A tailored catalyst for the sustainable conversion of glycerol to acrolein: Mechanistic aspect of sequential dehydration

Article abstract of DOI:10.1002/cssc.201402057

Developing a catalyst to resolve deactivation caused from coke is a primary challenge in the dehydration of glycerol to acrolein. An open-macropore- structured and Bronsted-acidic catalyst (Marigold-like silica functionalized with sulfonic acid groups, MS-FS) was synthesized for the stable and selective production of acrolein from glycerol. A high acrolein yield of 73 % was achieved and maintained for 50 h in the presence of the MS-FS catalyst. The hierarchical structure of the catalyst with macropores was found to have an important effect on the stability of the catalyst because coke polymerization and pore blocking caused by coke deposition were inhibited. In addition, the behavior of 3-hydroxypropionaldehyde (3-HPA) during the sequential dehydration was studied using density functional theory (DFT) calculations because 3-HPA conversion is one of the main causes for coke formation. We found that the easily reproducible Bronsted acid sites in MS-FS permit the selective and stable production of acrolein. This is because the reactive intermediate (3-HPA) is readily adsorbed on the regenerated acid sites, which is essential for the selective production of acrolein during the sequential dehydration. The regeneration ability of the acid sites is related not only to the selective production of acrolein but also to the retardation of catalyst deactivation by suppressing the formation of coke precursors originating from 3-HPA degradation.

Full text of DOI:10.1002/cssc.201402057

CHEMSUSCHEM
FULL PAPERS
DOI: 10.1002/cssc.201402057
A Tailored Catalyst for the Sustainable Conversion of
Glycerol to Acrolein: Mechanistic Aspect of Sequential
Dehydration
[
a]
[a]
[a]
[a]
[b]
Danim Yun, Tae Yong Kim, Dae Sung Park, Yang Sik Yun, Jeong Woo Han, and
[a]
Jongheop Yi*
Developing a catalyst to resolve deactivation caused from coke
is a primary challenge in the dehydration of glycerol to acrole-
in. An open-macropore-structured and Brønsted-acidic catalyst
quential dehydration was studied using density functional
theory (DFT) calculations because 3-HPA conversion is one of
the main causes for coke formation. We found that the easily
reproducible Brønsted acid sites in MS-FS permit the selective
and stable production of acrolein. This is because the reactive
intermediate (3-HPA) is readily adsorbed on the regenerated
acid sites, which is essential for the selective production of
acrolein during the sequential dehydration. The regeneration
ability of the acid sites is related not only to the selective pro-
duction of acrolein but also to the retardation of catalyst deac-
tivation by suppressing the formation of coke precursors origi-
nating from 3-HPA degradation.
(Marigold-like silica functionalized with sulfonic acid groups,
MS-FS) was synthesized for the stable and selective production
of acrolein from glycerol. A high acrolein yield of 73% was ach-
ieved and maintained for 50 h in the presence of the MS-FS
catalyst. The hierarchical structure of the catalyst with macro-
pores was found to have an important effect on the stability of
the catalyst because coke polymerization and pore blocking
caused by coke deposition were inhibited. In addition, the be-
havior of 3-hydroxypropionaldehyde (3-HPA) during the se-
Introduction
[
10]
Since the demand for a clean and green technology has in-
creased, countless efforts have been made to replace conven-
tional petrochemical processing. Biomass could be a solution
to satisfy the needs of the times because it has a potential for
producing fine chemicals or building blocks without increasing
ly viable. Hence, the conversion of glycerol and the related
[8–11]
catalysts have been extensively studied.
Thanks to previous studies, a better understanding has been
developed regarding the mechanism of glycerol conver-
[
10,12,13]
sion
lysts.
and the effectiveness of Brønsted acid cata-
During glycerol conversion, a variety of reactions
[
1,2]
[7,14–17]
greenhouse gas emissions.
Therefore, various studies con-
cerning the production of value-added chemicals from bio-
can take place, but the desired reaction are two sequential de-
hydrations consisting of the dehydration of glycerol to 3-hy-
droxypropionaldehyde (3-HPA) and 3-HPA to acrolein. When
Brønsted acid catalysts are used, the first dehydration, known
[3–6]
based compounds have been reported in recent years.
In
particular, the conversion of glycerol to acrolein has attracted
a considerable amount of attention as an alternative route to
the production of acrolein from petroleum-based propyl-
[14,18]
as the rate-determining step,
can be easily induced. This
[
7–9]
ene.
