Gas phase dehydration of glycerol to acrolein: Coke on WO3/TiO2 reduces by-products

Article abstract of DOI:10.1016/j.molcata.2016.05.022

Glycerol is a renewable feedstock for specialty chemicals that is co-produced when oil and fats are transesterified to biodiesel or hydrolysed to fatty acids. Acrolein is a target specialty chemical but catalysts deactivate with time due to coke and thus require frequent regeneration cycles to maintain activity. WO₃/TiO₂ catalyst dehydrated glycerol for 14 h while the acrolein selectivity increased as coke formed on the catalyst. After the first hour of reaction the acrolein selectivity was 36% and it reached steady state after 6 h at a selectivity of 73%. The major by-products were acetone, propanaldehyde, acetaldehyde, acetol and formic acid. Coke selectivity dropped from 50% in the first hour to 9% after 14 h. Elemental analysis (CHNS/O) of the catalyst withdrawn from the reactor during the experiments confirmed that polyaromatics formed on the surface. Based on NH₃ and SO₂ adsorption calorimetry the number of strong acid sites remained constant during 14 h, while the number of the weak and medium acid sites dropped. The basic sites disappeared after 2 h; coincidentally, less by-products (acetone) formed. Samples treated with pyridine, were analysed by FT-IR and they had fewer Br?nsted sites but no Lewis acid sites. Carbon deposited on walls of the cylindrical pores (based on Barrett-Joyner-Halenda method) and reduced the pore diameter.

Full text of DOI:10.1016/j.molcata.2016.05.022

Journal of Molecular Catalysis A: Chemical 421 (2016) 146–155
Contents lists available at ScienceDirect
Journal of Molecular Catalysis A: Chemical
journal homepage: www.elsevier.com/locate/molcata
Gas phase dehydration of glycerol to acrolein: Coke on WO /TiO
3
2
reduces by-products
a,∗
a
a
b
c
Marjan Dalil , Davide Carnevali , Mahesh Edake , Aline Auroux , Jean-Luc Dubois ,
Gregory S. Patiencea
a
Department of Chemical Engineering, Polytechnique Montréal, C.P. 6079, Succ. CV Montréal, H3C 3A7 Québec, Canada
Université Lyon 1, CNRS, UMR 5256, IRCELYON, Institut de recherches sur la catalyse et l’environnement de Lyon 2, avenue Albert Einstein,
b
F-69626 Villeurbanne, France
c
ARKEMA, 420 Rue d’Estienne d’Orves, 92705 Colombes, France
a r t i c l e i n f o
a b s t r a c t
Article history:
Glycerol is a renewable feedstock for specialty chemicals that is co-produced when oil and fats are trans-
esterified to biodiesel or hydrolysed to fatty acids. Acrolein is a target specialty chemical but catalysts
deactivate with time due to coke and thus require frequent regeneration cycles to maintain activity.
WO3/TiO2 catalyst dehydrated glycerol for 14 h while the acrolein selectivity increased as coke formed
on the catalyst. After the first hour of reaction the acrolein selectivity was 36% and it reached steady state
after 6 h at a selectivity of 73%. The major by-products were acetone, propanaldehyde, acetaldehyde,
acetol and formic acid. Coke selectivity dropped from 50% in the first hour to 9% after 14 h. Elemental
analysis (CHNS/O) of the catalyst withdrawn from the reactor during the experiments confirmed that pol-
yaromatics formed on the surface. Based on NH3 and SO2 adsorption calorimetry the number of strong
acid sites remained constant during 14 h, while the number of the weak and medium acid sites dropped.
The basic sites disappeared after 2 h; coincidentally, less by-products (acetone) formed. Samples treated
with pyridine, were analysed by FT-IR and they had fewer Brønsted sites but no Lewis acid sites. Carbon
deposited on walls of the cylindrical pores (based on Barrett–Joyner–Halenda method) and reduced the
pore diameter.
Received 29 February 2016
Received in revised form 10 May 2016
Accepted 20 May 2016
Available online 30 May 2016
Keywords:
Glycerol
Dehydration
Acrolein
WO3–TiO2
Coke
Fluidized-bed
©
2016 Elsevier B.V. All rights reserved.
1
. Introduction
the commercial process for production of acrolein is selective oxi-
dation of propylene in the gas phase over Bi/Mo-mixed oxide
catalysts [4].
Substitution of petroleum based feedstocks and chemicals with
bio-sustainable sources is a global priority. Biodiesel from veg-
etable oils, animal fats or cultured algae complement petro-diesel
transportation fuels [1–3]. Glycerol is a co-product of triglyceride
transesterification (biodiesel synthesis) and represents a mass frac-
tion of 10% of the total biodiesel produced. It is also an oleochemical
co-product of fatty acids obtained by hydrolysis of oils and fats. As
a consequence of the large scale production of biodiesel, the price
of crude glycerol has dropped since there was insufficient demand
to absorb the additional capacity.
