
rent application of glycerol is mainly confined to pharmaceuticals
and cosmetics and hence the demand is somewhat limited. The
availability of large amounts of cheap glycerol is the driving force
to develop new processes for its energetic or chemical utilisation
(e.g., [2]). Finding value-added alternatives to glycerol incineration
would improve economic viability of biodiesel manufacture and
the biofuel supply chain [1].
ity that induce selectivity problems in redox as well as acid–base
reactions. Commercial glycerol oxidation processes using hydrogen
peroxide are known to produce glycerol aldehyde [4,5] and dihy-
droxyacetone as an intermediate for glyceric acid and syrine [6,7]
in liquid phase. To overcome the reactivity/selectivity problem, re-
searchers tried to synthesise glycerol derivatives, e.g. acetals or
esters that allow the subsequent oxidation of only one type of car-
bon atom [8]. Hydrogenolysis of glycerol on Ru/C catalysts in the
presence of an acidic catalyst like amberlyst leads to 1,2- and 1,3-
propanediol at mild reaction conditions [9]. The overall reaction is
indeed a combination of dehydration and subsequent hydrogena-
tion via the formation of hydroxyacetone.
The best-known acid catalysed reactions with glycerol are
(i) condensation to form linear or cyclic glycerol dimers and
oligomers [10,11] and (ii) dehydration to form acrolein and other
compounds. To perform the dehydration of glycerol to acrolein,
strong catalyst acidity seems to be necessary. Various solid acid
catalysts including sulphates, phosphates and zeolites have been
tested for the dehydration of glycerol in either gaseous or liquid
phases [12–14]. Dehydration of glycerol (and other alcohols) was
Chemical utilisation of glycerol might include such attractive
chemicals as acrolein, propanediols or epichlorohydrin. Recently,
Solvay installed an epichlorohydrin process (Epicerol) using glyc-
Corresponding author. Fax: +49 30 6392 4454.
E-mail address: andreas.martin@catalysis.de (A. Martin).
*
◦
also carried out in supercritical water (T > 374 C, p > 221 bar)
0021-9517/$ – see front matter © 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2008.05.027

72
H. Atia et al. / Journal of Catalysis 258 (2008) 71–82
with H2SO4 (acrolein selectivity of 84% at 40% conversion [15]) or
with ZnSO4 as catalyst (75% selectivity at 50% conversion [16]).
Acrolein selectivity up to 80% at 90% conversion was reported with
Table 1
Results from BET measurements for catalysts and supports
Material
ABET
Percentage of deviation Pore volume
Average pore
diameter (nm)
−1
−1
(m2
g
)
relative to support
(cm3
g
)
◦
H2SO4 as catalyst at 400 C [17]. Studies by Degussa in the mid-
nineties showed that the reaction can also be carried out in liquid
A5
297.7
253.5
249.7
306.9
307.0
292.1
–
0.427
0.293
0.278
0.355
0.361
1.192
0.828
0.327
0.221
0.224
0.237
0.271
1.189
0.925
0.527
0.398
4.6
3.9
3.8
4.1
4.1
11.9
11.3
1.5
1.6
1.6
1.7
1.7
11.0
10.8
4.4
◦
◦
HPMo20/A5
HSiW20/A5
HPMo10/A5
HSiW10/A5
A12
−14.9
−16.1
+3.1
+3.1
–
water at 180–340 C and in gas phase at 250–340 C with sev-
eral acidic catalysts like HZSM-5 or HY zeolites, supported mineral
acids, and heteropolyacids. Acrolein yields up to 75% at complete
conversion were obtained [13]. Chai et al. communicated the use
of sulphated zirconia [18] or Nb2O5 [19] but these catalysts did
not show better performance. Very recently, Tsukuda et al. [20]
tested heteropolyacids supported over SiO2 and found that the in-
troduction of mesopores in silica support significantly affected the
catalytic activity. However, such strongly acidic catalysts deactivate
quickly as carbonaceous deposits block the catalyst surface, in par-
ticular in the presence of large amounts of micropores.
HSiW20/A12 250.2
S2
−14.3
648.8
–
HPMo20/S2
HSiW20/S2
HPMo10/S2
HSiW10/S2
S11
HSiW20/S11 259.5
AS4 383.6
HSiW20/AS4 309.2
418.4
422.8
442.0
528.1
320.7
−35.6
−34.8
−31.9
−18.6
–
−19.1
–
Heteropolyacids (HPA) are a class of materials with well-defined
structures and widely tuneable acidity, making them excellent cat-
−19.4
3.5
alysts for acid–base reactions. Their chemical composition is typi-
8−n
cally described as [Xn+Y12O40
]
·xH2O (with X = P, Si, B; Y = Mo,
contains the abbreviation of the HPA, 20 or 10 indicates the load-
ing percentage, while A5 represents the type of support and its
average pore diameter in nm.
W, V; n = oxidation state of X). Different polyanion structures
and molecular arrangements for heteropolyacids exist, like Ander-
son, Dawson and Keggin structures, but most of the catalytic re-
searchers have focused on the latter type. These solids are soluble
in polar solvents and have strong Brønsted acidity as proofed in
many studies [21]. Besides that, choice of an appropriate metal Y
that owns different oxidation states makes them also active in re-
dox catalysis. Heteropolyacids offer structural flexibility allowing
for exchange of water and polar organic molecules. On the other
hand, they lack thermal stability and have low specific surface area
(1–5 m2/g). This is the reason why heteropolyacids are often sup-
ported over acidic or neutral carriers like Al2O3 or SiO2 [21].
In the present investigation, several supported heteropolyacids
(HPA) were prepared, characterised and tested in dehydration of
glycerol in gas phase for their catalytic performance. Emphasis was
also laid on reaction engineering aspects like selectivity/conversion
dependency of the reaction on these catalysts and long-term sta-
bility experiments.
2.2. Catalyst characterisation
BET surface area, pore volume, and average pore diameter for
fresh and spent solids were derived from nitrogen adsorption
◦
isotherms measured at −196 C (Micromeritics ASAP 2020). Before
◦
the measurement, each sample was degassed at 200 C for 4 h. The
average pore diameters were calculated according to BJH method.
The X-ray powder diffraction measurements were carried out
on a STADI P automated transmission diffractometer (STOE, Darm-
stadt) with CuKα1 radiation and Ge monochromator. The pattern
◦
◦
was scanned in the 2θ range of 5–60 (step width 0.5 , 100 s per
step) and recorded with a STOE position sensitive detector (PSD).
The samples were prepared to flat plates. The phase analysis was
carried out with the Win Xpow software package, including the
powder diffraction file (PDF).
Raman spectroscopic investigations were carried out using a fi-
bre optical RXN-Spectrometer (Kaiser Optical Systems, 785 nm Laser)
with a Laser power of 50 mW.
2. Experimental
2.1. Catalyst synthesis
◦
Thermogravimetric analysis (TG) of fresh (dried at 120 C only)
as well as of spent catalysts was performed using a SETARAM TG-
DTA 92-12 instrument that allowed both TGA and DTA curves to be
recorded simultaneously. The experiments were carried out from
room temperature up to 800 C with a heating rate of 10 K/min
under airflow (1 l/h).
