WO3 supported on Zr doped mesoporous SBA-15 silica for glycerol dehydration to acrolein

Article abstract of DOI:10.1016/j.apcata.2016.02.016

Glycerol dehydration to acrolein is one of the most promising routes to valorize the surplus of such by-product of the biodiesel industry. In this context, a family of solid acid catalysts based on tungsten oxide supported over zirconium doped mesoporous SBA-15 silica has been synthesized and the influence of phosphoric acid treatment on the catalytic behavior has been also evaluated. These catalysts have been characterized by using the following techniques: XRD, N₂ adsorption-desorption at -196 °C, NH₃-TPD, pyridine adsorption coupled to FT-IR spectroscopy, X-ray photoelectron spectroscopy, Raman and diffuse reflectance UV-visible spectroscopies, and they have been tested in the gas-phase dehydration of glycerol. Under the operating conditions assayed, all catalysts were active and acrolein was always the main product. The catalytic performance of support was enhanced when WO₃ was supported and after treatment with H₃PO₄. The best catalytic results were achieved for 20W catalyst which showed values of glycerol conversion and acrolein yield of 97% and 41%, respectively, after 2 h on stream at 325 °C, and this activity was even maintained after 8 h. Moreover, the treatment with phosphoric acid improved the acrolein yield at 2 h on stream, but the deactivation of these catalysts was more pronounced probably because of the deterioration of textural and acid properties.

Full text of DOI:10.1016/j.apcata.2016.02.016

Accepted Manuscript
Title: WO3 Supported on Zr Doped Mesoporous SBA-15
Silica for Glycerol Dehydration to Acrolein
´
´
Author: J.A. Cecilia C. Garcıa-Sancho J.M. Merida-Robles J.
´
´
Santamarıa Gonzalez R. Moreno-Tost P. Maireles-Torres
PII:
S0926-860X(16)30082-5
DOI:
Reference:
http://dx.doi.org/doi:10.1016/j.apcata.2016.02.016
APCATA 15775
To appear in:
Applied Catalysis A: General
Received date:
Revised date:
Accepted date:
16-12-2015
11-2-2016
12-2-2016
´
´
Please cite this article as: J.A.Cecilia, C.Garcia-Sancho, J.M.Merida-Robles,
´
´
J.Santamaria Gonzalez, R.Moreno-Tost, P.Maireles-Torres, WO3 Supported on Zr
Doped Mesoporous SBA-15 Silica for Glycerol Dehydration to Acrolein, Applied
Catalysis A, General http://dx.doi.org/10.1016/j.apcata.2016.02.016
This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.

WO3 Supported on Zr Doped Mesoporous SBA-15 Silica for Glycerol Dehydration
to Acrolein
J.A. Cecilia, C. García-Sancho* cristinags@uma.es, J.M. Mérida-Robles, J. Santamaría
González, R. Moreno-Tost, P. Maireles-Torres
Universidad de Málaga, Departamento de Química Inorgánica, Cristalografía y
Mineralogía (Unidad Asociada al ICP-CSIC), Facultad de Ciencias, Campus de
Teatinos, 29071 Málaga, Spain
*Corresponding author.
1

Graphical Abstract
2

Highlights
 WO3 supported over Zr-SBA-15 catalysts are active in dehydration of glycerol,
being always acrolein the main product which is considered as a versatile
intermediate for the chemical industry.
 WO3 enhances the total acidity and the Brönsted acid sites ratio of support,
consequently improving the activity and stability of catalysts in glycerol
dehydration to acrolein.
 Glycerol conversion and acrolein yield are maintained after 8 h for W catalysts,
attaining the highest values for a WO3 loading of 20 wt% which may be related
to WO3-ZrO2 phases on catalyst surface.
 Although the H3PO4 treatment enhances acrolein yield for short reaction times,
high amounts of acid damage the textural and acidic properties of W catalysts,
decreasing the available acid sites.
3

Abstract
Glycerol dehydration to acrolein is one of the most promising routes to valorize the
surplus of such by-product of the biodiesel industry. In this context, a family of solid
acid catalysts based on tungsten oxide supported over zirconium doped mesoporous
SBA-15 silica has been synthesized and the influence of phosphoric acid treatment on
the catalytic behaviour has been also evaluated. These catalysts have been characterized
by using the following techniques: XRD, N2 adsorption-desorption at -196 ºC, NH3–
TPD, pyridine adsorption coupled to FT-IR spectroscopy, X-ray photoelectron
spectroscopy, Raman and diffuse reflectance UV–visible spectroscopies, and they have
been tested in the gas-phase dehydration of glycerol. Under the operating conditions
assayed, all catalysts were active and acrolein was always the main product. The
catalytic performance of support was enhanced when WO3 oxide was supported and
after treatment with H3PO4. The best catalytic results were achieved for 20W catalyst
which showed values of glycerol conversion and acrolein yield equal to 97% and 41%
respectively after 2 h on stream at 325 ºC and this activity was even maintained after 8
h. Moreover, the treatment with phosphoric acid improved the acrolein yield at 2 h on
stream, but the deactivation of these catalysts was more pronounced probably because
of the deterioration of textural properties and total acidity.
Keywords: tungsten oxide; phosphoric acid; acid catalysts; glycerol; acrolein;
zirconium doped mesoporous silica
4

1.
Introduction
Due to the increasing cost and depletion of fossil fuel reserves, renewable fuels
have been spread worldwide in last years, being biodiesel a remarkable example of
biofuels. Biodiesel is a mixture of alkyl esters produced by transesterification of
vegetable oils and animal fats with short-chain alcohol, obtaining glycerol as by-product
[1, 2]. Hence, a large surplus of glycerol has been generated by the biodiesel industry,
being essential its valorization to ameliorate the sustainability and economical viability
of the biodiesel industry [3, 4]. Glycerol is one of the most versatile and valuable
chemical in industry, since more than 1500 direct applications are already known,
especially in cosmetics, pharmaceuticals and food industries [5, 6]. Moreover, glycerol
is a molecule with a large potential since it offers many opportunities to produce added-
value chemicals through different reactions, such as hydrogenolysis, partial oxidation,
etherification, halogenation, esterification or dehydration [5, 7].
One of the most promising routes of glycerol valorization consists of its catalytic
dehydration to acrolein, an important and versatile intermediate for the chemical
industry [8]. Most of produced acrolein is employed for the synthesis of acrylic acid
which is a chemical with high demand, whose annual global growth rate is of 4%.
Likewise, acrolein is widely utilized in the synthesis of 3-methylthiopropionaldehyde, a
precursor of D,L-methionine which is an essential amino acid required in meat
production to accelerate animal growth [8], as well as in the production of
glutaraldehyde, 1,2,6-hexanetriol, quinoline, pentaerythritol, cycloaliphatic epoxy
resins, oil-well derivatives and water treatment [9]. In addition, it can be directly used as
5