The low price of glycerol makes it an attractive feedstock for
many specialty chemicals including acrolein. It is a large volume
intermediate for acrylic acid and has several industrial applications:
D,L-methionine, fragrances, polymers and detergents. Currently,
Schering Kahlbum AG dehydrated glycerol to acrolein in the gas
phase over lithium-phosphate and copper-phosphate catalysts in
1933 [5]. The acrolein yield reached 75% at temperatures between
◦
◦
300 C and 600 C. Zeolites, heteropolyacids (HPA), metal oxides
and supported mineral acids, dehydrate glycerol to acrolein in
either the gas phase or the liquid-phase. The selection of the best
catalyst depends on the surface area, pore size, acid strength and the
nature of the acid sites for a catalytic system. Acidity is the key fac-
tor to determine whether a catalyst is suitable to dehydrate glycerol
[6–12]. For the first time, Neher et al. [13] demonstrated the effect
of acidity on the catalyst performance for glycerol dehydration.
Heteropolyacids are good candidates for glycerol dehydration,
due to their high Brønsted acidity which approaches superacid
region (stronger acidity than pure H SO4 [14]). They are also
2
economical, environment friendly, and feature well defined struc-
tures and tunable acidity levels. The most common HPAs are HPW
∗ Corresponding author.
E-mail address: marjan.dalil@polymtl.ca (M. Dalil).
(
H PW12O40), H PW11VO40, HSiW (H SiW12O40), H SiMoO and
3 4 4 4 40
http://dx.doi.org/10.1016/j.molcata.2016.05.022
1
381-1169/© 2016 Elsevier B.V. All rights reserved.

M. Dalil et al. / Journal of Molecular Catalysis A: Chemical 421 (2016) 146–155
147
H PMo [14–16]. However, for catalytic applications, HPAs are
3
12
usually loaded on supported materials (such as TiO ) due to their
2
low specific surface area. The addition of support improves the sur-
face area of HPAs while their acidity remains high enough for the
desired reaction. Dubois et al. [17] patented the application of sup-
ported HPAs for the gas-phase dehydration of glycerol. The reaction
yield to acrolein was between 50% and 93% while the conversion of
glycerol was 79% and 100% [17,18].
The influence of acid/base properties of zirconia and titania
based catalysts on the selectivity of acrolein and by-products was
studied by Stosic et al. [19]. They observed that beside the acid sites,
the number of basic sites had a direct impact on acrolein selectiv-
ity in the gas-phase dehydration of glycerol. Therefore, in order to
increase the selectivity of acrolein, it is necessary to control not
only the strength and the amount of the acidic sites, but also to
hinder the number/strength/action of the basic sites. However, the
selectivity of acetol – as the major by-prodcut – was not correlated
to the acid–base properties.
Although very good catalysts are developed for the dehydration
of glycerol, the main disadvantage of current best catalysts is their
quick deactivation. More stable catalysts (69% acrolein yield after
2
4 h) have been developed by Katryniok et al. [20] but the glycerol
was highly diluted (10 wt%). Diluted feedstock require more energy
consumption for evaporation and a larger reactor would be needed
to yield in the same productivity. Erfle et al. [21] and Suprun et al.
Fig. 1. Schematic of the 46 mm ID fluidized bed, inlet gas manifold, quenches and
analytical.
[
22] reported that high reaction temperature, small catalyst pore
size and high acidity of the catalyst led to severe catalyst deacti-
vation. Erfle et al. [21] also analysed the spent catalyst by FT-IR
pyridine and detected cyclic anhydride compounds. The authors
stated that Brønsted sites deactivated the catalyst.
trapped in two quenches in series. A Varian HPLC (Metacarb 87H
column) measured the concentration of the condensables whilst a
Bruker GC equipped with Hyesep Q, Molsieve 5A and FFAP columns
analysed the acrolein and other light products e.g. acetaldehyde in
the gas phase. An online Pfeiffer mass spectrometer monitored the
permanent gases (Fig. 1).
The main objective of this study was to focus on the catalytic
behaviour of WO /TiO2 catalyst in the dehydration of glycerol to
3
acrolein. The increasing trend of the acrolein selectivity with reac-
tion time motivated us to perform a 14 h reaction in a fluidized-bed
reactor and withdraw catalyst samples at 1 h intervals. We also
measured the by-products distribution to assess the rate of carbon
deposition and its effect on catalyst performance. The main char-
acteristics of the catalyst such as acidity/basicity, type of the acid
sites, surface area and pore size distribution were correlated to coke
formation and acrolein selectivity during the 14 h time-on-stream.
◦
We preheated the catalyst bed (180 g) to 400 C for 30 min under
◦
a flow of argon. The reactor operated at 280 C for 14 h and at
1 h intervals, catalyst, liquid and gas samples were withdrawn for
analyses. The molar composition of the feed in the gas phase was
glycerol/oxygen/water/argon: 0.06/0.05/1.3/8.9. These conditions
were selected to study the coke build-up and its effect on the cat-
alysts acid and base sites as well as products selectivity.
The conversion of glycerol (Xgly.) as well as selectivities towards
products (Sp) and coke (Scoke) were calculated as follows:
2
. Methodology
in
gly.
out
gly.
n
− n
Xgly. (mol%) =
× 100
(1)
(2)
in
2.1. Catalytic reaction
n
gly.
np
mp
The fluidized bed reactor was a 46 mm ID quartz tube with a
Sp (mol%) =
×
× 100
mgly.
in
gly.
out
gly.
straight section (700 mm long) flanging out to 70 mm ID (300 mm
long) to reduce catalyst carried by the gas to the quench. The reac-
tor set-up has been described in detail in our previous publication
23]. A 20 ␮m ceramic frit, 50 mm from the bottom of the reac-
tor, distributed the feed gases. A 1.6 mm diameter nozzle passed
through the distributor and atomized the glycerol/water solution
n
− n
ncoke
in
gly.