Four commercially available heteropolyacids such as phos-
phomolybdic acid H3PMo12O40·xH2O (HPMo, Merck), phospho-
tungstic acid H3PW12O40·xH2O (HPW, Aldrich), silicotungstic acid
H4SiW12O40·xH2O (HSiW, Fluka) and ammonium phosphomolyb-
date (NH4)3PMo12O40·xH2O (NH4PMo, ABCR) are used as pre-
cursors for preparing supported catalysts by incipient wetness
method. More details on the catalyst preparation are described
below. Aqueous solutions with calculated amounts of the precur-
sors (10 or 20 wt%) were added separately to solid supports such
as Al2O3 (pore diameter of 5 nm, support denoted as A5; Fluka)
and SiO2 (2 nm, S2; Fluka) under stirring for 20 h. Excess water
◦
TPD of ammonia (TPDA) profiles were measured on a commer-
cial characterisation equipment (AMI-1, Altamira/Zeton) under He
◦
flow (30 ml/min) with a heating rate of 5 K/min up to 600 C
and 1.5 h hold time. After absorption in 0.05 N H2SO4, the total
amount of ammonia was quantified by back-titration with 0.05 N
NaOH versus Tashiro indicator. The desorption curves were moni-
tored with a thermal conductivity detector (TCD). Desorbed water
and possibly desorbed CO2 were captured in a trap with KOH split
before entering the TCD. Prior to recording the curves the samples
◦
was removed with a rotary evaporator at 60 C. The samples were
◦
◦
dried for 2 h at 120 C in air and finally calcined at 400 C for
4 h under airflow. After calcination a prompt usage for catalytic
tests was realised. Additionally, the samples were heated before
catalytic tests under nitrogen to the desired reaction temperatures
to remove physisorbed water.
◦
were heated to 400 C (10 K/min) and held for 1 h under flowing
He.
2.3. Catalytic tests
Another series of catalysts with 20% by weight of silicotungstic
acid (HSiW) was supported over aluminosilicate (pore diameter
of 4 nm, support denoted as AS4 (Siral® 10); Condea), alumina
(12 nm, A12; Fluka), and silica (11 nm, S11; Fluka) (Table 1). All
these samples were dried and calcined as described above. The
catalysts will be marked as following: the denotation of HPA20/A5
The glycerol dehydration tests were realised in a continuous
flow apparatus as shown in Fig. 1 with a glass tube reactor (length
14 cm, volume 35 cm3) equipped with an internal guide tube for a
thermocouple inside the catalyst bed. The reactor was placed in an

H. Atia et al. / Journal of Catalysis 258 (2008) 71–82
73
Fig. 1. Flow scheme of continuous bench scale set-up.
electrically heated furnace with an internal diameter matching the
2.4. Analytical methods and evaluation of results
outer diameter of the reactor to optimise heat transfer and tem-
perature profile. The particle size of the solids tested was in the
range from 315 to 500 μm. Besides the prepared supported cat-
alysts, bulk heteropolyacids as well as pure supports with same
particle size were tested. Usually the catalyst was diluted with in-
ert quartz by a factor of ca. 20 to avoid hot spots. Measurement
of the axial temperature profile in the catalyst bed proofed that
the reaction was carried out at isothermal conditions. All reactant
The cold trap contained a measured amount of water to im-
prove absorption of low-boiling compounds and some hydro-
quinone was added to avoid polymerisation of acrolein. The aque-
ous samples were analysed without preliminary extraction or sepa-
ration of water using a gas chromatograph (Shimadzu GC 17A) with
autosampler equipped with a capillary column (Chrompack FFAP
30 m × 0.30 mm × 0.2 μm) and flame ionisation detector (FID,
◦
◦
◦
and product containing tubes were heated to at least 200 C to
temperature program: 40 C (4 min)–20 K/min–200 C (23 min);
carrier gas He; 1 μl injection volume; split ratio 80). Quantification
was realised with 5% by weight of 1-butanol in water as inter-
nal standard by mixing 1000 μl of non-diluted sample and 200 μl
of internal standard. Authentic compounds were used for product
calibration (acrolein and C1–C3 oxygenates).
avoid condensation, and these set-points were validated by mea-
suring the internal temperature in the tubes. A HPLC pump (Gilson)
was used to feed a mixture of glycerol (analytical grade) and water
(10% glycerol by weight) into the flow plant. Nitrogen served as di-
lutant and the flow rate was adjusted with a mass flow controller
◦
Gaseous samples were analysed with two separate gas chro-
matographs (Shimadzu GC 17A). The content of the gas collecting
cylinder was flushed into a 6-port sampling valve and after injec-
tion the permanent gases were eluted with argon as carrier gas
(Bronkhorst). The feed was first evaporated in a preheater (250 C)
and then entered the reactor in downflow mode. Products were
◦
collected first in a cold trap (0 C) that was connected directly
to the reactor outlet. The mass balance for each trap was deter-
mined to check proper operation of the set-up and the collected
liquids were analysed offline by means of GC. Volatile compounds
that were not retained in the cold trap were conducted through a
gas-collecting cylinder and also analysed by offline GC. Gas phase
analysis allowed quantification of permanent gases, volatile hydro-
carbons and volatile oxygenates.
ꢀꢀ
on a combination of two packed columns (PoraPak N, 3 m × 1/8 ;
ꢀꢀ
Molsieve 13X, 3 m × 1/8 ) and analysed by thermal conductiv-
◦
◦
ity detector (TCD, temperature program: 40 C–10 K/min–80 C–
◦
40 K/min–120 C). The second GC was equipped with a 10-port
sampling valve and two parallel analytical lines with He as carrier
◦
◦
gas and FID (temperature program: 50 C–35 K/min–200 C). One
line used a FFAP capillary column (50 m × 0.32 mm × 0.2 μm) for
analysis of volatile oxygenates. The second line used an Alumina
PLOT column (25 m × 0.32 mm) for analysis of volatile hydrocar-
bons.
After setting new reaction parameters, the system was run for
at least 1 h to establish steady-state operation. For each set-point,
usually two independent samples were collected in cold traps.
At the same time, also the corresponding gas samples were col-
lected at the outlet of the traps. The standard test program for
each catalyst comprised four consecutive temperature set-points at
A total organic carbon (TOC) analyser (Shimadzu VCPN) served
for quantification of total carbon content (organic and inorganic
carbon) in the liquid samples to crosscheck GC analyses.
The contributions of all GC analyses for liquid and gaseous
phase were joined to evaluate carbon as well as gas balance. Mo-
lar streams for all reactants were calculated from concentration in
cold trap, dilution, and sampling time. Calculation of glycerol con-
◦
225, 250, 275, and 300 C. All other parameters like catalyst mass
(300 mg), feed composition (glycerol mass concentration 10% by
weight), and flow rate (0.028 ml/min of aqueous feed, 30 ml/min
of nitrogen) were kept constant.

74
H. Atia et al. / Journal of Catalysis 258 (2008) 71–82
Fig. 2. DTA plots for non-supported and alumina-supported heteropolyacids:
(a) HSiW, (b) HPMo, (c) HPMo20/A5, (d) HSiW20/A5, (e) A5 (heating rate 10 K/min,
airflow 1 l/h, 5–20 mg of catalyst).
Fig. 3. DTA plots for non-supported and silica-supported heteropolyacids: (a) HSiW,
(b) HPMo, (c) HPMo20/S2, (d) HSiW20/S2, (e) S2 (heating rate 10 K/min, airflow
1 l/h, 5–20 mg of catalyst).
version (Xglycerol) based on the molar streams of glycerol at the
inlet and outlet of the setup as follows:
˙
˙
nglycerol,in − nglycerol,out
Xglycerol
=
× 100%.
˙
nglycerol,in
The selectivities for the products (Si) were calculated from the
molar streams of product and converted feed according to
˙
ni
zi
Si =
×
× 100%,
˙
˙
nFeed,0 − nFeed
zFeed
where z represents the number of carbon atoms in product i
and in the feed molecule (glycerol). The discussed selectivities
are therefore carbon-based numbers. Carbon balance is evaluated
from non-converted glycerol and yields of identified and calibrated
products, therefore carbon balance does not equate to 100%. More
details on product identification are given in Section 3.6.
3. Results and discussion
Fig. 4. DTA plots for HSiW supported on different supports: (a) HSiW,
(b) HSiW20/S11, (c) HSiW20/A12, (d) HSiW20/AS4, (e) AS4 (heating rate 10 K/min,
airflow 1 l/h).
3.1. Differential thermal analysis (DTA)
a small modification in the band around 900 cm−1. This region
is cited for the stretching vibration (υas Mo–Ob–Mo) that is sen-
sitive for dehydration process induced by thermal treatment [25].