an effective aquatic biocide to control the growth of undesired microbial material and
aquatic weeds.
Acrolein is currently produced from partial oxidation of propylene catalyzed by
complex multicomponent BiMoOx catalysts [10]. Its production from dehydration of
glycerol is considered of great interest by researchers due to glycerol is a relatively
inexpensive raw material which is available from biodiesel industry. Although it is
known that acrolein can be obtained from thermal decomposition of glycerol, it is
convenient to employ an acid catalyst in order to improve the catalytic performance.
Thus, different acid catalysts have been tested for dehydration of glycerol to acrolein,
such as zeolites [11-14], heteropolyacids [15-17], sulfated zirconia [18, 19] or metal
oxide [20, 21]. Tungsten oxide can be found among the different active phases proposed
for this reaction. Thus, Kraleva et al. synthesized a W-SBA-15 catalyst (Si/W molar
ratio of 20) and achieved high activities even at low temperatures, obtaining a value of
glycerol conversion about 88% at 275 ºC. Its catalytic performance was also enhanced
after 1,2-tungstophosphoric acid impregnation [16]. Chai et al. have tested WOx
supported on Al2O3, ZrO2, and SiO2 and they showed that the WO3/Al2O3 and
WO3/ZrO2 catalysts were more effective than WO3/SiO2 to catalyze the glycerol
dehydration in terms of acrolein selectivity, being the most effective catalyst 30 wt%
WO3/Al2O3 sample since the acrolein yield of 61% was maintained after 10 h [22].
Similarly, other researchers have studied WO3 supported on ZrO2 [23-25] and the effect
of doping with Si atoms a WO3/ZrO2 catalyst and they achieved high values of glycerol
conversion and acrolein selectivity [25, 26]. Moreover, mixed oxides of tungsten and
niobium have been supported over different materials, demonstrating to be active
catalysts for this reaction [27-29]. Comparing niobium oxide with tungsten oxide,
6

Massa et al. observed that the catalytic activity of supported tungsten oxide was slightly
better, with an initial yield to acrolein around 78% against 75% for niobium oxide, at
complete conversion of glycerol after 3 hours of reaction [29]. In addition, they did not
found a synergistic effect using mixed W/Nb oxides.
On the other hand, it is well known that the acidity of the active phase is a key
parameter which affects the catalytic performance, being relevant the type of acid sites
since several authors have proved that Brönsted acid sites provide higher values of
acrolein selectivity than Lewis acid sites [30-32]. In this sense, phosphoric acid has
been widely used in literature to increase the total acidity of catalysts by providing new
Brönsted acid sites [33-35]. This fact has been corroborated by Struzhko et al., who
recently treated a mesoporous WOx∙ZrO2 solid with phosphoric acid and evaluated its
catalytic behavior in dehydration of glycerol, finding an increasing in the acrolein
selectivity, although this also depended on the WOx loading [36]. Likewise, Rao et al.
tested tungsten oxide supported on porous zirconium phosphate reaching a high
selectivity to acrolein (about 82%) at total glycerol conversion, at 300 ºC [37]. They
also demonstrated that the catalytic behavior mainly depended on both the fraction of
acid sites of moderate strength and those of Brönsted type. Finally, Suprun et al. studied
a family of transition metal supported on aluminophosphates (APO), being the highest
acrolein selectivity (52–58%) reached with a W-APO catalyst [38]. Likewise, they
evaluated the influence of different transition metal oxide on phosphated alumina on the
gas-phase dehydration of glycerol in such a way that the highest acrolein selectivity was
detected over W contained catalysts [39].
7

In addition, it should be pointed out that the deactivation of catalysts is an usual
drawback in the glycerol dehydration reaction [7, 8]. In some catalytic systems studied,
the porous network of the catalysts itself was one of the main source of deactivation
since small pores could complicate the diffusion of the products and favored the
formation of coke, which finally blocked the access to the active phase [7, 40].
Therefore, there is a great deal of interest to employ supports with high-surface area in
order to enhance the dispersion of the active phase and improve the diffusion of
reactants and products through the porous network.
The aim of this work was to synthesize solid acid catalysts based on tungsten
oxide, as active phase, supported on zirconium doped mesoporous silica with a SBA-15
structure. These catalysts were tested in the gas phase dehydration of glycerol to
acrolein, and the effect of the treatment with phosphoric acid was also evaluated. The
influence on the catalytic performance of experimental parameters, such as reaction
temperature and WOx loading, has been studied.
2.
Experimental
2.1.
Synthesis of catalysts
A zirconium doped mesoporous SBA-15 silica with a Si/Zr molar ratio of 5 was
synthesized following the method previously reported by Cecilia et al. [34], employing
Pluronic P123 (Sigma-Aldrich, average Mn~5800), as structure directing agent, and
tetraethyl orthosilicate (Sigma-Aldrich, 98%) and zirconium n-propoxide (Sigma-
Aldrich, 70% in n-propanol) as sources of silicon and zirconium, respectively. The final
molar composition of the synthesis gel was P123/SiO2/ZrO2/HCl/H2O=
1/55/11/350/11100. The resulting suspension was stirred at 40 ºC for 72 h and the solid
8