Furthermore, an increasing cost ratio of acrolein to
can be attributed to the secondary OH group of glycerol being
preferentially adsorbed onto the Brønsted acid sites, which
glycerol in recent years makes the conversion also economical-
[7,14]
should occur prior to the dehydration to 3-HPA.
Although
Brønsted acid catalysts show a high activity and selectivity for
acrolein early in the reaction, the catalysts are quickly deacti-
+
+
[
a] D. Yun, T. Y. Kim, D. S. Park, Y. S. Yun, Prof. J. Yi
World Class University (WCU) program of
chemical Convergence for Energy&Environment
Institute of Chemical Processes
[8,19]
vated due to extensive coke deposition.
Therefore, solving
the problem of coke formation is necessary to achieve glycerol
dehydration on a commercial scale.
School of Chemical and Biological Engineering
Seoul National University
Seoul 151-741 (Republic of Korea)
Fax: (+82) 2-885-6670
There are two approaches to solve the problem of this type
of deactivation. One involves reducing the amount of coke
produced during the reaction and the other is enhancing in-
trinsic resistivity to coke deposition. For the former, a key to
the solution involves 3-HPA conversion. 3-HPA is considered to
be a major intermediate formed during the reaction as it is
E-mail: jyi@snu.ac.kr
[
b] Prof. J. W. Han
Department of Chemical Engineering
University of Seoul
Seoul 130-743 (Republic of Korea)
[12,15]
a precursor of acrolein.
It is very reactive, and the addition-
+
al undesired conversions can easily take place, resulting in
[
] These authors contributed equally to the work.
[10,12,15]
coke formation.
However, the behavior of 3-HPA during
Supporting Information for this article is available on the WWW under
http://dx.doi.org/10.1002/cssc.201402057.
glycerol dehydration is not well understood currently because
ꢀ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2014, 7, 2193 – 2201 2193

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
the detection of 3-HPA is almost impossible during the reac-
tion due to the highly reactive nature of 3-HPA. A theoretical
study could contribute to a better understanding of this prob-
lem, but up to now such studies have been limited to the reac-
[
13,14]
tion mechanism of glycerol.
It should be possible to ob-
serve the behavior of 3-HPA during the reaction using density
functional theory (DFT) calculations.
The amount of coke produced can be reduced through the
realization of appropriate active sites, but coke formation
cannot be entirely prohibited considering the overall reaction.
Therefore, enhancing the intrinsic resistivity of a catalyst to
coke formation is essential for achieving a stable production of
acrolein. In this regard, several attempts to control the textural
properties of catalysts have been reported to improve coke re-
Scheme 1. Description of MS-FS catalyst on various scales.
The functionalization of the surface with propanesulfonic acid
groups confers pure Brønsted-acidic characteristics to the cata-
lyst. The importance of Brønsted acid type has been empha-
[7,14–17]
sized in previous reports.
In the case of acidity, a narrow
[
20]
sistivity. Tsukuda et al.
studied the influence of mesopore
distribution of acid strength should be advantageous for the
selective formation of the product. In this regard, the propane-
size of the silica support on catalytic stability. They demonstrat-
ed that the silica with the largest pores had the best re-
sistance, thereby emphasizing the importance of pore
size in the stable production of acrolein. On the other
[
21]
Table 1. Effect of modification of the terminal carbon atom on DEDPE of propane-
hand, Hoang et al. examined suitable pore structures
in terms of residence time of reactant or intermediate.
They found that compounds containing alkyl and aro-
matic groups were primarily formed over a catalyst with
a longer residence time, which resulted not only in
a low selectivity for acrolein but also in a rapid deactiva-
tion of the catalyst by coke deposition. Although en-
hancing effects caused by large pores and suitable pore
structure were deduced, the problem associated with
deactivation was not solved.