DTA served to evaluate the thermal stability of the prepared
supported catalysts within the target temperature range of 200–
350 C for catalytic tests. The temperatures for thermal decomposi-
◦
◦
Maksimovskaya et al. reported that near 400 C HPMo12O39 trans-
tion of bulk HPMo, NH4PMo, HSiW and HPW reported in literature
◦
forms into MoO2HPO4 and MoO3 or [PMo12O38][PMo12O39] (so-
called anhydride 2) [26]. The formed molybdenum oxide is be-
ing assigned to β-MoO3 [25]. Therefore, the exothermic peak for
are 375, 400, 445 and 465 C, respectively [21]. Figs. 2–4 show
the DTA curves for alumina-, silica-, and aluminosilicate-supported
heteropolyacids, including reference data for bulk heteropolyacids
and pure supports.
◦
bulk phosphomolybdic acid at 444 C might be due to the crys-
◦
tallisation of β-MoO3. Above 450 C anhydride 2 decomposes to
The DTA curves shown in Figs. 2 and 3 for alumina (A5) and
◦
yield (MoO2)2(P2O7) and α-MoO3 [25], corresponding to the broad
silica (S2) supports gave one endothermic peak at 109 and 97 C
◦
◦
exothermic peak at 475 C. The exothermic peak at 520 C is due
to the crystallisation of the resulting α-MoO3 [25].
for each support. These peaks are due to physisorbed water.
The DTA curve (Fig. 2) for the bulk phosphomolybdic acid
◦
The DTA for bulk silicotungstic acid displays two endothermic
shows three endothermic peaks near 40, 73 and 116 C, which
◦
peaks at 73 and 208 C (Fig. 2); the first peak is due to the loss
are accompanied by a considerable weight loss caused by loss
of water of crystallisation [22,23]. No weight changes or ther-
mal effects were observed in the temperature range from 150
of physisorbed water, while the second endothermic peak is due
to the removal of water from this hydrated heteropolyacid [27].
Babad-Zakhryapin and Moiseev claimed that the removal of water
from this hydrate gives an endothermic effect within this temper-
◦
to 350 C. Hodnett and Moffat found that at higher temperature
◦
(≈400 C) water is formed from acidic protons and oxygen be-
◦
ature region (200–285 C) [28]. However, no important structural
longing to the Keggin units [23]. This type of the so-called con-
stitutional water amounts to ≈1.5 moles per Keggin unit. Mak-
simovskaya and Bondareva found the formation of the so-called
changes are observed in this temperature region. Anhydrous 12-
◦
silicotungstic acid is intact up to 400 C [29]. The exothermic peak
◦
◦
at 540 C is due to crystallisation of a new compound. This com-
anhydride 1 (HPMo12O39) at 380 C due to such loss of the consti-
tutional water [24]. Rocchiccioli-Deltcheff et al. carried out Raman
investigations after thermal treatment up to 380 C and observed
pound is assigned to WO3(Si) that could be considered as a solid
solution of silicon in tungsten trioxide [30].
◦

H. Atia et al. / Journal of Catalysis 258 (2008) 71–82
75
Similar to the bulk materials, the DTA plots for the alumina-
supported catalysts HSiW20/A5 and HPMo20/A5 show an en-
◦
dothermic peak due to loss of water (ca. 120 C) accompanied by
a weight loss in TGA (Fig. 2). These samples show an additional
◦
endothermic peak in a very narrow range of 460 to 475 C in-
dependent of the heteropolyacid compound and no appreciable
◦
weight change at temperatures up to 800 C is noticed. The same
behaviour was observed for supported heteropolyacids with 10%
loading. The appearance of the second endothermic peak might be
due to interaction between the support alumina and heteropoly-
acid.
All silica-supported samples show endothermic peaks around
◦
132 C due to loss of water (Fig. 3). In contrast to the alumina-
supported samples, an exothermic peak appears in the range from
◦
475 to 557 C that indicates the decomposition of the Keggin an-
ion. These peaks are broader and one can conclude that these data
mainly represent the influence of the thermal stability of different
heteropolyacids as they interact with the support weakly. It is con-
firmed by literature data that bulk HPW and HSiW heteropolyacids
are more stable than HPMo and NH4PMo. Immobilisation on silica
obviously increases the thermal stability of the Keggin structures
in comparison with their parent bulk heteropolyacids. Silica is a
well-suited support for HPA if compared to alumina [21,31–33]. On
the other hand, some research groups derived contradictory con-
clusions for heteropolyacid stability over silica [34,35].
Fig. 5. X-ray diffractograms for alumina-supported heteropolyacids.
Alumina seems to interact more strongly with the active com-
pounds than silica does. This interaction could be explained as
follows: when the solution and the support are in contact, the sur-
+
face hydroxyl groups of Al2O3 become protonated, yielding MOH2
(M is the metal ion). As a result, the surface gets a positive charge
−1
that can interact strongly with heteropolyanion (HPAn)
forming
M(HPAn) [36] that is being distorted. In this case, the HPA would
obviously loose acidity. Silanol OH groups may similarly participate
+
in HPA bonding to form (SiOH2 )(HPAn)−1. In dependence on the
nature of the support such interaction leads to the formation of
active intermediates with different strength of Brønsted acid sites
and stabilised structures. As a result, two effects may get com-
pensated: On silica, due to weak interaction mainly exothermic
decomposition of HPA accompanied by loss of constitutional water
is observed, whereas on alumina an additional stabilisation effect
due to stronger interaction between support and HPA occurs. This
interaction has to be resolved before thermal destruction of the
HPA occurs. If the first thermal effect is larger than the second one,
the overall effect will be inverted and endothermic peaks should
be observable. However, the structure of all supported HPAs seems
Fig. 6. X-ray diffractograms for silica-supported heteropolyacids.
ing silicotungstic acid on A12, S11 and AS4 resulted in materials
with surface areas of 250, 259 and 309 m2/g, respectively.
Lowering the loading of heteropolyacid from 20 to 10 wt% al-
ways led to increased surface area. This effect seems to be stronger
on silica than on alumina or aluminosilicate. The decrease in the
surface area may be attributed to wet impregnation method. The
data in Table 1 show that the silica-supported catalysts loose about
one third of their initial specific surface area during impregna-
tion. This is understandable when looking at the pore volume and
in particular at the average diameter of pores: the pores for the
alumina are more than double in size compared to silica contain-
ing materials. The latter contain more micropores, and first many
of them were plugged with the heteropolyacids as the Keggin
units are small enough (diameter ≈ 1.2 nm) to enter even microp-
ores [37].
◦
to be maintained at least up to 400 C, and hence this tempera-
ture should be suitable for calcination of the catalyst precursors at
least.
Fig. 4 shows DTA curves for catalysts with 20% HSiW supported
on alumina (pore diameter 12 nm), silica (11 nm) and aluminosil-
icate (4 nm) as well as for the bulk HSiW and support AS4. They
◦
reveal a single endothermic peak in the range from 73 to 116 C
due to loss of water. The aluminosilicate-supported catalyst pos-
◦
◦
sesses two endothermic peaks at 121 C and 490 C for loss of
water and the interaction between hydroxyl groups of aluminosili-
cate and the heteropolyanion.
3.3. X-ray diffraction (XRD)
3.2. Nitrogen adsorption
PMo,
The most intensive reflections for bulk heteropolyacids NH4
HPMo, HSiW, HPW are expected at 2θ = 26.5, 27.3, 8.0, and 10.3,
respectively. However, for alumina-supported heteropolyacids, the
reflections detected neither correspond to the heteropolyacids nor
to other crystalline phases even at a loading of 20% by weight and
hence the samples are completely X-ray amorphous (Fig. 5).
On the other hand, most of the diffractograms for hetero-
polyacids supported on silica S2 exhibit definite peaks (Fig. 6).