product was recovered by filtration, washed with deonized water and dried at 60 ºC, and
finally calcined in air at 550 ºC for 6 h, with a heating rate of 1 ºC∙min−1. This material
was denoted as SiZr.
Tungsten oxide was supported on the SiZr support by incipient wetness
impregnation, by using ammonium metatungstate hydrate (Sigma-Aldrich, ≥66.5% W)
aqueous solutions to achieve WO3 loadings ranging between 10 and 40 wt%. Then,
these materials were dried at 60 ºC, and calcined at 400 ºC for 4 hours (heating rate of 2
ºC∙min−1) in order to form the WO3 phase. The catalysts were labeled as xW, where x
stands for wt% of WO3.
The 20W catalyst, before calcination, was treated with different amounts of
H3PO4, obtaining 20WPy catalysts, where y symbolizes the W/P molar ratio. Next, they
were dried at 60ºC and calcined at 400 ºC for 4 hours (20WPy).
2.2. Characterization of catalysts
Powder diffraction patterns were collected on an PANalytical automated
diffractometer, EMPYREAN model, using Cu-Kα_1,2 (1.5406 Å) radiation and a last
generation PIXcel detector. The database employed were ICSD, PDF 2-2004 and COD
of High Score Plus programme of PANalytical.
The textural parameters were determined from the nitrogen adsorption–
desorption isotherms at -196 ºC, obtained by using an automatic ASAP 2020 model of
gas adsorption analyser from Micromeritics. Prior to N2 adsorption, the samples were
outgassed at 200 ºC and 10-4 mbar for 10 h. Surface areas were determined by using the
Brunauer–Emmet–Teller (BET) equation and a nitrogen molecule cross section of 16.2
9

Å2. The Density Functional Theory method (DFT) was employed to determine the pore
size distribution.
X-ray photoelectron spectra were obtained with a Physical Electronics PHI 5700
spectrometer with non-monochromatic Mg Kα radiation (300 W, 15 kV, and 1253.6 eV)
with a multi-channel detector. Spectra were recorded in the constant pass energy mode
at 29.35 eV, using a 720 µm diameter analysis area. Charge referencing was measured
against adventitious carbon (C 1s at 284.8 eV). A PHI ACCESS ESCA-V6.0 F software
package was used for acquisition and data analysis. A Shirley-type background was
subtracted from the signals. Recorded spectra were always fitted using Gaussian–
Lorentzian curves in order to determine the binding energies of the different element
core levels more accurately.
Raman spectra were recorded on a FT-Raman spectrometer, using a Nd:YAG
laser as the excitation source at 1064 nm and the laser power was set to 150 mW and a
standard spectral resolution of 4 cm-1. Raman spectroscopy was performed on powder
samples without any previous treatment. UV–visible spectroscopy studies were carried
out in the diffuse reflectance mode with a Shimadzu MPC3100 spectrophotometer and
BaSO4 as reference.
The temperature-programmed desorption of ammonia (NH3-TPD) was carried
out to evaluate the total surface acidity of the catalysts. The catalysts (80 mg) were
evacuated with helium at 550 ºC and then ammonia was adsorbed at 100 ºC. Finally, the
NH3-TPD was performed by raising the temperature from 100 to 550 ºC, under a helium
flow of 40 mL min-1, with a heating rate of 10 ºC min-1 and maintained at 550 ºC for 15
10

min. The evolved ammonia was analyzed by using a TCD detector of a gas
chromatograph (Shimadzu GC-14A).
FTIR spectra of adsorbed pyridine were recorded on a Shimadzu Fourier
Transform Infrared Instrument (FTIR8300). Self supported wafers of the samples with a
weight/surface radio of about 15 mg cm−2 were placed in a vacuum cell greaseless
stopcocks and CaF2 windows. The samples were evacuated at 300 ºC and 10−2 Pa
overnight, exposed to pyridine vapours at room temperature (vapour pressure of 200
mbar) for 10 min and then outgassed at 100 ºC, 200 ºC and 300 ºC.
Carbon content of spent catalysts was measured with a LECO CHNS 932
analyzer. Thermogravimetric analyses (TGA) were carried out in a TGA/DSC 1
(Mettler–Toledo), measuring the weight variation under an air flow of 50 ml·min-1 in
the temperature range of 30-900 ºC with a heating rate of 10 ºC·min−1.
2.3. Catalytic tests
The dehydration of glycerol was carried out, at atmospheric pressure, in a fixed-
bed continuous-flow stainless steel reactor (9.1 mm in diameter, and 230 mm in length),
operated in the down-flow mode. The reaction temperature was measured with an
interior placed thermocouple in direct contact with the catalyst bed. For the activity
tests, 0.5 g of catalyst (particle size 0.85–1.00 mm), diluted with SiC to 3 cm3 volume,
were used. Catalysts were pre-treated in situ at atmospheric pressure under a N2 flow of
15 mL min-1, at 325 ºC for 30 min. The glycerol solution (10 wt% in water) was
supplied by means of a Gilson 307SC piston pump (model 10SC) at 0.1 mL min-1 feed
rate in a N2 flow (15 mL min-1). The evolution of the catalytic performance was studied
by collecting liquid samples between 2 hours and 8 hours in a vial cooled in an ice trap.
11

These vials were sealed and analyzed by means of gas chromatograph (Shimadzu GC-
14B), equipped with a flame ionization detector and a capillary column (TRB-14
Teknokroma).
The glycerol conversion (mol%), the selectivity of the identified products and
the yields have been calculated as follows:
nGly,in  n
C

%



Gly,out 100
nGly,in
ni
zi
Si
yi

%


100
nGly,in  nGly,out zGly.
C  Si
100

% 

where nin and nout represent the mol of glycerol at the inlet and outlet of the reactor. ni is
the molar concentration of the product i at the outlet of the reactor and zi and zGly are the
number of carbon atoms of the product i and glycerol, respectively. The carbon balance
(mol%) was calculated by summing up the unreacted glycerol and the total quantities of
detected and calibrated products. The selectivity of non-detected products was evaluated
by difference between 100 and all quantified detected products.
3.
Results and discussion
3.1. Characterization of catalysts
Powder X-ray diffraction has been used to identify crystalline phases present in
these catalysts. Thus, XRD patterns at low-angle of xW catalysts (Figure 1A) show an
intense peak at ca. 2= 1.2º corresponding to the (1 0 0) reflection, confirming the
12