1
3
Scoke (mol%) =
×
× 100
(3)
out
gly.
[
n
− n
where nin and n
out
are the molar flow rates of glycerol at the reac-
gly.
gly.
(
28 wt%) directly into the WO /TiO2 bed. The injector was 20 mm
tor entrance and exit. In Eq. (2), np is the molar flow rate of each
product. mp and mgly. represent the number of carbon atoms of the
products and glycerol. Eq. (3) is a form of Eq. (2) in which the values
of m have been replaced as 1 and 3 for coke and glycerol. The molar
flow rate of coke was calculated based on the results obtained from
TGA analysis.
3
higher than the distributor. The quality of the spraying is crucial
to prevent catalyst from agglomerating and blocking the injector
which leads to the liquid accumulation at the bottom of reactor. To
improve the quality of spraying, a flow of inert gas (Ar) assisted the
liquid feed during the injection. The volumetric ratio of argon to
glycerol solution (G/L) was qualitatively evaluated to be 400. The
pressure drop in the injector at this condition reached to 0.4 bar.
Although the spray quality was better at higher gas to liquid ratios,
the catalyst was more exposed to attrition at higher gas velocities.
An electrical furnace heated the reactor and a 6-point thermo-
couple monitored the temperature of the catalyst bed and each
zone of the reactor. Condensable products of the reaction were
2.2. Characterization techniques
We compared the catalyst samples collected during the reac-
tion with fresh WO /TiO . A TA-Q50 thermogravimetric analyzer
3
2
(TGA) measured the weight loss of catalyst as a function of temper-
ature. We loaded 20 mg of catalyst to a 10 ␮m aluminium crucible. A

1
48
M. Dalil et al. / Journal of Molecular Catalysis A: Chemical 421 (2016) 146–155
Table 1
Effect of operating conditions on product distribution for glycerol dehydration over WO3/TiO2 after 2 h time-on-stream.
◦
T ( C)
yO2 (mol%)
Gly. con. (wt%)
RT (s)
Sacrolein (%)
Sacetaldehyde (%)
Spopanal (%)
Sacetone (%)
Conv. (%)
300
300
300
300
300
300
300
300
280
280
280
280
280
280
280
280
1.5
1.5
1.5
1.5
0.5
0.5
0.5
0.5
1.5
1.5
1.5
1.5
0.5
0.5
0.5
0.5
28
28
14
14
28
28
14
14
28
28
14
14
28
28
14
14
1.5
1.2
1.5
1.2
1.5
1.2
1.5
1.2
1.5
1.2
1.5
1.2
1.5
1.2
1.5
1.2
43
40
50
41
30
36
48
35
58
55
60
60
45
40
50
48
8
8
6
9
6
7
5
8
5
6
3
5
3
7
3
5
7
9
4
6
8
9
5
7
4
6
5
6
8
8
5
7
4
5
3
98
96
98
95
100
96
99
97
97
95
98
95
100
99
100
97
3
12
12
6
9
4
6
4
7
10
11
7
9
Platinel II thermocouple-placed 2 mm above the sample pan mon-
itored the temperature. The resolution and accuracy of the balance
was 0.1 ␮g and >± 0.1%, respectively. To assure that the recorded
weight loss was only related to the carbon combustion, a stream
3. Results and discussion
3.1. Effect of reaction conditions
−
1
◦
of 40 mL min nitrogen first purged the sample at 300 C. After a
To determine the optimal reaction condition for glycerol dehy-
4
1
5 min isothermal hold, air substituted the nitrogen at the same
dration, we executed a 2 by full factorial experimental design. In
◦
flow rate and the furnace ramped to 500 C and held the tempera-
ture for 60 min.
process optimization, selectivity of the catalyst for acrolein was the
response variable. The factors were temperature, gas phase O con-
2
A
JEOL JSM-7600TFE Field Emission Scanning Electron
centration, mass fraction of glycerol in the solution, and residence
time. All tests were at ambient pressure and lasted 2 h.
Microscopy imaged the samples. Graphite adhesive tape held
the powder on the sample holder. We used both a LEI (lower
secondary electron image) and LABE (Low-Angle Backscattered
Electron) detectors.
3
.1.1. Reaction temperature
The operating temperatures were 280 C and 300 C. The cata-
◦
◦
An Autosorb-1 porosimeter (Quantochrome) measured the N
2
lyst was very active at both reaction temperatures (conversion >95)
◦
adsorption/desorption isotherms at −196.15 C. We calculated the
surface area based on the multi-point BET (Brunauer, Emmett and
Teller) method and the pore size and pore volume by the BJH
Barrett–Joyner–Halenda) method.
The type of the acid sites was determined by Fourier transform
infrared spectroscopy (FT-IR) with a Spotlight 400-PerkinElmer FT-
IR spectrometer. Prior to the analysis, the catalyst samples were
treated with pyridine as the probe molecule. The IR range of the
analysis was 600–4000 cm with resolution of 16 cm .
A heat flow calorimeter (Setaram C80) linked to a conventional
volumetric apparatus measured the acidity and basicity of the
◦
and even at 280 C, the conversion of glycerol was nearly complete.