NH4PMo20/S2 shows reflections at 2θ = 26.4 and 10.7 that can
Textural properties (BET surface area, pore volume and average
pore diameter) for heteropolyacids immobilised on five different
supports derived from nitrogen physisorption isotherms are given
in Table 1. The specific BET surface areas for various heteropoly-
acids supported on alumina (A5) with 20% loading reached 249–
270 m2/g. The silica-supported heteropolyacids with 20% loading
had surface areas from 418 to 487 m2/g. Additionally, immobilis-

76
H. Atia et al. / Journal of Catalysis 258 (2008) 71–82
be assigned to the Keggin-type structure of (NH4)2.6(H3O)0.4
-
(PO4Mo12O36) (PDF 70-129). For HPMo20/S2, signals at 2θ = 27.3
and 23.3 indicate the presence of MoO3 phase (PDF 89-7112). The
formation of anhydrous heteropolyanion occurs in the temperature
◦
region up to 350 C. Then, the loss of constitutional water starts
◦
up to ca. 435 C. The HPA decomposes into mainly β-MoO3 with
small portion of α-MoO3. During heating to 500 C only α-MoO3
◦
can be observed [38]. Reflections of β-MoO3 and some of α-MoO3
◦
are clearly to see in the pattern of HPMo at 350 C [38].
In the diffractogram for HPW20/S2 the signals at 2θ = 10.3 and
25.3 belong to the Keggin-type H3PW12O40 phase (PDF 75-2125).
These data prove that molybdenum-containing heteropolyacids are
the least stable compounds in this group in coincidence with re-
sults from thermal analysis. As an exception, HSiW20/S2 shows no
reflections in the diffractogram at all, indicating that HSiW has
been dispersed even over the silica support in the same manner
as observed with alumina support (see below).
The diffractograms for silicotungstic acid (HSiW20) supported
on alumina A12, silica S11 and aluminosilicate AS4 are not mean-
ingful. HSiW20/S11 and HSiW20/A12 samples are X-ray amorphous
and show no crystalline phase. On the other hand, non-calcined
HSiW20/AS4 showed three humps at 2θ = 28.2, 38.6, and 49.2
that were assigned to Al2O3·H2O (boehmite). After calcination at
Fig. 7. Raman spectra for HPMo supported over alumina and silica (1 = bands from
glass tube).
However, alumina-supported samples are always found to be X-ray
amorphous.
◦
400 C, the intensity of these humps significantly decreases maybe
3.4. Raman spectroscopy
due to dehydroxylation as inferred from the endothermic peak in
DTA. Based on the size of the Keggin anion (diameter ≈ 1.2 nm)
and the available surface area of alumina A5 and aluminosilicate
AS4 (298 and 384 m2/g), theoretically a loading of about 80% of
HPA would saturate the surface. Therefore, 20% loading may lead
to species that either are highly dispersed or decomposed on the
surface. Kozhevnikov found that at higher loadings, 30–50% of HPA
is present as crystalline phase on silica with 200–300 m2/g but
at low loadings mainly fine dispersed species are formed. Obvi-
ously, X-ray detectable HPA crystal phase is developed on silica
surface only above 20% loading [21]. High concentration of the
active phase saturates the support surface and then some HPA
molecules could not access the support mesopores and crystallise
as separate phase that could be detected by XRD.
Raman technique is known to be well suited for observation
of Keggin structures [40] to prove the presence of HPA structures.
Fig. 7 represents Raman spectra for HPMo20/A5 and HPMo20/S2.
The Raman spectrum for bulk HPMo gives bands at 996, 983,
−1
882, 603, and 246 cm
which are assigned to symmetric (υs)
and asymmetric (υas) vibrations of terminal oxygen υs(Mo–Ot) and
υas(Mo–Ot), of corner shared bridged oxygen υs(Mo–Ob–Mo), of
edge shared bridged oxygen υs(Mo–Oc–Mo) and of oxygen in the
central tetrahedron υs(Mo–Oa) [38]. Comparing the Raman spec-
trum of HPMo20/S2 with bulk HPMo spectrum reveals bands at
−1
996, 981, 975, 818, 606 and 243 cm that were assigned to Keggin
3−
−1
heteropolyanion (PMo12O40
)
except the band at 818 cm that is
The surface coverage of HSiW on HSiW20/A5, HSiW20/A12,
HSiW20/S2, HSiW20/S11, HSiW20/AS4 was calculated according
to a literature method using the following equation: 0.2/(0.8 ×
MSiW × SBET), where MSiW is the molecular weight of SiW
(=3310.66 g/mol) loaded on the silica and SBET is the sur-
face area of the support, respectively [20]. The calculated sur-
face density values were 0.254 (HSiW20/A5), 0.258 (HSiW20/A12),
0.120 (HSiW20/S2), 0.235 (HSiW20/S11) and 0.197 (HSiW20/AS4)
μmol/m2, respectively. The low value for HSiW20/S2 indicates
higher dispersion over the silica and this could explain why only
HSiW20/S2 did not show any XRD pattern. As seen from the pre-
vious equation, surface density is inversely proportional to the
molecular weight of the HPA. Thus, as HSiW has higher molecular
weight compared with HPMo, silicotungstic acid would be more
dispersed over silica support as found by the XRD studies and this
could explain why HPMo and HPW showed reflections while HSiW
did not.
From these studies, it can be concluded that the interaction
of the heteropolyacids is much stronger with alumina than with
silica. A more reasonable explanation refers to the surface zeta
potential of the supports than aforementioned hydrogen bonds.
Alumina has a typical isoelectric point around pH 10 and silica
around pH 2 [39]. As discussed above, at the preparation con-
ditions in aqueous phase, alumina should be positively charged
and therefore strong interaction with the negatively charged het-
eropolyanions is possible. Silica surface is negatively charged and
hence the heteropolyanions rather tend to agglomerate. If this hap-
pens, the heteropolyacids can form X-ray detectable crystallites.
assigned to MoO3 formed by partly degradation of HPMo [38]. For
−1
HPMo20/A5, a broad band at 951 cm
is detected, which is re-
lated to the stretching vibrations of υs(Mo–Ob–Mo) groups and is
influenced by hydrolysis/dehydration processes [38,41]. Some NMR
studies could support these findings. Vázquez et al. [42] reported
in their studies on HPMo supported over alumina two lines present
at −3.25 and −10.76 ppm. The first chemical shift is assigned
to the Keggin unit and the second could not be strictly assigned
but probably it may be related to the strong HPMo interaction
with OH groups of alumina support. A similar shift (−10.76 ppm)
was observed by Rao et al. [32] that was assigned may be due
to HPMo interacting with the support surface or partially reduced
HPMo [43] or unidentified species. These species might be dimeric
6−
6−
[P2Mo21O71
]
resembling to [P2W21O71
]
that are formed by
3−
partial transformation of [PW12O40
]
supported on alumina [44].
Fig. 8 illustrates that the most intense bands in bulk HSiW
−1
−1
appear at 998 cm
(stretching of W=O), 974 cm
(bending of
−1
W–Oc–W), and 555 cm
are assigned to WO3, but could also stem from im-
purities. The spectrum of HSiW20/S2 shows peaks at 998 cm
(bending of O–Si–O). The signals at 329
−1
and 227 cm
−1
−1
and 970 cm for the W=O stretching and W–O–W bending mode,
indicating the preservation of the Keggin structure in the HSiW on
the surface of silica. The use of alumina as support for the same
heteropolyacid in HSiW20/A5 leads to a shift in W–O–W bend-
ing mode to 962 cm−1. Such a distortion effect within the Keggin
structure could evidence a strong interaction of HSiW with alu-
mina surface.

H. Atia et al. / Journal of Catalysis 258 (2008) 71–82
77
acidic sites. On the other hand, the overall acidity of the silica
surface significantly increased after impregnation with heteropoly-
acid as the amount of acidic sites over silica is larger compared to
alumina-supported sample. The weak interaction of silica with the
HPA keeps its Brønsted acid character and leads to an increased
◦
proportion of medium acid sites. Desorption peaks above 400 C in
the TPDA curves (registered by TCD) of the supported HPAs may
point to their destruction.
3.6. Product identification
Product analyses were carried out as mentioned above by using
several GC units. Acrolein is always the predominant product in
both liquid and gaseous phase besides some by-products in differ-
ent quantities (acetaldehyde, propionaldehyde, acetone, methanol,
ethanol, allyl alcohol, hydroxyacetone, acetic acid, 1,2-propandiol,
and propionic acid). Some minor by-products were not identified.