mesoporous structure, even after incorporation of the highest amount of tungsten oxide
(40 wt.%). However, this diffraction peak is attenuated when catalysts are treated with
phosphoric acid (20WPy), disappearing for 20WP0.2 catalyst, pointing out the collapse
of the mesostructure of the SiZr support. This fact has been previously observed by Wu
et al. [41], who found a less ordered structure after treatment of a mesoporous SBA-15
silica with phosphoric acid, although, in the present work, it is more marked in the case
of the 20WPy catalysts, which can be attributed to the presence of tungsten oxide. With
regard to the XRD patterns in the high-angle region (Figure 1B), the characteristic
reflections of monoclinic and triclinic WO3 phases are detected for all catalysts (Powder
Diffraction Line reference codes 98-065-3651 and 98-008-0055, respectively) [42, 43],
being mainly observed in 31-37º region (Supplementary Information, Fig. 1S). These
diffraction peaks, as expected, are much more intense for WO3 loadings higher than 10
wt%. However, their intensity is considerably decreased after phosphoric acid
treatment, which could be probably due to the formation of amorphous W and P phases.
In no case, tetragonal ZrO2 phase has been found for these catalysts, so that zirconium
apparently stays incorporated into the siliceous framework.
The mesoporous character of the xW catalysts has been confirmed from the
corresponding nitrogen adsorption–desorption isotherms at −196 ºC, since all catalysts
display Type IV isotherms according to the IUPAC classification. Moreover, the shape
of the isotherms is preserved after incorporation of different amounts of tungsten oxide
(10-40 wt%) over the SiZr support, and even after treatment with phosphoric acid
(Supplementary Information, Figure 2S). The determination of textural parameters
reveals that the incorporation of increasing loadings of WOx species on the support
leads to a decrease of specific surface area and pore volume in comparison with those of
13

the SiZr support (Table 1). This could be explained by the presence of WO3
nanoparticles, detected by XRD, partially blocking the pores, thus modifying the pore
size distribution of the SiZr support (Figure 2A). Thus, the SiZr support shows a wide
range of pore sizes, from micropores to mesopores; however, when the WO3 particles
are generated, both the micropores and the pores with sizes comprised between 2 and 4
nm decreased, with the concomitant increase of the pores within the range of 4 to 6 nm.
These two effects lead to an increase of the average pore diameters with respect to the
support (Table 1). Likewise, the treatment with H3PO4 provokes a serious damage in the
mesostructure of these materials, decreasing markedly the surface area and pore
volume, even up to 15 m2·g-1 for 20WP0.2, and raising the average pore diameter, so
that the treatment with phosphoric acid is detrimental for the textural properties, as
Cecilia et al. [34] previously showed, corroborating the blockage of smaller pores
(Table 1 and Figure 2B).
The Raman spectra of xW catalysts show major bands at 805, 715, 326 and 270
cm-1 which can be ascribed to crystalline WO3 nanoparticles, whose intensity rises with
the WO3 loading (Figure 3) [44, 45]. Thus, the bands at 805 and 715 cm-1 are associated
to bridging W – O – W stretching frequencies, whereas that the band at 270 cm-1 is
ascribed to the corresponding bending mode [46]. The weak band at 326 cm-1
corresponds to the bending vibration mode of O – W – O bonds of surface WOx species
[45]. Moreover, a broad and low intense band is discernible about 975 cm-1 which is
related to W=O vibration in hydrated polyoxotungstate clusters interacting with the
support [46, 47]. The presence of WOx species have been associated with strong
Brönsted acid sites [48]. Moreover, the characteristic bands of segregated ZrO2
nanoparticles for xW catalysts are not observed, therefore the existance of ZrO2
14

particles could be discarded as the XRD results previously shown. On the other hand,
after the acid treatment, the 20WPy catalysts do not display any noticeable band.
XPS technique has been employed to get insights about the chemical state of
tungsten in these catalysts (Table 2 and Supplementary Information, Fig. 3S). The
binding energy (BE) values for Si 2p (103.3-103.5 eV) and Zr 3d5/2 (183.2-183.6 eV)
are typical of zirconium doped siliceous framework, as previously reported [34, 49, 50].
In the O 1s region, an asymmetric band was observed implying the existence of
different oxygen environments which can be deconvoluted into two components: the
most intense at 532.7-532.9 eV is associated to oxygen in silica [51, 52], and that a
lower BE, 530.4-531.0 eV, corresponds to Si―O―Zr bonds (at 530.7 eV) [49]. For
WO3 loading higher than 10 wt%, the intensity of this latter band is increased and
slightly shifted to lower BE values (Table 2), which may be due to the contribution of O
1s from WO3 (530.4 eV) [53]. On the other hand, the BE of W 4f core level is within the
range 35.3-36.3 eV, which is ascribed to W6+ regardless the tungsten loading [53, 54].
The typical line splitting for the W 4f core level due to the spin – orbit coupling is even
difficult to discern, showing the W 4f photoelectronic peak a FWHM value of c.a. ~
4.12 eV. That means that tungsten could be distributed over many non-equivalent sites,
i.e. tungsten interacting with both zirconium and silicon sites, even the presence of W5+
cannot be rule out since 4f7/2 and 4f5/2 doublet falls in the same spectral region.
Likewise, it can be observed (Table 2) that the W/(Si+Zr) molar ratio determined
by XPS increases, as expected, with the WO3 loading up to 30 wt%. However, this
atomic ratio is lower than the corresponding bulk values (Table 2). Therefore, XPS
analysis suggests that tungsten oxide could be mainly located inside the porous structure
15

of the support. This fact was previously observed from the analysis of textural
parameters which showed a sharp decline of surface area of catalysts as the metallic
loading was increased. Another important conclusion arising from the XPS data is that
Si/Zr atomic ratio is increased as the tungsten oxide rises but reaching a plateau for
loadings higher than 20 wt%. This behavior seems to indicate that the tungsten oxide is
interacting primarily with zirconium domains and, as the metallic loading is increased,
these tungsten oxide particles grow in size and spread around such initial seeds of
tungsten species. These results are supported by both the XRD results, where the
formation of WO3 particles was detected even at lower tungsten loading, and Raman
analysis, which showed the presence of both structures in the xW catalysts.
UV–vis diffuse reflectance absorption spectra of catalysts have been carried
out in order to obtain information about the chemical nature and coordination states of
W (Figure 4). The UV-vis spectrum of the SiZr support does not resemble to the bulk
zirconia, due to the lack of the characteristic sharp absorption at 210 nm [52].
Nevertheless, it shows a band with a maximum at 293 nm and a broad band extending
up to roughly 375 nm. This latter spectral feature has been associated to oligomeric
zirconium species [55], whereas the band at 293 nm corresponds to isolated Zr4+ with
low coordination [55, 56].
From the UV-vis spectra of the xW catalysts, the existence of isolated [WO4]2-
species could be discarded, since its associated band at 230 nm is absent [57]. Instead,
xW catalysts show a broad band extending from 300 up to 450 nm, with a maximum at
about 300 nm, which is red-shifted by increasing the tungsten loading, and a shoulder at
~ 375 nm, indicating that different tungstate species are formed. This broad band is
16