◦
Acrolein selectivity was higher at 280 C and fewer by-products
including acetaldehyde, propanaldehyde and acetone were formed
at the lower temperature. Additionally, at lower temperature, the
HPLC detected traces (less than 2%) of hydroxyacetone in the sam-
ples from the liquid trap.
(
◦
◦
Selectivity of acrolein was lower at 300 C compared to 280 C.
When the temperature increases, acrolein decomposes to lighter
products such as acetaldehyde and formaldehyde. Moreover, at
high temperatures, more carbon deposits on the catalyst and it
deactivates more rapidly.
−
1
−1
◦
catalyst at 150 C. The instrument was equipped with a Barocel
capacitance manometer to monitor pressure. The probes (ammonia
for acidity and sulphur dioxide for basicity) were purified by suc-
cessive freeze–pump–thaw cycles. Prior to the test, 100 mg sample
was pre-treated overnight at 250 C under vacuum. To record the
differential heats of adsorption, small doses of the adsorbate were
introduced repeatedly onto the catalyst and the pressure reached
3
.1.2. Oxygen concentration
The aim of applying molecular oxygen in the gas-phase dehydra-
tion of glycerol is to decrease the formation of aromatic by-products
phenol) and also the by-products originating from hydrogenation
◦
(
of dehydrated products such as propanal and acetone. The pres-
ence of molecular oxygen also reduces coke formation and thereby
maintains the catalyst’s activity [12].
6
6 Pa. The samples outgassed for 30 min at the same temperature,
◦
and the procedure was repeated at 150 C and 27 Pa. The difference
between the amounts adsorbed in the first and second steps rep-
resents the irreversibly adsorbed quantity (Virr.) of a respective gas
from which we estimate the number of acidic/basic sites.
We fed a mole fraction of 0.5% and 1.5% molecular oxygen to
the reactor. The glycerol conversion was the same at both con-
ditions, however, acrolein selectivity was higher at the higher
oxygen concentration while the selectivity of acetone, acetalde-
hyde and propanaldehyde was lower. We noticed that the effect
of oxygen was much higher for acetone than for acetaldehyde or
propanaldehyde (Table 1). Acetone may form via the hydrogenol-
ysis of hydroxyacetone (acetol) which is a product of single-step
glycerol dehydration when the primary OH is released [12,24].
A EURO EA instrument measured the H/C content of the fresh
and used catalyst (CHNS/O). The instrument combusted C, H, S
◦
and N at 600 C. A carrier gas – He – swept the products out of
the combustion chamber. The gases passed over heated copper to
remove unreacted oxygen during the initial combustion and a ther-
mal conductivity detection (TCD) measured their concentration.
We calibrated the instrument with acetanilide.
3
.1.3. Glycerol concentration
The solid catalyst should be insoluble in water and hydrother-
To record the transmission electron microscopy (TEM) micro-
graphs of fresh WO /TiO , the samples were supported on a copper
3
2
mally stable when water vapour is present during the reaction. In
the screening experiments, the mass fraction of glycerol in the liq-
uid phase was 14% and 28%. The selectivity of acrolein was higher
grid and a JOEL JEM-2100F field emission electron microscope
recorded the images.

M. Dalil et al. / Journal of Molecular Catalysis A: Chemical 421 (2016) 146–155
149
minor peaks of methylglyoxal in the first hour samples but it disap-
peared with reaction time. Methylglyoxal is formed via oxidation
of 1-hydroxyacetone.
Acetone was the dominant by-product during the first hour of
reaction and its selectivity was 10%. It dropped to 4% after 2 h and
to below 1% after 11 h. Acetone was likely produced through the
hydrogenation (or hydrogen transfer) of hydroxyacetone forming
1
,2-propanediol (traces observed). Another possible route to ace-
tone is the hydrogenolysis of hydroxyacetone.
Acetaldehyde forms from the retroaldol reaction of 3-
hydroxypropanal (a glycerol dehydration intermediate). The
selectivity of acetaldehyde decreased by 50% during the 14 h test.
The selectivity of formic acid was constant during the reaction and
it is produced via the oxidation of formaldehyde. Formaldehyde is
formed through the retroaldol reaction of 3-hydroxypropanal. We
did not detect any traces of formaldehyde in our product analysis as
◦
it is very unstable (boiling point: −19 C) and perhaps quickly oxi-
dizes to formic acid or degrades to CO and H . CO, CO and H were
Fig. 2. Acrolein selectivity versus reaction conditions for 14 wt% glycerol/water
2
2
2
solution.
present in our gas phase analysis. CO can be the product of side
reactions or in situ coke combustion. The concentration of H as an
2
intermediate increased slightly with time. Some acrolein oxidized
to make acrylic acid but it was less than 1%. Propanal selectivity
was 2% throughout the entire reaction (Table 2).
With reaction time, acrolein selectivity increased, which sug-
gests that either glycerol or acrolein tend to convert to coke and
by-products. As coke grew, the non-selective sites were destroyed
and the catalyst became more selective to acrolein. Consistent
with our kinetic experiments, acrolein selectivity increases while in
addition to coke selectivity, by-product selectivity of acetaldehyde,
formic acid, propanal and acetone decreases (Fig. 5a).
During the first hour, we sampled the catalyst, gas and liquid
products at shorter intervals (5–10 min). The acrolein selectivity
increased as coke accumulated on the catalyst while selectivity of
coke and by-products decreased. However, after 6 h, the selectivity
of acrolein was constant although the coke continued to accumu-
late with time-on-stream (Fig. 6a and b).