However, a defined portion of acrolein, acetaldehyde and propi-
onaldehyde escaped from the cold trap but were quantified from
the product gas in addition to liquid phase. Absence of O2 vali-
dated proper sampling and the absence of CO and CO2 indicated
that reforming of glycerol did not occur. Routinely evaluated mass
balances for each cold trap accounted for 92–98% (without gaseous
compounds). Carbon balances were calculated from detected prod-
ucts in gas and liquid phase for each set-point and added up to
60–95% (values can be calculated from Table 2). Missing carbon
is due to (i) some unknown minor peaks in chromatograms and
(ii) mostly deposits on the catalyst surface depending on the re-
action conditions (glycerol oligomers, polymers from acrolein, etc.)
as also indicated by GC–MS after extraction spent catalyst sam-
ples. Carbonaceous deposits on the spent catalyst are not included
in carbon balance due to uncertain quantification. Casually made
TOC analyses of carbon content in liquid samples revealed good
agreement with GC data.
Fig. 8. Raman spectra for bulk HSiW and HSiW supported on alumina and silica
(oven dried samples; HSiW400 = calcined at 400 C).
◦
3.7. Influence of temperature on conversion and selectivity
The catalytic tests with the pure heteropolyacids in the tem-
◦
perature range from 225 to 300 C showed that HSiW was the
◦
most active solid reaching 63% conversion at 275 C. No catalyst
Fig. 9. Profiles for TPD of ammonia for supports silica S2 and alumina A5 and cata-
lysts HSiW20/A5 and HSiW20/S2.
revealed complete conversion. HPMo deactivated rapidly and HPW
and NH4PMo showed a moderate increase in conversion with tem-
perature. There was no general trend observable for the four het-
eropolyacids and maximum selectivity for acrolein never exceeded
50%.
3.5. Temperature programmed desorption of ammonia (TPDA)
◦
Alumina support A5 calcined at 400 C shows much higher total
The tests with the silica-supported heteropolyacids showed that
these materials rapidly deactivate after few hours. The deactivation
could occur due to decomposition of the HPA active compound or
blocking of the catalyst surface by carbonaceous deposits (see be-
low). Activity of the silica-supported heteropolyacids decreases in
the order HSiW20/S2 > HPW20/S2 > NH4PMo20/S2 > HPMo20/S2.
Data for these catalysts are listed in Table 2. Increasing the pore
size of the support (from 2 to 11 nm) enhances the catalytic activ-
acidity from titration (0.34 mmol NH3/g) than silica (0.07 mmol/g),
as expected (Fig. 9). The maxima for ammonia desorption from
alumina and silica supports were observed at temperatures of ap-
◦
proximately 180 and 320 C, respectively. A small additional peak
◦
around 195 C is detected for alumina. The classification of strength
◦
◦
of acid sites is weak (150–300 C), medium (300–500 C), and
strong (500–650 C) [45].
◦
Modification of the supports with HPA seems to induce new
medium acidic sites at least. After impregnating the supports with
HSiW, the overall acidity measured by titration of HSiW20/A5 stays
unchanged (0.34 mmol/g), whereas the amount of adsorbed am-
monia rises from 0.07 to 0.44 mmol/g for the HSiW20/S2 catalyst.
The TPDA profile of HSiW20/A5 shows three peaks appearing at
◦
ity and hence the conversion is complete for HSiW20/S11 at 275 C.
Nevertheless, rapid deactivation was observed in the test with the
latter material.
Fig. 10 shows conversion with increase in temperature for
HSiW20/A5, HPMo20/A5, HSiW20/A12, as well as for the pure sup-
ports A5 and A12. In contrast to bulk and silica-supported cat-
alysts, all heteropolyacids supported on alumina A5 as well as
pure alumina A5 itself show a consistent increase in glycerol con-
version. No such heavy deactivation was observed as for silica-
supported samples. The surface specific rate of acrolein formation
on alumina-supported catalysts is approximately one order of mag-
nitude higher than for silica-supported samples (Table 2). At the
◦
approximately 180, 215 and 310 C. The TPDA curve of HSiW20/S2
◦
reveals only two peaks at 180 and 320 C. The pure alumina sup-
port shows much more total acidity compared to the silica, there-
fore, the HPA anion is expected to interact to a large extent. In this
case, a portion of the HPA would loose its acidity due to the dis-
tortion of the Keggin structure. The addition of heteropolyacid did
not introduce any new sites, but led to a replacement by other

78
H. Atia et al. / Journal of Catalysis 258 (2008) 71–82
Table 2
Results from glycerol dehydration with alumina (A5)- and silica (S2)-supported catalysts
T
[ C]
X
[%]
Y (known
products) [%]
Rate of acrolein
formation [μmol/(m2 h]
S (individual products)
Acetone [%] Acrolein [%]
HPMo20/A5
◦
Ethanol [%]
Allyl alcohol [%]
Hydroxyacetone [%]
Total [%]
225
250
275
300
60.5
83.5
97.3
98.0
21.2
42.5
86.9
84.7
0.15
0.38
1.04
1.04
8.8
8.5
6.5
7.2
10.6
19.1
45.0
43.8
3.8
4.0
2.3
2.4
6.7
6.9
8.2
6.4
3.0
6.4
10.9
5.2
35.0
50.9
89.3
85.1
NH4PMo20/A5
225
250
275
300
47.6
71.5
95.6
99.6
17.2
40.6
66.3
50.4
0.15
0.43
0.79
0.33
7.0
4.5
4.4
2.5
14.3
27.1
37.1
14.9
4.4
2.7
2.2
2.8
4.1
7.5
7.0
5.1
3.3
6.7
6.4
4.4
36.2
56.8
69.4
50.6
HPW20/A5
19.8
225
250
275
300
17.3
72.7
99.2
97.4
16.8
37.5
76.8
58.9
0.08
0.38
1.14
0.47
40.3
15.2
4.3
22.1
3.7
2.4
3.5
0.6
0.8
1.8
1.4
5.2
4.2
8.0
6.1
96.9
51.5
77.4
60.5
23.5
51.7
21.7
4.4
HSiW20/A5
14.3
225
250
275
300
22.9
74.2
98.4
99.5
4.9
26.3
83.9
83.1
0.08
0.49
1.50
1.41
2.7
1.2
2.6
2.0
0
0
0
21.5
35.4
85.3
83.5
27.4
63.6
58.9
0.5
1.6
1.2
0.6
1.5
1.2
3.2
7.9
4.3
HPMo20/S2
4.3
225
250
275
300
47.0
21.8
8.9
10.4
14.2
6.6
0.03
0.07
0.03
0.15
7.1
12.2
14.0
1.7
6.1
8.6
8.9
1.3
1.7
8.0
5.9
3.8
1.1
5.6
7.8
5.5
22.1
65.0
73.7
38.6
23.9
27.8
19.8
55.7
21.5
NH4PMo20/S2
225
250
275
300
43.8
34.1
48.9
70.7
12.2
21.1
29.7
40.1
0.04
0.09
0.15
0.19
10.5
18.8
10.7
5.5
7.4
20.8
24.4
22.3
5.7
10.0
5.7
1.5
4.3
7.0
8.7
1.2
3.3
5.2
6.6
27.8
62.2
60.7
56.7
2.6
HPW20/S2
21.6
225
250
275
300
83.6
52.9
55.5
67.1
26.6
21.5
23.9
27.6
0.22
0.19
0.19
0.24
5.4
5.4
6.6
3.0
2.8
2.6
3.4
1.4
0.1
0.1
0.2
0.4
0.7
1.1
2.0
3.0
31.8
40.5
43.2
41.1
29.9
28.0
29.1
HSiW20/S2
5.0
225
250
275
300
65.1
70.0
79.2
64.8
5.4
11.9
23.8
36
0.05
0.13
0.24
0.36
1.4
1.2
3.3
3.3
0.9
1.2
2.5
2.1
0.2
0.1
0.1
0.3
0.3
0.8
1.4
4.4
8.2
16.9
30.0
36.0
12.7
21.0
39.0
Note. Catalyst mass = 300 mg, glycerol mass concentration = 10 wt%, flow rate (liquid) = 0.028 ml/min, 30 ml/min of nitrogen, T = reaction temperature, X = conversion,
Y = yield, S = selectivity.
chosen modified contact time of 0.15 kg h mol−1, with most cat-
◦
alysts complete conversion was obtained near 275 C. Comparison
of HSiW20/A5 with HSiW20/A12 shows that increasing the support
pore size enhances the catalytic activity as already observed for
silica-supported catalysts. HSiW20/AS4 catalyst shows similar per-
formance as alumina-supported catalysts, reaching maximum con-
◦
version at 275 C without noticeable deactivation at higher tem-
perature, too. Based on these results, it can be concluded that the
nature of the carrier has a significant influence on the catalyst ac-
tivity and its lifetime.