consequence of the overlapping of bands arising from the low-condensed oligomeric
tungsten oxide species (300 nm) and WO3 crystallites (shoulder at 375 nm) [57-60].
These results corroborate the data extracted from Raman and XRD analyses, which
showed both the presence of polytungsted species and WO3 crystallites.
A key parameter in the glycerol dehydration is the catalyst acidity, and therefore
the total acidity of xW and 20WPy catalysts has been evaluated by means of NH3-TPD
(Table 1). Although the SiZr support shows a notable acidity mainly associated to Zr4+
ions with low coordination, this value is considerably enhanced after supporting WO3
species. Nevertheless, the total acidity decreases for WO3 loadings higher than 10 wt%,
being more pronounced for the 40W catalyst, whose acidity is even lower than that of
the SiZr support. Likewise, from the broad NH3-TPD profiles (Figure 5) it can be
inferred the existence of acid sites with a wide range of strength since it is generally
accepted that the acid strength depends on the NH3 desorption temperature, considering
weak (100–200 ºC), medium (200–400 ºC) and strong acid sites (>400 ºC) [49, 61].
Thus, the presence of tungsten oxide leads to an increase of the amount of ammonia
desorbed at high temperature, consequently pointing out the generation of stronger acid
sites. On the other hand, the H3PO4 treatment increases the acidity of 20W catalyst up to
797 μmoles NH3·gcat-1 for a W/P molar ratio equal to 1, decreasing for W/P= 0.6 and 0.2
(741 and 139 μmoles NH3·gcat-1, respectively). This drastic reduction may be due to the
serious damage suffered by the mesostructure of 20WP0.2 catalyst after this treatment.
On the other hand, pyridine adsorption coupled to FTIR spectroscopy has
been employed to evaluate the nature (Brönsted and/or Lewis) of acid sites, since the
selectivity to acrolein is correlated to the presence of Brönsted acid sites [30-32]. The
17

concentration of both types of acid sites has been estimated from the integrated
absorption at 1550 and 1450 cm−1, using the extinction coefficients EB=0.73 cm·µmol−1
and EL=1.11 cm·µmol−1 for Brönsted and Lewis sites, respectively [62]. The
concentration of Brönsted and Lewis acid sites measured by FTIR spectra after pyridine
adsorption are presented in Table 1S (Supplementary Information). The SiZr support
evidences both types of acid sites [63], and a maximum concentration is reached for the
10W catalyst, whereas for higher WO3 loadings, a decline is observed. For the xW
catalysts, the Lewis acid sites have been ascribed to W=O terminal groups [29], whereas
the Brönsted acid sites are associated to W – (OH) – W or W – OH bonds [29, 44]. In
the case of Zr-WOx/SiO2 catalysts, Kim et al. ascribed the presence of strong Brönsted
acid sites to the formation of Zr-WOx nanoclusters, whereas the generation of tungstates
of smaller sizes from such species led to a gradual decrease of Brönsted acidity [64].
Moreover, it has been reported that the formation of Brönsted acid sites depends on the
coverage of support surface, and the highest concentration of Brönsted acid sites is
attained for a monolayer of tungsten oxide [29, 44]. This fact has been explained by the
increasing number of boundaries existing between the domains of tungsten oxide,
because such W – (OH) – W or W – OH groups are more abundant in this region. On
the other hand, Kitano et al. have demonstrated that monoclinic WO3, detected in
tungsten supported on alumina, did not show Brönsted acidity [44]. The acid character
of WO3 nanoparticles was also studied by Cortés-Jácome et al., who found the
formation of WO3 nanocrystallites after annealing WO3/ZrO2 at 800 ºC, which causes a
concomitant decrease in the acid site density, being more reduced the amount of
Brönsted acid sites than Lewis ones [65]. Furthermore, Massa et al. observed the
formation of Lewis acid sites above the monolayer, due to the presence of
18

coordinatively unsaturated oxo sites [29]. These findings allow us explaining the
gradual change observed in the type and number of acid sites, firstly, by taking into
account the tungsten loading and the support surface, and assuming that all WO3 was
present as WO6 octahedra covering an area of 0.22 nm2 [44, 66], so none catalyst
reaches a WO3 monolayer (Table 2S). Secondly, Raman and UV-visible spectroscopies
have shown the existence of polytungstates on the support surface, and XPS analysis
has suggested that tungsten species are mainly interacting with zirconium domains.
Therefore, it seems reasonable to propose that the 10W catalyst exhibits the highest
amount of polytungstate species due to its highest amount of both types of acid sites
(Figure 6). As the tungsten loading is increased, these polytungstate species transform
into larger crystalline WO3 particles, and thereby decreasing the amount of Lewis and
Brönsted acid sites.
However, it can be observed that the B/B+L ratio increases with the tungsten
oxide content up to 30 wt%, showing the same trend that the intrinsic acidity (Table 1).
Interestingly, the Brönsted acid sites exhibit a higher strength than Lewis sites. This fact
could be related to the interaction between Zr – WOx as Yong Kim et al. previously
proposed [64].
Likewise, new Brönsted acid sites are generated after phosphoric acid treatment
in such a way that Lewis acid sites are not detected for a W/P molar ratio of 1 (Table 1
and 1S). However, the B/B+L ratio decreases when a higher amount of phosphoric acid
is incorporated compared to 20WP1, though being still higher than the corresponding
ratio of the 20W catalyst.
3.2. Catalytic Results
19