Fig. 3. Acrolein selectivity versus reaction conditions with a mass fraction of 28%
glycerol/water solution.
3.3. Characterization of fresh and coked WO /TiO
3
2
for the reactions performed with a mass fraction of 14% glycerol
3.3.1. TGA
(
Figs. 2 and 3). Higher water content in the liquid feed leads to
TGA analysis of the samples withdrawn at 1 h intervals estab-
lished a profile for coke formation. Fluidized beds are ideal mixers
so there are no axial or radial concentration gradients and the coke
on the catalyst represents the average of the bed. After 5 min, the
mass fraction of coke on the catalyst was 0.2% and it reached 1%
after 1 h. At the end of reaction (14 h), the coke fraction reached
2.5%, which represents 4.5 g on the 180 g of catalyst (Fig. 5b).
During the first hour, the coke selectivity reached up to 87%. It
dropped to 50% after 1 h and the acrolein selectivity increased to
36%. The coke selectivity continued to drop after 6 h but it reached
a steady state thereafter (Fig. 6b). After 14 h, the coke selectivity
was 9%. The acrolein selectivity plateaued at 73% after 6 h, which
confirms that it is related to the coke built-up.
the formation of less coke. Furthermore, in the presence of water
vapour, some Lewis acid sites converts to Brønsted sites which
favour acrolein production.
3
.1.4. Residence time
In the range of conditions we tested, residence time (RT) had no
effect on glycerol conversion, nor on acrolein selectivity. As a result,
the fluidizing gas velocity could be set as high as possible in order
to avoid the decomposition of acrolein in the reactor.
3
.2. 14 h catalytic test
WO /TiO catalyst remained fully active during 14 h of reac-
The carbon balance was closed to better than 88% based on sum-
ming the moles of detected products, unreacted glycerol and the
coke. The carbon loss could be related to the minor unknown peaks
in the GC chromatogram or decomposition of some products at
the time of injection to GC. Measuring the coke during the entire
experiment allowed us to achieve a better mass balance, particu-
larly during the early stages when the coking rate was the highest
(Fig. 7).
3
2
tion and converted 97% of glycerol during the first 2 h of injection;
after 3 h all of the gylcerol reacted. High conversion of glyc-
erol could be due to the presence of molecular oxygen in our
reaction feed (mole fraction of 0.5% in the gas feed). Co-feeding
oxygen improves glycerol conversion and maintains the cata-
lyst active [6,25,24]. Based on our liquid and gas phase analysis
of reaction products, propanaldehyde (propanal), hydroxyacetone
(
acetol), acetaldehyde, acetone and formic acid were the major
by-products. Selective dehydration of glycerol to acrolein occurs
when secondary carbon atom loses its OH group. Another reac-
tion pathway involves removal of OH group from primary carbons
of glycerol to give 1-hydroxyacetone (acetol) (Fig. 4). We observed
3.3.2. Elemental analysis (CHNS/O)
Based on the CHNS/O analysis, the mass fraction of carbon for
the 1 h sample was 0.8% and it increased to 2.2% after 14 h. The
TGA results were similar for carbon content (Table 3). Although the

1
50
M. Dalil et al. / Journal of Molecular Catalysis A: Chemical 421 (2016) 146–155
Table 2
Product distribution for 14 h glycerol dehydration over WO3/TiO2 catalyst. The quantity of coke deposition is reported for 180 g of the catalyst.
TOS (h)
Sacrolein (%)
Sacetaldehyde (%)
Sacetol (%)
Sformic acid (%)
Spopanal (%)
Sacetone (%)
Coke (g)
1
1
1
1
1
2
3
4
5
6
7
8
9
1
1
1
1
1
/12
/6
/4
/2
3
3
4
3
3
3
4
4
3
3
3
3
2
3
3
2
2
2
2
2
2
1
3
2
3
1
1
1
1
1
1
1
2
2
3
2
2
2
2
–
–
2
3
1
1
1
1
1
1
1
1
1
–
–
–
–
–
–
1
1
1
2
3
2
1
1
2
1
1
1
1
1
1
1
1
7
9
9
10
10
4
3
3
2
2
1
1
2
1
1
1
–
–
0.2
0.5
0.8
1.2
1.8
2.3
2.4
2.8
3.1
3.1
3.2
3.5
3.8
4.1
4.2
4.3
4.3
4.5
11
36
45
55
55
63
67
74
71
70
71
71
70
72
73
0
1
2
3
4
Table 3
Elemental analysis data revealing H/C ratio.
3
.3.3. FE-SEM
The surface of the fresh catalyst was rough and the particle
Time (h)
1
H
C
H/C
C (TGA)
size distribution was heterogeneous (Fig. 8a). However, the particle
shape was the same after 14 h (Fig. 8c) and it did not agglomerate.
For both fresh and used samples, a layer of fine powder (dp < 1 ␮m)
on the surface covered the particles. Non-uniformity in the size of
the particles and also the thin layer formed on the surface was due
to the crushing of WO3/TiO2 catalyst which was originally in pellet
form (Fig. 8b and d).
0.21
0.58
0.82
2.2
0.26
0.27
0.97
2.4
1
4
amount of carbon and hydrogen increased continuously, the H/C
ratio of carbonaceous compounds formed on the catalyst increased
from zero to 0.26 after 1 h and remained constant up to 14 h. The
H/C ratio suggests that these compounds were polyaromatic coke.