Considering the supports, the catalyst activity as well as life-
time seem to decrease in the order alumina ꢀ aluminosilicates >
silica. At the same time, the results also demonstrate that im-
pregnation of all kinds of heteropolyacids onto supports leads to a
significantly increased performance compared to bulk heteropoly-
acids. Supporting heteropolyacids causes the dispersion of HPA as
found by XRD, in particular for alumina-supported samples. The
dispersion leads to an increase in the total amount of catalytically
active protons. The use of these dispersed catalysts permits a bet-
Fig. 10. Influence of temperature on glycerol conversion for alumina and alumina-
ter accessibility of glycerol molecules to active sites and thus high
conversions can be obtained. In view of this, HSiW20/A5 showed
supported heteropolyacids (10% of glycerol in water, W /F = 0.15 kg h mol−1).

H. Atia et al. / Journal of Catalysis 258 (2008) 71–82
79
to the presence of higher number of Keggin protons of HSiW
(four) as compared to HPW or HPMo (three) [46]. Varisli et al.
found that HSiW is more stable than HPW at temperatures higher
◦
than 200 C in dehydration of ethanol [47]. On the other hand,
phosphomolybdic compounds always showed the lowest selectivity
for acrolein whereas selectivity for acetaldehyde, propionaldehyde,
acetone, and acetic acid increased. These results might be due to
the redox properties introduced by molybdenum that enables ad-
ditional reaction pathways at the cost of acrolein formation [20].
Furthermore, it seems that pores larger than 4 nm lead to more
effective catalysts because they provide sufficient space for Keg-
gin units on the inner surface and their interaction with adsorbed
glycerol molecules. If a large proportion of the heteropolyacids was
fixed in mesopores (2 to 12 nm in this study), the formation of
crystallites with sufficient size to generate XRD pattern could be
hindered. This possibly explains the absence of XRD patterns for
the alumina containing catalyst that contain a large portion of
mesopores. On the other hand, for materials with a large por-
tion of micropores (silica), the heteropolyacid compounds mainly
are located on the outer surface and are not restricted in form-
ing crystallites. In addition, the influence of internal mass transfer
limitation will be reduced with larger pores [20].
Although heteropolyacids are probably partially decomposed
over the alumina surface, they show high catalytic activity and
selectivity. Alejandre et al. found that the surface tungsten Lewis
acidic centres are more acidic on alumina than on other sup-
ports [48]. IR experiments revealed that the presence of such tung-
sten oxide species induces Brønsted acidity. At catalytic reactions
where water vapour is frequently present at relatively high tem-
perature (≈573 K), it probably hydrates a portion of the surface
forming Brønsted acidity centres. So the Lewis acidity corresponds
to the anhydrous form and Brønsted acidity to the hydrated form.
Such hydration could happen to HPMo but due to its low ther-
mal stability and additional redox properties, its catalytic activity
is lower than HSiW.
Fig. 11. Acrolein selectivity versus conversion over alumina supports and alumina–
supported HPAs (10% of glycerol in water, W /F = 0.15 kg h mol−1).
higher activity than HSiW20/S2 (see also Table 2). However, the ef-
fect of changing the HPA mass fraction on the described supports
from 10 to 20% on catalytic conversion seems to be negligible. This
observation is valid for alumina as well as silica-supported sam-
ples.
The most prominent benefit of using these supports rises from
the significant impact on product distribution. With catalysts sup-
ported on silica S2 (pore diameter 2 nm), the plots of selectivity
versus glycerol conversion did not lead to reasonable results due
to the rapid deactivation, probably sampling in shorter time dis-
tances could give more reliable data but this could not be realised
due to the given sampling procedure (time interval of ca. 1 h
used). Nevertheless, some of these data are listed in Table 2, too.
HSiW as active compound is superior to the other S2-supported
catalysts with regard to acrolein formation. A maximum selectiv-
ity for acrolein of 39% at 63% glycerol conversion was achieved
for HSiW20/S2. Usage of HSiW20/S11-supported catalyst boosts the
acrolein yield dramatically as selectivity rises to 73% at 67% con-
version.
3.8. Influence of contact time on conversion and selectivity over
H4SiW12O40
HSiW is the most effective heteropolyacid while alumina and
aluminosilicate are superior supports compared to silica in terms
of better performance. In order to exploit such potential effects, the
influence of combination of these materials on the performance
was investigated more in detail by varying the residence time
The pure alumina supports A5 and A12 themselves showed a
considerably high catalytic activity in the studied reaction, prob-
ably due to their acidic properties. However, the selectivity for
acrolein on these materials (A5: 30% at complete conversion; A12:
40% at 93% conversion) was significantly less than for supported
HPA (Fig. 11). Among the four catalysts supported on alumina with
5 nm pore diameter, HSiW20/A5 was the most selective catalyst
as it gives 64% selectivity for acrolein at complete conversion at
◦
at a constant temperature of 275 C (Fig. 12). With HSiW20/AS4,
HSiW20/A12 and HSiW20/A5 the catalytic activity increases con-
sistently with residence time and complete conversion is reached
at approximately 0.2 s (Fig. 12a). Acrolein selectivity on these cat-
alysts seems to pass through a maximum by further increase of
residence time (Fig. 12b). This might be due to consecutive reac-
tions like acrolein degradation or polymerisation. A change in the
colour of spent catalysts indicated the formation of deposits on the
catalyst surface.
◦
275 C. To simplify the diagram only data for selected catalysts
were plotted. Changing the pore size from 5 to 12 nm further
enhances the selectivity of acrolein formation. In general, the se-
lectivity for acrolein increases with glycerol conversion and con-
tact time, respectively. This clearly indicates a consecutive reaction
pathway.
The maximum acrolein selectivity reached 74% with HSiW20/
AS4 (X = 99%), 71% with HSiW20/A12 (X = 91%) and 51% with
HSiW20/A5 (X = 100%). Obviously, HSiW20/AS4 was the best per-
HSiW on aluminosilicate leads to high acrolein selectivity
of 67% at nearly complete conversion similar to the alumina-
supported heteropolyacids. As observed for all other catalysts, the
◦
forming catalyst at 275 C. The data also clearly suggest that in-
◦
maximum selectivity is obtained at 275 C that seems to be indeed
crease in the pore size changes the product distribution, improves
lifetime and increases acrolein yield. An enlargement of the mean
pore diameter of the support alumina from approximately 5 to
12 nm leads to a gain of approximately 20% in acrolein selectiv-
ity. Though HSiW20/AS4 has a comparatively low pore diameter
of 4 nm, it still keeps its stability and activity. Long-term tests
with different alumina and aluminosilicate supports in steam at-
an optimum temperature with regard to acrolein formation.
In comparison to all other tested heteropolyacids, HSiW al-
ways led to the highest acrolein selectivity whether it was used
as bulk material or impregnated onto supports. The outstanding
performance of HSiW is linked to high acidity and hydrolytic sta-
bility even in aqueous environment without forming acetals [21].