The WO3-based catalysts have been tested in the gas-phase dehydration of
glycerol at 325 ºC, during 8 h of time-on-stream (TOS). The only identified products
under the experimental conditions used in the present work were acrolein, acetaldehyde
and hydroxyacetone.
Firstly, the influence of the WO3 loading has been evaluated on the catalytic
performance (Table 3 and Figure 4S, in Supplementary Information). The catalytic
results point out that both SiZr support and xW catalysts are very active, although the
presence of WO3 ameliorates the selectivity to acrolein (Table 3). The 40W catalyst
shows the lowest stability among the xW catalysts, as well as lower acrolein selectivity
after 8 h of TOS (Time On Stream). The deactivation of 40W catalyst might be related
to its low concentration of acid sites, which are readily poisoned by carbonaceous
deposits, in spite of the fact that the carbon content of this spent catalyst is the lowest
among the tested catalysts (Table 3). The NH3-TPD plots (Figure 5) of the xW catalysts
showed a similar distribution of acid sites, being the most abundant those considered as
weak acid sites [67] (Fig. 5S). This similar low proportion of strong acid sites could
explain the analogous percentage of carbon found in spent catalysts. Furthermore,
several authors have highlighted the key role played in catalyst deactivation by both the
presence of strong acid sites [26, 37, 67, 68] and the porous network [69-73]. Therefore,
the deactivation of xW catalysts would be influenced not only by the distribution of acid
sites but also by the mesoporosity of catalysts, which facilitates the diffusion of
reactants and products in the pore network.
The turnover frequency (TOF) at time zero was calculated in order to minimize
the effect of catalytic deactivation on the intrinsic catalytic activity. This extrapolation
20

was made by supposing a first order deactivation kinetic [70, 72], X(t) = X(0) exp (-kd
t), where X(t) and X(0) accounts for consumption of glycerol at time t and t = 0,
respectively, and kd represents the deactivation constant. Thus, the TOF was calculated
by dividing the rate of glycerol consumption per the total acidity measured by NH3–
TPD (Table 2S). It could be expected that catalysts with the most suitable acid
properties should exhibit the highest TOF values. However, the highest TOF value was
attained for the 40W catalyst, nearly more than twice the rest of catalysts, in spite of its
lower total acidity. Nevertheless, it should be considered that the intrinsic surface
acidity is similar for the xW catalysts (Table 1), and therefore, the location of acid sites
might play an important role. This fact can be drawn in Figure 7, where both TOF(0)
and total acidity are represented as a function of external surface of catalysts. As
expected, the total acidity runs in parallel to the external surface, but the TOF(0) follows
the inverse trend, that is, it decreases as the external surface increases. This would point
out that the acid sites of the 40W catalyst could be located preferentially in the outer
surface, ready to activate the glycerol molecules, whereas the rest of catalysts could
have part of the acid sites inside the pores, not being easily accessible to reactant
molecules. The evolution of mmol of glycerol converted per unit time and square meter
of external surface (mmol Glycerol · h-1 m-2) as a function of W atoms per square
nanometer gives rise to a straight line, corroborating the aforementioned assumption
(Fig. 6S). This availability of acid sites would be also related to the velocity of catalyst
deactivation, since this is faster for the 40W catalyst, which shows the highest value of
kd (Table 2S). Choi et al. found that TOF values were higher for aluminosilicates with a
low surface acid density, since in that case, if the acid sites were sufficiently far apart,
21

they could be more easily accessible for glycerol molecules and therefore the interaction
between neighbor glycerol molecules would be avoided [70].
Regarding the acrolein selectivity, it has been pointed out that Brönsted acid
sites are more adequate than Lewis ones in order to obtain the highest acrolein yield [29,
69, 70, 74, 75], whereas both Lewis acid and basic sites are involved in the formation of
hydroxyacetone [68, 76, 77]. Acetaldehyde selectivity stems from different routes
including the decomposition of hydroxyacetone over acidic sites [11, 26], the
homogenous decomposition of 3-hydroxypropanal desorbed from an acid site [69] or at
expense of acrolein at high temperature [22]. Table 3 gathers the products distribution
after 2 and 8 h of TOS. The acrolein yield is increased with the tungsten loading, but
both hydroxyacetone and acetaldehyde yields do not follow the same trend, since they
remain constant regardless of TOS. Thus, the 20W catalyst shows the highest acrolein
yield, reaching a value of 41 mol% after 2 h, and maintaining a 38 mol% after 8 h of
TOS. Thus, Chai et al. found that tungsten oxide dispersed on ZrO2 are active for
dehydration of glycerol to acrolein, being maximum the acrolein selectivity when the
WO3 loading was in range of 15–30%, similar to the data obtained for xW catalysts
[22]. As it has been previously mentioned, glycerol is converted into acrolein on
Brönsted acid sites, and therefore, it is expected that the selectivity towards acrolein
enhances with the B/B+L ratio. Figure 6 shows that both B/B+L ratio Brönsted ratio and
acrolein yield follow a similar trend, and the maximum acrolein yield would be
achieved for a catalyst with a WO3 loading within the range 20–30 wt%. Although these
catalysts predominantly possess Lewis acid sites, the main product detected is acrolein
instead of hydroxyacetone. This fact was already observed in the case of a zirconium
doped MCM-41 silica, and it was justified by the generation of new Brönsted acid sites
22

due to the presence of water (the solvent of the glycerol) which was able to hydrolyze
some Si – O – Zr bonds forming new Brönsted acid sites [49]. In the case of tungsten
based catalysts, it has been proved that alcohols, hydrogen or other reductants can
generate W(6-δ) from W6+ [48]. This partial reduction implies an excess of negative
charge which can be neutralized by protonation of oxygen atoms, thus generating
Brönsted acid sites. In this dehydration process, glycerol, acrolein, acetaldehyde and
other hydrocarbons could cause such partial reduction, generating new species of
reduced tungsten which would have a Brönsted character and contributing to the
formation of acrolein [75].
The comparison of different catalytic systems is not straightforward, mainly due
to the different experimental conditions used (reaction temperature and time,
concentration of glycerol solution, WHSV or GHSV, carrier gas, inert or oxygen,
among others). For instance, although Lauriol-Garbey et al. previously found high
values of glycerol conversion and acrolein yield (100% and 76%, respectively) in
presence of WO3/ZrO2 catalysts doped with SiO2 after 30 h at 300 ºC [26], they
employed a higher catalyst loading, in such a way that glycerol fed per time unit and
mass of catalyst, 0.33 µmol glycerol·gcat-1·s-1, was much lower than 3.91 µmol
glycerol·gcat-1·s-1 used in this work. Therefore, the mesoporous structure of SiZr support
improves the catalytic performance of tungsten oxide, generating 1.60 µmol
acrolein·gcat-1·s-1 from 3.79 µmol converted glycerol·gcat-1·s-1 at 325 ºC after 2 h of TOS
in presence of the 20W catalyst. Several authors have also studied different WO3/ZrO2
based catalysts for dehydration of glycerol, usually reaching high values of glycerol
conversion and acrolein yield, but using different feeding of glycerol per time and mass
of catalyst, as for example Ginjupalli et al. (1.51 µmol converted glycerol·gcat-1·s-1 and
23