However, the type of coke was the same throughout the reaction
3.3.4. TEM
TEM micrographs exhibited agglomeration of fine particles
(
Table 3).
in the structure of WO /TiO2 (Fig. 9a and b). We observed a
3
Fig. 4. Reaction mechanism for steady-state dehydration of glycerol to acrolein and by-products.

M. Dalil et al. / Journal of Molecular Catalysis A: Chemical 421 (2016) 146–155
151
Fig. 6. Selectivity towards acrolein and coke exhibited reverse trends (a) at first
hour of reaction and (b) during the 14 h time-on-stream.
Fig. 5. (a) Selectivity of by-products decreased by time-on-stream. Formic acid and
propanaldehyde were rather constant during the reaction. (b) The profile of carbon
built-up for coke deposition over WO3/TiO2.
representative lattice fringe pattern for fresh catalyst indicating
that WO /TiO particles were crystalline (Fig. 9c). The crystalline
3
2
structure was also confirmed by the selected-area electron diffrac-
tion (SAED) pattern (Fig. 9d).
3.3.5. N2 adsorption
The hysteresis loops for the N adsorption/desorption isotherms
2
are type H1 as defined by IUPAC. In H1 type hysteresis, two of the
branches are almost vertical and nearly parallel in the range of the
gas uptake [26]. The shape of the hysteresis loop is associated with
the structure of the pores. Type H1 represents cylindrical shaped
pores. This type of hysteresis has been reported for pure titania
by Bavykin and Walsh [27] and Bing et al. [28]. The pore volume
3
−1
3
−1
of fresh catalyst was 0.22 cm g and it reduced to 0.18 cm g
after 2 h and remained constant. Thereafter, the coke reduced the
mean pore diameter of the catalyst from 17.3 nm to 12.4 nm at
6
h. The decrease in the pore volume and diameter was related to
the coverage of the catalyst pores by coke (Table 4). The adsorp-
tion/desorption isotherms of the 14 h sample was the same as fresh
sample indicating that the pores remained cylindrical (Fig. 10).
This data suggests that the coke forms in the pores and covers the
internal walls of the cylinders rather than plugging the entrance
(
Fig. 11).
The nitrogen adsorption/desorption isotherms exhibited an
inflection at 0.6 < P/P < 0.9 indicating that the fresh catalyst was
Fig. 7. Carbon balance versus time.
0

1
52
M. Dalil et al. / Journal of Molecular Catalysis A: Chemical 421 (2016) 146–155
Fig. 8. FE-SEM images of (a and b) fresh catalyst and (c and d) used catalyst.
2
−1
19
−2
−1
mesoporous. The surface area of fresh catalyst was 30.5 m g and
it decreased with time (coke loading) [23]. After 1 h, when the mass
fraction of coke was 1% (1.8 g of carbon), the surface area dropped
3.8 × 10 atoms m . The number of C layers for 2.5% carbon depo-
2
sition on 30.5 m g of catalyst, is equivalent to a monolayer after
14 h.
2
−1
2
−1
to 27 m g and after 14 h it was 23 m g with 2.5% coke (4.5 g)
Table 4).
To calculate the number of the carbon layers on surface, we
(
3.3.6. Acid/base properties (micro-calorimetric results)
The adsorption of probe molecules by micro-calorimetry indi-
cates the acid/base features of the catalyst. NH3 was used as a
basic molecule to determine the acidity while SO titrated the basic
sites (Air Liquide >99.9%). The initial sharp rise in the volume of
considered the formation of a graphite type carbon. Graphite
with hexagonal structure, has cell dimensions of a = 2.461 A˚ and
2
c = 6.708 A˚ . So, there are 2 carbon atoms per basal plane or
Fig. 9. TEM images of fresh catalyst: (a) 0.5 ␮m, (b) EDAX, (c) 10 nm and (d) SAED. The crystalline structure of WO3/TiO2 is observed in (c) and (d).

M. Dalil et al. / Journal of Molecular Catalysis A: Chemical 421 (2016) 146–155
153
Fig. 10. N2 adsorption/desorption isotherms for fresh and used WO3/TiO2.
Table 4
Table 5
Textural properties of WO3/TiO2 catalyst.
Virr. and Vtot. calculated from adsorption isotherms of SO2 and NH3 on fresh and used
WO3/TiO2.
SA (m2
g
−1
)
ꢀ
p (cm3
g
−1
)
dp (nm)
Time (h)
Sample NH3 adsorbed amount
SO2 adsorbed amount
0
1
2
6
30.5
27.1
26.3
25.5
23.0
0.22
0.20
0.18
0.17
0.18
17.3
17.5
17.3
12.4
12.4
Vtot. (␮mol g−1) Virr. (␮mol g−1) Vtot. (␮mol g−1) Virr. (␮mol g−1)
Fresh
143.2
125.4
116.9
117.2
86.5
64.3
61.3
60.2
8.8
6.8
–
3.8
2.2
–
2
6
1
h
h
4 h
1
4
3.1
0.8
1
4 h samples were the same indicating that the number of acid
sites remained constant after 6 h (Table 5). The NH3 adsorption
isotherms of 6 h and 14 h samples are identical in Fig. 12. Initially,
the acrolein selectivity was negligible; it increased with time to
reach a steady rate and was constant after 6 h, which is the identi-
cal trend for NH and SO adsorption. Thus the change in selectivity
3
2
relates to the change in acidity and basicity of the catalyst.