[
a]
sulfonic acid and sulfuric acid: Cluster-based DFT calculations
[
b]
[c]
Model Structure Terminal DEDPE
Variation
ꢀ
1
ꢀ1
group
–
[kJmol ] [kJmol
]
CH
3
3
(CH
2
2
)
)
2
3
SO
SO
3
H
H
475.9
475.8
467.2
–
CH
(CH
3
CH
3
0.1
8.7
Herein, we propose the use of an open-macropore-
structured and Brønsted-acidic silica (Marigold-like silica
functionalized with propanesulfonic acid groups, MS-FS)
as a sustainable and selective catalyst for acrolein pro-
duction. The appropriate Brønsted acid sites for the se-
quential dehydration of glycerol were obtained through
functionalization with propanesulfonic acid groups. In
the dehydration of glycerol, the open-macropore struc-
ture of the catalyst showed a clear enhancement in re-
sistance to hard coke formation and to pore blocking by
coke. Using DFT calculations, we found a relationship be-
tween the behavior of 3-HPA and the recyclability of
Brønsted acid sites during sequential dehydration. The
MS-FS catalyst permitted the deprotonated Brønsted
acid sites to be easily regenerated, which led to the se-
lective dehydration of 3-HPA into acrolein. Moreover, the
formation of coke precursors originating from the degra-
dation of 3-HPA was inhibited. For these reasons, the
MS-FS catalyst showed outstanding selectivity for acrole-
in formation and was also quite stable (with yields ap-
proaching 73% after 50 h).
SiH
3
(CH
2
)
3
SO
3
H
SiH
3
AlH
2
(CH
2
)
3
SO
3
H
AlH
2
472.7
451.8
427.2
3.2
24.1
–
ZrH
3
(CH
2
)
3
SO
3
H
ZrH
–
3
H
2
SO
4
SiH
3
SO
4
H
SiH
3
422.4
369.7
323.3
4.8
42.5
AlH SO H
2
4
AlH2
Results and Discussion
ZrH
3
SO
4
H
ZrH
3
103.9
We designed and synthesized MS-FS (Scheme 1). The
marigold-like structure accommodates the large open-
pore space between the silica walls and can enhance
[a] Computational details using cluster-based DFT calculation can be found in the
ꢀ
Supporting Information. [b] With respect to following reaction: HA+NH
3
!A +
+
[22]
4 3 2 2 3 2 4
NH . [c] Difference of DEDPE compared to CH (CH ) SO H or H SO .Table 2
the mass transport of both the reactant and product.
ꢀ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2014, 7, 2193 – 2201 2194

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
sulfonic acid moiety can ensure uniformity of acid strength be-
cause the propyl chain in the moiety separates the acid site
from the surface of the catalyst. The variability of acid site
[
23,24]
strength induced from the surface structure
can be dimin-
ished, resulting in acid sites with identical strengths. We con-
firmed this effect of propanesulfonic acid group by performing
simple cluster-based DFT calculations (Table 1). The deproto-
nation energy (DEDPE, related to the strength of acid sites)
of propanesulfonic acid group was essentially unaltered
ꢀ
1
(
ꢁ24 kJmol ) by the modification with a moiety attached to
the terminal carbon atom. In contrast, a similar modification
with sulfuric acid resulted in a significant change in DE
DPE
ꢀ
1
(
ꢁ104 kJmol ).
Figure 1 shows electron microscope images of the MS-FS
catalyst. It can be clearly seen that MS-FS is a nanosphere with
a size ranging from 400–600 nm. The surface of MS-FS is cov-
ered by winding walls of silica, and deep canyons are formed
between the walls (Figure 1a and c). In SEM images, the esti-
mated distance between the silica walls ranges from 10 nm to
over 100 nm. The thickness of each wall is estimated to be
13.1–17.5 nm based on Figure 1a, but it can be thinner be-
cause of the platinum coating required for SEM analysis. In Fig-
ure 1b, the wall thickness at the edge is about 3–5 nm. In the
same image, pores formed between the silica walls are clearly
observed. These pores appear to be formed from the inside to
the outside of the nanospheres. In addition, very small white
dots and fringes can be seen over the entire nanosphere. They
are 2–5 nm-sized mesopores distributed in the silica walls and
the inside of hierarchical structured pores, which was further
confirmed by nitrogen physisorption analysis. Interestingly, al-
though the propanesulfonic acid moieties were used to func-
tionalize the catalyst, the overall catalyst structure was almost
the same as that for a nonfunctionalized silica nanosphere
Figure 1. Electron microscope images of MS-FS catalyst: (a,c) SEM and
(b) TEM.
are regularly distributed not only on the inside of hierarchically
structured pores but also in the silica walls. This result is in
good agreement with the observation of white dots and fring-
es in the TEM image (Figure 1b). The amount of macropores
decreased because the thickening of the walls made the dis-
tance between the walls shorter. From these results, we con-
firmed that functionalization was successfully carried out and
that hierarchically structured meso–macropores were well de-
veloped in the MS-FS catalyst.
[Marigold-like silica (MS), Figure S1].