1.13 µmol produced acrolein·gcat-1·s-1) [23], Massa et al. (2.85 µmol converted
glycerol·gcat-1·s-1 and 2.24 µmol produced acrolein·gcat-1·s-1) [29] and Chai et al. (3.79
µmol converted glycerol·gcat-1·s-1 and 2.67 µmol produced acrolein·gcat-1·s-1) [22].
Consequently, the 20W catalyst can be considered as an interesting catalyst due to its
stability and productivity to convert glycerol to acrolein, since the amount of acrolein
produced per time unit and mass of catalyst is an important parameter to its possible
use.
The influence of reaction temperature on the catalytic behavior has been also
studied for the 20W catalyst (Figure 8). It is observed that glycerol conversion increases
with the reaction temperature from 300 to 350 ºC, although similar values of acrolein
yield are attained after 2 h of TOS. Nonetheless, acrolein yield is slightly enhanced at
higher reaction times when the reaction temperature is increased from 300 up to 325 ºC,
but it does not follow this trend when the temperature is further risen up to 350 ºC,
being even more important its deactivation after 8 h of TOS. Thus, a higher amount of
non-identified products is obtained when reaction temperature is 350 ºC. On the other
hand, acetaldehyde yield increases with the reaction temperature from 11% at 300ºC up
to 16% at 350 ºC, after 2 h of TOS, similarly to that reported by other authors [49, 78].
With respect to hydroxyacetone yield, insignificant differences are observed in the
range of 300-350 ºC.
On the other hand, it has been previously demonstrated that treatment with
phosphoric acid improves the acidity of some metal oxides and the amount of Brönsted
acid sites, which are more selective to acrolein [34-36]. Thus, the 20W catalyst has been
treated with different amounts of H3PO4 and tested in dehydration of glycerol at 325 ºC
(Figure 9). It can be observed that deactivation is more severe for catalysts treated with
24

phosphoric acid, mainly for the 20WP0.2 catalyst, which may be due to its lowest total
acidity. In regard to acrolein yield, an increase is detected with H3PO4 content for short
reaction times, in such a way that a value of 51% is attained for 20WP0.2 catalyst after
2 h of TOS, in spite of a lower glycerol conversion. This increase is probably related to
the Brönsted acid sites concentration of these materials, since the B/B+L molar ratio
respect to total number of acid sites is higher after treatment with phosphoric acid
(Table 1), and it has been previously demonstrated that this type of acid sites are more
selective to acrolein [30, 31]. However, this more pronounced deactivation, in
comparison with the 20W catalyst, limits the potential of the treatment with phosphoric
acid of WO3 based catalysts, under these experimental conditions, in spite of the slight
increase of acrolein yield after 2 h of TOS, because of the 20W catalyst stays stable
after 8 h of TOS. Moreover, it should be noted that acetaldehyde yield is almost zero
after H3PO4 treatment, as previously detected by Cecilia et al. for vanadium oxide based
catalysts treated with phosphoric acid [34]. On the other hand, the hydroxyacetone yield
remains virtually unchanged after this treatment (Table 3).
In order to justify the stability of the 20W catalyst in the dehydration of glycerol
at 325 ºC after 8 h of TOS, the spent catalyst (R20W) has been characterized by XPS
and XRD (Table 2). The binding energy and atomic ratio values are similar to those of
the fresh 20W catalyst, although the triclinic WO3 phase is not already detected in the
diffractogram of the spent catalyst (Fig. 10A and 10C), being only observed the
monoclinic WO3 phase, mainly in the region of 31-37º (Figure 10C). Therefore, it can
be affirmed that the reaction conditions affect to the crystalline phases of the catalyst,
but those changes do not influence on the catalytic performance since deactivation was
not observed for the 20W catalyst. With regard to the spent 20WP0.2 catalyst, although
25

the values of BE are also maintained after reaction (Table 2), significant differences for
atomic concentrations are detected by XPS. Thus, this spent catalyst displays a
considerable amount of carbon (59%), which would explain its severe deactivation and
the decrease of the Si/Zr molar ratio by the presence of carbonaceous deposits on the
catalyst surface. However, an increase of the W/Si+Zr and W/P molar ratios is observed
after catalytic test, so it is likely that a sintering of W and P phases takes place during
reaction. This fact has been corroborated by XRD (Fig. 10B), since new diffraction
lines belonging to W and W-P phases are observed (W17O47, PDF code 01-079-0171;
H34O38W7, PDF code 98-020-0767 and O20PW6, PDF code 98-005-4054) which were
not present in the fresh 20WP0.2 catalyst, indicating that they are amorphous or well
dispersed on the catalyst surface before reaction, but they crystalize after 8 h of TOS at
325 ºC. Thus, the presence of these crystalline phases could explain the fast deactivation
that this catalyst suffers during dehydration of glycerol.
Likewise, the thermal stability of carbonaceous deposits in both catalysts, 20W
and 20WP0.2, has been studied by TGA (Figure 7S). Although a first weight loss is
found at low temperature probably due to adsorbed water, the main contribution is
located between 300-650 ºC, indicating the presence of different types of carbonaceous
deposits on the catalysts surface. In the case of the 20W catalyst, the range of
temperatures is narrower and it is centered about 500 ºC, which may be due to the
existence of unsaturated hydrocarbon or polyaromatic molecules more resistant to
combustion [49]. However, the 20WP0.2 catalyst also displays a shoulder between 300-
400 ºC which could be ascribed to carbon deposits easily oxidized like aliphatic
molecules [30]. Taking into account these results, the regeneration of spent catalysts
26

could be accomplished by thermal treatment at 550 ºC under oxygen, thus allowing the
removing of the carbonaceous deposits.
4.
Conclusions
A family of tungsten oxide and tungsten oxide-phosphorous supported on a
zirconium doped mesoporous SBA-15 silica has demonstrated to be active catalysts for
glycerol dehydration. The incorporation of WO3 to the SiZr support increases the total
acidity and the Brönsted acid sites proportion, improving the stability in glycerol
dehydration. Thus, the 20W catalyst displays the highest glycerol conversion and
acrolein yield values (97% and 41% after 2 h, and 90% and 38% after 8 h of TOS,
respectively, at 325 ºC) which may be related to the existence of WO3-ZrO2 phases on
catalyst surface, as deduced from XPS data. Although the H3PO4 treatment increases the
total acidity and the B/B+L molar ratio, when small amounts of acid are utilized,
enhancing acrolein yield at the beginning of reaction, the deactivation of 20WPy
catalysts is much more pronounced due to the important reduction of specific surface
area and total acidity, decreasing the number of available acid sites, mainly for
20WP0.2 catalyst.
Acknowledgements
The authors are grateful to financial support from the Spanish Ministry of
Economy and Competitiveness (CTQ2012-38204-C03-02 project), Junta de Andalucía
(RNM-1565) and FEDER (European Union) funds.
27