According to the SO2 adsorption measurements, the fresh cat-
alyst has few basic sites (medium to weak strength) compared to
the acid sites (Table 5 and Fig. 12). As coke accumulated, the basic
sites disappeared in less than 2 h and the formation of acetone and
Fig. 11. Carbon deposited on the wall of the pores of WO3/TiO2, reducing the pore
diameter.
1
,2-propanediol dropped considerably (Fig. 5).
The differential heats of adsorption confirm the presence of
adsorbed NH corresponds to irreversible adsorption, whereas the
3
strong, medium and weak acid sites on the fresh catalyst (Fig. 13).
We derived the number of acid and basic sites from Fig. 13. The
three domains of acid sites for Qdiff are defined as weak, medium
and strong [29]:
rising slope after 0.05 Torr relates to reversible adsorption (Fig. 12).
Fresh WO /TiO adsorbed the most ammonia because it
3
2
has most acid sites and the highest strength (high heats of
adsorption). As coke forms, the catalyst adsorbs fewer probe
molecules. The number of acid sites decreased in the order: fresh
WO /TiO > 2 h > 6 h ꢀ 14 h. The greatest drop was from the fresh
•
•
90 < Q < 120: weak
120 < Q < 150: medium
3
2
catalyst to the 2 h sample. The volume adsorbed by the 6 h and
Fig. 12. Volumetric isotherms of (a) NH3 and (b) SO2 adsorption on fresh and spent WO3–TiO2 catalyst.

1
54
M. Dalil et al. / Journal of Molecular Catalysis A: Chemical 421 (2016) 146–155
Fig. 13. Differential heats of NH3 and SO2 adsorption as a function of surface cov-
erage for fresh and spent WO3/TiO2 catalyst.
Fig. 15. FT-IR spectra recorded after pyridine adsorption at room temperature and
desorption at 100 C for fresh and used catalysts. (L: Lewis and B: Brønsted acid
◦
sites).
strong
medium
weak
basic
+
−1
−1
(
1
PyH ) and Brønsted sites appear at 1640 cm , 1620 cm
and
−1
540 cm . The band at 1490 cm corresponds to a mixture of
−
1
1
0
8
6
4
2
0
Lewis and Brønsted acids and the pyridine that was coordinated
to Lewis sites on fresh dehydrated catalyst (Py-L) appeared at
−
1
1450 cm [30]. For the 6 h sample, the band corresponding to B + L
sites was still dominant while the Brønsted acid bands decreased.
The peaks of Lewis acidity disappeared while the peaks of Brønsted
sites reduced with reaction time. We attribute the drop in the
Brønsted and Lewis sites to coke formation. However, there was no
noticeable difference in the band intensity between 6 h and 14 h
samples suggesting that although the coke forms continuously,
after 6 h the acidity of the catalyst remained constant. Conse-
quently, the selectivity of acrolein and by-products was stable
after 6 h and the reaction reached steady-state. The coke selectiv-
ity (Fig. 6b) approached a plateau after 6 h which corresponds to
the same trend for the acid/base properties of the catalyst, acrolein
selectivity and coke. This also proves that Lewis acid sites do not
contribute to the acrolein selectivity.
0
2
6
14
time, h
Fig. 14. Number of acid sites of different strengths and basic sites versus reaction
time.
4. Conclusions
•
150 < Q: strong
Commercial biofuels manufacture from vegetable oil or animal
fat has increased the supply of glycerol and subsequently, its price
has dropped five fold (or more) in some regions of the world. Glyc-
erol is an attractive feedstock to produce acrolein. Obstacles to
commercialize the glycerol-to-acrolein process relate to achieving
high selectivity, reducing the by-products and catalyst deactivation
but also finding enough supply of glycerol at a price low enough.
Coke formation reduced the number of medium and weak acid
sites of the catalyst (Fig. 14), while after 14 h, the number of strong
sites did not vary noticeably. Because of the presence of the strong
sites, the catalyst remained fully active after 14 h and selectively
produced acrolein.
WO /TiO2 catalyst selectively dehydrated glycerol to acrolein at
3
3
.3.7. FT-IR pyridine
a selectivity of 73% in a fluidized bed at 100% conversion even
though coke formed on the surface. It increased steadily from 3% to
the steady state value after 6 h on-stream. Concurrently, the selec-
tivity of acetone and acetaldehyde decreased: acetone selectivity
dropped from 10% in the first hour to zero after 14 h. The acetone
selectivity was related to the hydrogen transfer reaction. As less
coke formed, the acetone and propanal selectivity decreased. In the
first hour coke selectivity was 50% and it dropped to 9% after 14 h.
Coke covers the active sites that catalyse the parallel reactions and
for this reason acrolein selectivity increased with time-on-stream.
After 6 h both the coke rate and acrolein selectivity were constant.
The Lewis and Brønsted sites reduced with reaction time based on
TPD-NH measures the strength of the acid sites but not the type
3
of these sites (Brønsted or Lewis acid). The FT-IR pyridine spectra
of WO /TiO differentiates between the Brønsted and Lewis acid
sites (Fig. 15). Lewis acid sites may react with steam to generate
Brønsted sites [11]. Brønsted sites dehydrate glycerol to acrolein
by protonating its secondary hydroxyl.