Details concerning the pore structure and textural properties
of the MS-FS catalyst were confirmed by nitrogen physisorp-
tion analysis. In the nitrogen isotherm plot (Figure 2a) of the
MS and MS-FS catalysts, the broad hysteresis loop from 0.4 to
1
.0 (P/P ) indicates that meso- and macrosizded pores are pres-
0
ent in the catalysts. A significant increase in adsorbed nitrogen
in the range of 0.7 to 1.0 clearly shows that there is a consider-
able amount of macropores in both catalysts. The most nota-
ble difference between the MS and MS-FS catalysts is
Successful functionalization of MS-FS was also confirmed by
elemental analysis (Table S1). In MS-FS, the weight percentage
of sulfur was estimated to be 1.4%. In addition, the structure
the amount of adsorbed nitrogen near P/P =0.4. In
0
the case of nonfunctionalized MS, a sharp increase in
adsorbed nitrogen was observed, which implies the
presence of large amounts of mesopores. In contrast,
functionalized MS-FS does not show this sharp in-
crease. The pore size distribution clearly shows this
difference between the absence and presence of
functional groups (Figure 2b). The amount of 2–
5
nm-sized mesopores was definitely decreased as
a result of functionalization. This can be attributed to
the length of the propanesulfonic acid moiety (0.68–
0
.75 nm), which is sufficiently long to block small
pores. In addition, this distinct difference in the
amount of mesopores indicates that small mesopores MS-FS(c) and MS(a).
2
Figure 2. (a) N adsorption-desorption isotherm and (b) pore size distribution curve of
ꢀ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2014, 7, 2193 – 2201 2195

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
of the functional groups was confirmed by FTIR analysis. In Fig-
ꢀ
1
ure S2, there is no peak at 2575 cm , which indicates that the
mercaptopropyl groups were fully oxidized to sulfonic acid
[
25,26]
ꢀ1[27,28]
groups.
CꢀH vibrations at 2970 and 2873 cm
and
ꢀ
1[28]
a hydroxyl band between 3200 and 3400 cm
were identi-
fied and attributed to the presence of propyl groups and sul-
fonic acid groups, respectively.
The activity of the MS-FS catalyst was investigated for glyc-
erol dehydration. Although the reaction was run for 50 h, the
glycerol conversion was maintained at 100% and the selectivi-
ty for acrolein was high (72.8%; Figure 3). Considering previous
studies, which attempted to solve the deactivation problem by
Figure 3. Glycerol conversion (*) and acrolein selectivity (
) of MS-FS.
&
The used catalysts were analyzed by using the nitrogen ad-
sorption method, and a dramatic change in porosity was
found. Figure 4a and b shows the pore size distribution of
[
29]
[30–32]
co-feeding oxygen or adding a promoter,
the selectivity
for acrolein and the stability of MS-FS are outstanding. To fur-
ther investigate the reason why
MS-FS showed such a high sta-
bility, we carried out compara-
tive studies with regard to i) hier-
archically structured meso–mac-
ropores and ii) characteristics of
Brønsted acid sites.
Extensive coke formation on
the acid sites is a typical prob-
[
8,19]
lem in glycerol dehydration.
Coke deposition could lead to
the blockage of pore entrances,
resulting in rapid deactivation.
Therefore, proper textural prop-
erties such as a large pore size
and hierarchical pore structures
are the major factors in terms of
catalyst stability. For this reason,
we selected zeolite HZSM-5 as
a control group with micropo-
rous structures and compared
the catalytic activities of both
catalysts. Fresh HZSM-5 has
a crystal structure, as evidenced
by high-resolution (HR)-TEM (Fig-
ure S3) and XRD (Figure S4) anal-
ysis. Nitrogen adsorption–de-
sorption isotherms showed no
distinct hysteresis loop (Fig-
ure 4c), which indicates that the
fresh HZSM-5 catalyst has a typi-
cal microporous structure unlike Figure 4. Pore size distribution of (a) HZSM-5 and (b) MS-FS catalysts for fresh samples (c) and those used for
1
0 (a) and 30 h (b). Nitrogen adsorption–desorption isotherms of (c) HZSM-5 and (d) MS-FS catalysts.
the MS-FS catalyst.
As shown in Table 2, HZSM-5
was initially active and selective
for the production of acrolein (conversion 92.7% and selectivi-
[
8]
Table 2. Catalytic activity of MS-FS and HZSM-5 catalysts in the dehydra-
ty 73.0%), similar to previous results. However, the catalyst
was rapidly deactivated even though its selectivity was main-
tained. The conversion of glycerol after a 10 h reaction was
only 21.9%. The major by-product, acetol, formed with the
same selectivity (9.3%) using MS-FS, and other by-products
formed at yields of nearly 0% for both catalysts.
tion of glycerol with regard to TOS. Amount of catalyst: 0.3 g; glycerol
feed rate: 2.0 mLh ; reaction temperature: 2508C.