References
[1]
[2]
E. Shahid, Y. Jamal, Renew. Sust. Energ. Rev. 15 (2011) 4732-4745.
E. Aransiola, T. Ojumu, O. Oyekola, T. Madzimbamuto, D. Ikhu-Omoregbe,
Biomass Bioenerg. 61 (2014) 276-297.
[3]
B. Katryniok, S. Paul, V. Belliere-Baca, P. Rey, F. Dumeignil, Green Chem. 12
(2010) 2079-2098.
[4]
[5]
M. Ayoub, A. Abdullah, Renew. Sust. Energ. Rev. 16 (2012) 2671-2686.
M. Pagliaro, M. Rossi, The future of glycerol: New Uses of a Versatile Raw
Material, (2008).
[6]
C. Callam, S. Singer, T. Lowary, C. Hadad, J. Am. Chem. Soc. 123 (2001)
11743-11754.
[7]
B. Katryniok, S. Paul, M. Capron, F. Dumeignil, ChemSusChem 2 (2009) 719-
730.
[8]
[9]
L. Liu, X. Ye, J. Bozell, ChemSusChem 5 (2012) 1162-1180.
K. Bowmer, G. Smith, Weed Res. 24 (1984) 201-211.
[10] G.W. Keulks, L.D. Krenzke, T.M. Notermann, Selective Oxidation of
Propylene, Adv. Catal., 1979, pp. 183-225.
[11] A. Corma, G. Huber, L. Sauvanauda, P. O'Connor, J. Catal. 257 (2008) 163-171.
[12] C. Jia, Y. Liu, W. Schmidt, A. Lu, F. Schuth, J. Catal. 269 (2010) 71-79.
[13] A. de Oliveira, S. Vasconcelos, J. de Sousa, F. de Sousa, J. Filho, A. Oliveira,
Chem. Eng. J. 168 (2011) 765-774.
[14] C. Carrico, F. Cruz, M. Santos, H. Pastore, H. Andrade, A. Mascarenhas,
Micropor. Mesopor. Mat. 181 (2013) 74-82.
[15] L. Shen, Y. Feng, H. Yin, A. Wang, L. Yu, T. Jiang, Y. Shen, Z. Wu, J. Ind.
Eng. Chem. 17 (2011) 484-492.
[16] E. Kraleva, R. Palcheva, L. Dimitrov, U. Armbruster, A. Bruckner, A.
Spojakina, J. Mater. Sci. 46 (2011) 7160-7168.
[17] R. Liu, T. Wang, Y. Jin, Catal. Today 233 (2014) 127-132.
[18] F. Cavani, S. Guidetti, L. Marinelli, M. Piccinini, E. Ghedini, M. Signoretto,
Appl. Catal. B-Environ. 100 (2010) 197-204.
28

[19] L. Yang, J. Joo, N. Kim, K. Jung, J. Yi, Korean J. Chem. Eng. 27 (2010) 1695-
1699.
[20] N. Shiju, D. Brown, K. Wilson, G. Rothenberg, Top. Catal. 53 (2010) 1217-
1223.
[21] L. Tao, B. Yan, Y. Liang, B. Xu, Green Chem. 15 (2013) 696-705.
[22] S. Chai, B. Yan, L. Tao, Y. Liang, B. Xu, Catal. Today 234 (2014) 215-222.
[23] S. Ginjupalli, S. Mugawar, N. Rajan, P. Balla, V. Komandur, Appl. Surf. Sci.
309 (2014) 153-159.
[24] S. Chai, L. Tao, B. Yan, J. Vedrine, B. Xu, Rsc Adv. 4 (2014) 4619-4630.
[25] R. Znaiguia, L. Brandhorst, N. Christin, V. Baca, P. Rey, J. Millet, S. Loridant,
Micropor. Mesopor. Mat. 196 (2014) 97-103.
[26] P. Lauriol-Garbey, S. Loridant, V. Belliere-Baca, P. Rey, J. Millet, Catal.
Commun. 16 (2011) 170-174.
[27] M. Massa, A. Andersson, E. Finocchio, G. Busca, J. Catal. 307 (2013) 170-184.
[28] K. Omata, S. Izumi, T. Murayama, W. Ueda, Catal. Today 201 (2013) 7-11.
[29] M. Massa, A. Andersson, E. Finocchio, G. Busca, F. Lenrick, L. Wallenberg, J.
Catal. 297 (2013) 93-109.
[30] A. Alhanash, E. Kozhevnikova, I. Kozhevnikov, Appl. Catal. A-Gen. 378 (2010)
11-18.
[31] S. Chai, H. Wang, Y. Liang, B. Xu, Green Chem. 9 (2007) 1130-1136.
[32] G. Foo, D. Wei, D. Sholl, C. Sievers, ACS Catal. 4 (2014) 3180-3192.
[33] J. Ekhsan, S. Lee, H. Nur, Appl. Catal. A-Gen. 471 (2014) 142-148.
[34] J.A. Cecilia, C. García-Sancho, J.M. Mérida-Robles, J. Santamaría-González, R.
Moreno-Tost, P. Maireles-Torres, Catal. Today 254 (2015) 43-52.
[35] Y.Y. Lee, K.A. Lee, N.C. Park, Y.C. Kim, Catal. Today 232 (2014) 114-118.
[36] V. Struzhko, E. Senchylo, G. Kosmambetova, V. Gritsenko, V. Solomakha, P.
Strizhak, Theor. Exp. Chem. 49 (2014) 390-395.
[37] G. Srinivasa Rao, N. Pethan Rajan, M. Hari Sekhar, S. Ammaji, K.V.R. Chary,
J. Mol. Catal. A-Chem. 395 (2014) 486-493.
[38] W. Suprun, H. Papp, Catalysis in Industry 3 (2011) 70-75.
29