We recorded the background spectra of the samples before
the main analysis. Then we treated the samples with pyridine at
room temperature and desorbed the excess pyridine at 100 C for
3
2
◦
1
h. After cooling to room temperature, the instrument recorded
the spectra. The bands of the spectra relate to pyridinium cations

M. Dalil et al. / Journal of Molecular Catalysis A: Chemical 421 (2016) 146–155
155
FT-IR pyridine analysis. Strong acid sites remained constant after
4 h. This could be related to the continuous catalyst regenera-
[12] J.-L. Dubois, C. Duquenne, W. Hoelderich, J. Kervennal, WO2006087084 A2,
006.
2
1
[
[
13] A. Neher, T. Haas, D. Artnz, H. Klenk, W. Girke, US 5387720 A, 1993.
14] M.N. Timofeeva, Appl. Catal. A: Gen. 256 (1–2) (2003) 19–35.
tion by molecular oxygen or because of coke formation on basic or
medium acidic sites before reaching the strong acidic sites. Strong
acid sites survive during the reaction time whereas by-products for-
mation decreased with decrease in medium acidic and basic sites
of the catalyst.
[15] Y. Wu, X. Ye, X. Yang, X. Wang, W. Chu, Y. Hu, Ind. Eng. Chem. Res. 35 (8) (1996)
546–2560.
2
[
[
16] I.V. Kozhevnikov, J. Mol. Catal. A: Chem. 262 (1–2) (2007) 86–92.
17] J.-L. Dubois, Y. Magatani, K. Okumura, WO2009127889 A1, 2009.
[18] J.-L. Dubois, Y. Magatani, K. Okumura, WO2009128555 A3, 2009.
[
19] D. Stosic, S. Bennici, J.-L. Couturier, J.-L. Dubois, A. Auroux, Catal. Commun. 17
2012) 23–28, ISSN 1566-7367.
(
References
[
20] B. Katryniok, S. Paul, M. Capron, C. Lancelot, V. Belliere-Baca, P. Rey, F.
Dumeignil, Green Chem. 12 (2010) 1922–1925.
[
[
1] H. An, W.E. Wilhelm, S.W. Searcy, Biomass Bioenergy 35 (9) (2011) 3763–3774.
2] E.F. Aransiola, T.V. Ojumu, O.O. Oyekola, T.F. Madzimbamuto, D.I.O. Ikhu-
Omoregbe, Biomass Bioenergy 61 (2014) 76–297.
[21] S. Erfle, U. Armbruster, U. Bentrup, A. Martin, A. Bruckner, Appl. Catal. A: Gen.
391 (1–2) (2011) 102–109.
[22] W. Suprun, M. Lutecki, T. Haber, H. Papp, J. Mol. Catal. A: Chem. 309 (1–2) (2009)
71–78.
[23] M. Dalil, D. Carnevali, J.-L. Dubois, G.S. Patience, Chem. Eng. J. 270 (2015)
557–563.
[24] A. Ulgen, W.F. Hoelderich, Appl. Catal. A: Gen. 400 (1–2) (2011) 34–38.
[25] A. Ulgen, W. Hoelderich, Catal. Lett. 131 (1–2) (2009) 122–128.
[26] K.S.W. Sing, D.H. Everett, R.A.W. Haul, L. Moscou, R.A. Pierotti, J. Rouquerol, T.
Siemieniewska, Pure Appl. Chem. 57 (4) (1985) 603–619.
[27] D. Bavykin, F.C. Walsh, Titanate and Titania Nanotubes: Synthesis, Royal Society
of Chemistry, 2010.
[28] G. Bing, S. Hangyan, S. Kangying, Z. Yaowu, N. Wensheng, J. Chem. Sci. 121 (3)
(2009) 317–321.
[29] A. Auroux, Calorimetry and Thermal Methods in Catalysis, Springer, 2013.
[30] C. Martin, G. Solana, V. Rives, G. Marci, L. Palmisano, A. Sclafani, Catal. Lett. 49
(3–4) (1997) 235–243.
[
[
3] M. Nasrin, A. Alireza, R. Reza, M. Tangestani, R. Eghtesadi, A.S. Nejad, Biomass
Bioenergy 39 (2012) 449–453.
4] D.D. Eley, H. Pines, P.B. Weisz (Eds.), Volume 27 of Advances in Catalysis, Aca-
demic Press, 1979.
[
[
[
[
5] Schering Kahlbum AG, FR 695931 A, 1930.
6] J.-L. Dubois, C. Duquenne, W. Hoelderlich, WO2006087083A2, 2006.
7] S.-H. Chai, H.-P. Wang, Y. Liang, B.-Q. Xu, Green Chem. 9 (2007) 1130–1136.
8] E. Tsukuda, S. Sato, R. Takahashi, T. Sodesawa, Catal. Commun. 8 (9) (2007)
1
349–1353.
[9] H. Atia, U. Armbruster, A. Martin, J. Catal. 258 (1) (2008) 71–82.
[
[
10] K. Benjamin, P. Sebastien, C. Mickael, L. Christine, B.-B. Virginie, R. Patrick, D.
Franck, Green Chem. 12 (2010) 1922–1925.
11] A. Alhanash, E.F. Kozhevnikova, I.V. Kozhevnikov, Appl. Catal. A: Gen. 378 (1)
(
2010) 11–18.