Verhoef et al. reported that supported silicotungstic acid is more
active than tungstophosphoric acid in esterification reactions due
◦
mosphere at 240 C revealed an increase in hydrothermal stability
for aluminosilicate compared to aluminas [49]. The improved sta-

80
H. Atia et al. / Journal of Catalysis 258 (2008) 71–82
Fig. 12. (a) Conversion and (b) acrolein selectivity versus residence time with
HSiW20/A12, HSiW20/A5 and HSiW20/AS4 (275 C, 0.6 g of catalyst).
Fig. 13. Long-term runs with HSiW20 supported on different materials. (a) Conver-
◦
◦
sion, (b) acrolein selectivity versus time-on-stream (275 C, 0.15 kg h mol−1).
bility of the textural properties of aluminosilicates like support AS4
(Siral® 10) may compensate the lower diameter of the pores. Be-
sides that, may be a pore size of 4 nm is some kind of a threshold
value as reported by Tsukuda et al. [20], who found that silica with
3 nm was the least active support.
3.9. Long-term stability of H4SiW12O40 catalysts
The above discussed experiments with bulk and silica-supported
HPAs at standard reaction conditions under variation of tempera-
ture led to rapid deactivation, either due to heavy coking or also
probably due to structural changes in the active phase. On the
other hand, the alumina- and aluminosilicate-supported samples
were apparently stable during these tests. Spent silica-containing
samples were deeply black supporting a deactivation by coking,
whereas spent alumina- and aluminosilicate-supported catalysts
usually changed to slight brown only. This suggests a deactiva-
tion mechanism due to the blocking of the solid surface with
high-boiling compounds. The oligomerisation of glycerol or non-
saturated products like acrolein is a possible route towards such
deposits. Both reactions are definitely connected to the catalyst
acidity and pore size.
Therefore, alumina- and aluminosilicate-supported HSiW cat-
alyst samples (HSiW20/A5, HSiW20/A12 and HSiW20/AS4) were
checked for their long-term performance without interruption to
evaluate deactivation. The course of conversion and selectivity is
shown in Figs. 13a and 13b. Initial conversion of glycerol with
all these catalysts was complete, but decreased rapidly to ap-
proximately 50% within 75 h on-stream with HSiW20/A5. Both
the other catalysts HSiW20/A12 and HSiW20/AS4 showed signifi-
cantly higher stability as they lost less than 20% of activity within
200 h. After 300 h on-stream, HSiW20/AS4 still showed a conver-
sion above 70%. At the same time, acrolein selectivity was nearly
constant for all the catalysts. Indeed the product distribution did
not show any significant change during the complete test. This be-
haviour points at a deactivation mechanism forming deposits on
the surface that reduces the total number of accessible active sites
whereas the nature of the active sites that controls the selectivity
remains unchanged.
Fig. 14. DTA for spent catalysts HSiW20/A12, HSiW20/S11, HSiW20/AS4 (heating rate
10 K/min, airflow 1 l/h, 5–20 mg of catalyst).
of initial catalyst activity is possible easily by in situ oxidative re-
generation.
3.10. Characterisation of spent catalysts
So far it was not possible to clarify the nature of the de-
posits directly on the surface of the catalysts, e.g. by means of
spectroscopic methods. Nevertheless, carbonaceous deposits were
formed over the surface of the catalysts during the catalytic reac-
tion. The results from thermal analysis TG/DTA for spent catalysts
HSiW20/A12, HSiW20/S11 and HSiW20/AS4 are shown in Fig. 14.
Usually a weight loss of approximately 10% was measured but not
included in the C-balance. A precise quantification was not car-
ried out due to the operational mode (e.g. non-stationary regime,
changes in reaction temperature, and varying measurement time).
The DTA curves of the spent catalysts show two exothermic peaks
◦
near 350 and 450 C that can be assigned to two different types of
◦
deposits. For HSiW20/S11, the peak at 450 C is larger, indicating
Finally, a regeneration test was carried out with HSiW20/AS4
after conversion decreased from 100 to 75% within 300 h. Nitro-
gen flow was replaced by a mixture of 99% of N2 and 1% of O2 and
temperature was raised to 325 C for 24 h. After switching back
to standard reaction conditions, conversion recovered to 100% and
acrolein selectivity reached 60%. This demonstrates that recovery
that more thermal stable deposits are formed. On the other side,
for HSiW20/AS4 and HSiW20/A12 the less thermal stable deposit
species represented by the peak at 350 C are predominant. Hence,
it should be possible to regenerate those catalysts more easily than
silica-supported materials. Besides that, there are also hints in lit-
erature that porosity plays a role in deactivation, especially on sil-
◦
◦

H. Atia et al. / Journal of Catalysis 258 (2008) 71–82
81
ica that usually contains measurable amounts of micropores [20].
Furthermore, acidic properties seem to have a significant effect,
and silica-supported heteropolyacids showed increased number of
acidic sites.
During the long-term test with HSiW20/AS4 also changes in
the textural properties of the surface occur. The specific surface
area drops from 308 to 166 m2/g (46% loss) and the pore volume
decreases from 0.43 to 0.28 cm3 (35% loss). This observation also
proves the blocking of the surface by deposits.
of Brønsted sites, i.e. hydroxyl groups that also seem to play a
role for dispersion of the heteropolyacids on the surface of the
support during preparation. However, higher acidity seems to ac-
celerate deactivation of the catalyst. Alumina- and aluminosilicate-
supported heteropolyacids were more active than silica-supported
catalysts and bulk heteropolyacids. This could be due to the strong
interaction between the heteropolyacids and the support surface.
A strong interaction possibly leads to high dispersion of the HPAs
on the surface rather than agglomeration and formation of HPA
crystallites.
4. Summary and conclusions
With all tested catalysts, acrolein was always the predominant
product. Generally, all tungsten containing catalysts showed out-
standing performance with regard to activity and selectivity com-
pared to Mo and P containing materials. This can be explained
by the additional redox capability introduced to the catalysts by
molybdenum. The maximum selectivity for acrolein reached 75%
DTA measurements proved that the bulk HPAs are stable up
◦
to approximately 400 C. Therefore, all supported catalysts were
calcined at this temperature. When supporting the heteropoly-
acids over alumina, silica, and aluminosilicate, the thermal stabil-
ity is even increased. The decomposition of bulk heteropolyacids
showed exothermic peaks mainly due to loss of water. Alumina-
supported heteropolyacids showed additional endothermic peaks
indicating strong interaction between active compound and sup-
port that overweighs the previous effect. On the other side, silica-
supported heteropolyacids still showed exothermic peaks due to
weak interaction.
XRD results showed that reflections are only obtained on silica-
supported materials where the interaction is low and permits the
agglomeration of larger crystallites. Molybdenum containing solids
seemed to be less thermal stable than tungsten containing com-
pounds as XRD showed partial decomposition forming molybde-
num oxide. Raman spectra also inferred the partial decomposition
of molybdenum containing heteropolyacids on alumina as well as
over silica. Polymolybdates and molybdenum oxide were identified
as decomposition products. On the other hand, for tungsten con-
taining supported catalysts no new Raman bands appeared, which
indicates higher stability than for molybdenum containing com-
pounds.
◦
at complete conversion on HSiW/AS4 at 275
C. The reforming of
glycerol as a possible side reaction at such conditions was com-
pletely suppressed.
With regard to application, it has to be pointed out strongly
that the maximum selectivity for acrolein is obtained at complete
conversion and hence a favourable single pass conversion is possi-
ble. Long-term tests up to 300 h on-stream were carried out. It is
also beneficial that the catalysts could be regenerated easily in situ
with a small amount of oxygen in the feed at moderately elevated
temperature within a short time period.
Acknowledgments
The authors wish to thank Dr. U. Bentrup for Raman studies,
Dr. M. Schneider for XRD investigations, Mrs. S. Evert for ammo-
nia TPD analysis and Mr. R. Eckelt for TG/DTA data and nitrogen
adsorption measurements. Financial support by the Egyptian Min-
istry of Higher Education is gratefully acknowledged.
BET data revealed that impregnation of heteropolyacids on the
surface leads to a drop in specific surface area due to plugging
of the pores. Specific surface area of the silica-supported cata-
lysts is higher, and these silica materials have a significant amount
of micropores. Alumina- and aluminosilicate-supported heteropoly-
acids are rather mesoporous. In general, alumina support was more
acidic than silica support. After impregnating these supports with
heteropolyacids the acidity of the resulting catalysts reached simi-
lar values.
Four crucial parameters are responsible for the performance of
heteropolyacids in the dehydration of glycerol in gas phase: the
specific surface area, pore size of the support, surface acidity, and
nature of the heteropolyacid.
References
[1] http://www.theglycerolchallenge.org/.
[2] A. Corma, S. Iborra, A. Velty, Chem. Rev. 107 (2007) 2411.
[3] M. McCoy, Chem. Eng. News 84 (2006) 7.
[4] P. McMorn, G. Roberts, G.J. Hutchings, Catal. Lett. 63 (1999) 193.
[5] S. Carrettin, P. McMorn, P. Johnston, K. Griffin, C.J. Kiely, G.J. Hutchings, Phys.
Chem. Chem. Phys. 5 (2003) 1329.
[6] W.C. Ketchie, M. Murayama, R.J. Davis, Top. Catal. 44 (2007) 307.
[7] A. Abbadi, H. van Bekkum, Appl. Catal. A 148 (1996) 113.
[8] J. Deutsch, A. Martin, H. Lieske, J. Catal. 245 (2007) 428.
[9] T. Miyazawa, S. Koso, K. Kunimori, K. Tomishige, Appl. Catal. A 318 (2007) 244.
[10] J. Barrault, J.-M. Clacens, Y. Pouilloux, Top. Catal. 27 (2004) 137.
[11] M. Richter, Y.K. Krisnandi, R. Eckelt, A. Martin, Catal. Commun. 9 (2008) 2112.
[12] H. Adkins, W.H. Hartung, Org. Synth. Coll. 1 (1941) 15.
[13] A. Neher, T. Haas, D. Arntz, H. Klenk, W. Girke, U.S. Pat. 5 387 720 (1995).
[14] E. Schwenk, M. Gehrke, F. Aichner, U.S. Pat. 1 916 743 (1933).
[15] S. Ramayya, A. Brittain, C. DeAlmeida, W. Mok, M.J. Antal, Fuel 66 (1987) 1364.
[16] L. Ott, M. Bicker, H. Vogel, Green Chem. 8 (2006) 214.
[17] M. Watanabe, T. Iida, Y. Aizawa, T.M. Alda, H. Inomata, Bioresour. Technol. 98
(2006) 1285.
[18] S. Chai, H. Wang, Y. Liang, B. Xu, Green Chem. 9 (2007) 1130.
[19] S. Chai, H. Wang, Y. Liang, B. Xu, J. Catal. 250 (2007) 342.
[20] E. Tsukuda, S. Sato, R. Takahashi, T. Sodesawa, Catal. Commun. 8 (2007) 1349.
[21] I.V. Kozhenikov, Chem. Rev. 98 (1998) 171.
[22] J.-M. Tatibouët, C. Montalescot, K. Brückman, J. Haber, M. Che, J. Catal. 169
(1997) 22.
[23] B.K. Hodnett, J. B Moffat, J. Catal. 88 (1984) 253.
[24] R.I. Maksimovskaya, V.M. Bondareva, Zh. Neorg. Khim. [Russ. J. Chem., Engl.
Transl.] 39 (1994) 1298.
[25] C. Rocchiccioli-Deltcheff, A. Aouissi, M.M. Bettahar, S. Launay, M. Fournier,
J. Catal. 164 (1996) 16.
[26] R.I. Maksimovskaya, G.M. Maksimov, G.S. Litvak, Russ. J. Chem. Bull. Int. Ed. 25
(2003) 103.
[27] L.R. Pizzio, M.N. Blanco, Microporous Mesoporous Mater. 103 (2007) 40.
[28] A.A. Babad-Zakhryapin, V.P. Moiseev, Izv. Akad. Nauk SSSR, Ser. Khim. 5 (1964)
799.
At this stage it can be concluded first that impregnation onto
supports generally strengthens the thermal stability of the het-
eropolyacids (depending on the availability of surface hydroxyl
groups and zeta potential of the support). Immobilisation makes
◦
them effective catalysts for dehydration up to 300 C. Expanding
the specific surface area leads to a significant increase of active
sites and therefore of catalyst activity. A strong dispersion effect
on alumina containing materials was observed and these catalysts
seem to deactivate more slowly compared to silica-supported ones.
The reaction also runs on the bulk heteropolyacids and the bulk
carriers, but desired product selectivity and stability are rather
poor. Only supported catalysts lead to promising results. However,
catalysts with a mass fraction of 20% by weight showed no advan-
tage over samples with 10% loading. The pore size also has a strong
impact on activity and selectivity. To a certain extent, the enlarge-
ment of the pores leads to an increase in performance. Shape se-
lectivity is not expected to influence the conversion of glycerol. The
surface acidity is mainly connected to the number and strength

82
H. Atia et al. / Journal of Catalysis 258 (2008) 71–82
[29] A.A. Babad-Zakhryapin, Izv. AN SSSR, Otd. Khim. N. (1963) 215.
[30] A.A. Babad-Zakhryapin, N.S. Gorunov, Izv. AN SSSR, Otd. Khim. N. (1963) 14.
[31] J.B. Mofatt, S. Kasztelan, J. Catal. 109 (1988) 206.
[32] K.M. Rao, R. Gobetto, A. Iannibello, A. Zecchina, J. Catal. 119 (1989) 512.
[33] C. Hu, Y. Zhang, L. Xu, G. Peng, Appl. Catal. A: Gen. 177 (1999) 237.
[34] R. Fricke, G. Öhlmann, J. Chem. Soc. Faraday Trans. I 82 (1986) 263.
[35] C. Rocchiccioli-Deltcheff, M. Amirouche, G. Hervé, M. Fournier, M. Che, J.M. Tat-
ibouët, J. Catal. 126 (1990) 591.
[36] Y. Wu, X. Ye, X. Yang, X. Wang, W. Chu, Y. Hu, Ind. Eng. Chem. Res. 35 (1996)
2546.
[37] A. Popa, V. Sasca, E.E. Kis, R.M. Neducin, M.T. Bokorov, J. Halasz, J. Optoelectron.
Adv. Mater. 7 (2005) 3169.
[40] A. Brückner, G. Scholz, D. Heidemann, M. Schneider, D. Herein, U. Bentrup, M.
Kant, J. Catal. 245 (2007) 369.
[41] C. Rocchiccioli-Deltcheff, M. Fournier, R. Franck, R. Thouvenot, Inorg. Chem. 22
(1983) 207.
[42] P.G. Vázquez, M.N. Blanco, C.V. Cáceres, Catal. Lett. (1999) 205.
[43] M. Pope, Heteropoly and Isopoly Oxometallates, Springer, Berlin, 1983.
[44] E. Caliman, J.A. Dias, S.C.L. Dias, A.G.S. Prado, Catal. Today 107 (2005) 816.
[45] J.C. Yori, J.M. Grau, V.M. Benítez, J. Sepúlveda, J. Appl. Catal. A: Gen. 286 (2005)
71.
[46] M.J. Verhoef, P.J. Kooyman, J.A. Peters, H. Bekkum, Microporous Mesoporous
Mater. 27 (1999) 365.
[47] D. Varisli, T. Dogu, G. Dogu, Chem. Eng. Sci. 62 (2007) 5349.
[48] A.G. Alejandre, P. Castillo, J. Ramírez, G. Ramis, G. Busca, J. Appl. Catal. A:
Gen. 216 (2001) 181.
[38] S. Damyanova, J.L.G. Fierro, Chem. Mater. 10 (1998) 871.
[39] M. Kosmulski, Chemical Properties of Material Surfaces, Marcel Dekker, New
York, 2001.
[49] A. Martin, F. Luck, U. Armbruster, L. Patria, J. Radnik, M. Schneider, Top.
Catal. 33 (2005) 155.