ꢀ1
Catalyst
Glycerol conversion [%]
Selectivity [%]
TOS=1–2 h TOS=9–10 h acrolein acetol acetaldehyde
HZSM-5
MS-FS
92.7
100.0
21.9
100.0
73.0
73.4
9.3
9.3
0.0
0.0
ꢀ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2014, 7, 2193 – 2201 2196

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
HZSM-5 and MS-FS, respectively. In the case of HZSM-5, after
10 h reaction, the micro-sized pores were completely blocked
due to coke deposition. In contrast, macropores of 20–100 nm
in MS-FS were still present even after 30 h reaction. The differ-
ence between HZSM-5 and MS-FS is also presented in the iso-
therms of fresh and used samples (Figure 4c and d). The
amount of adsorbed nitrogen for used HZSM-5 was significant-
ly decreased in the range below 0.2 (P/P ; region related to mi-
0
cropores), and the isotherm was down-shifted. On the other
hand, most of the nitrogen adsorption (P/P >0.6) for the used
0
MS-FS was still maintained. The results explain why the MS-FS
catalyst showed a remarkable stability in spite of extensive
coke formation. The hierarchically structured meso–macro-
pores in MS-FS confer an advantage in that pore blocking is re-
duced, resulting that acid sites located inside the pores can
remain functioning for a long reaction time.
In addition, the properties of the coke on the two catalysts
were different from each other. Figure 5 shows the tempera-
ture-programmed oxidation (TPO) results for the used cata-
Figure 6. (a,b) SEM, (c,d) HR-TEM, (e) high-angle annular dark-field scanning
transmission electron microscope (HAADF-STEM), and (f–h) EDS mapping
images of the SZ/MS catalyst.
ure 6e–h), sulfated zirconia was highly dispersed on the MS
support. Only a broad amorphous silica peak corresponding to
SZ/MS is observed in Figure S4, which is also consistent with
the sulfated zirconia being highly dispersed in the SZ/MS cata-
lyst. The loaded amount of sulfur was estimated by elemental
analysis (Table S1). The weight percentage of sulfur in the SZ/
MS is same as that in the MS-FS catalyst. This indicates that
the amount of functional groups is identical in both fresh MS-
FS and SZ/MS catalysts.
Figure 5. TPO-MS profiles of MS-FS and HZSM-5 after 10 h glycerol dehydra-
tion.
lysts. The oxidation of coke in MS-FS starts at 2208C, which is
a lower temperature compared with that for HZSM-5, 2708C.
In addition, the major oxidation peak appears at 3308C in the
MS-FS catalyst whereas that of HZSM-5 appears at 5098C. This
is because the coke in HZSM-5 is likely to be polymerized or
condensed due to the microporous structure of the catalyst.
Also, by comparing the integrated area of TPO profiles in both
catalysts, it can be seen that the amount of coke produced is
similar. As frequently pointed out, the stability of MS-FS is su-
perior to HZSM-5 although a similar amount of coke was de-
posited during the reaction. These results can be explained by
The acid sites of the sulfated zirconia exhibit Brønsted-acidic
characteristics that are structurally similar to the sulfonic acid
[33]
moiety of MS-FS. The major differences in acid sites between
SZ/MS and MS-FS are acid strength and uniformity of acid site
distribution. Temperature-programmed desorption of NH3
(NH -TPD) results for MS-FS and SZ/MS are presented in
3
Figure 7. NH desorption in the case of MS-FS started at 1008C
3
[
21]
a study reported by Hoang et al. In the meso–macropores of
MS-FS, the residence time of intermediates or coke precursors
can be reduced, which results in a decrease in the probability
of coke polymerization.
and ended at around 3008C. In the case of SZ/MS, the desor-
bed NH was detected in the broad temperature range from
3
120–4808C. The higher temperature of NH desorption in the
3
case of SZ/MS suggests that the SZ/MS catalyst exhibit a slight-
ly stronger acidity than the MS-FS catalyst. Moreover, MS-FS
A comparison of the Brønsted acid sites was carried out
using sulfated zirconia supported on the marigold-like silica
showed a narrower NH desorption peak than SZ/MS, which in-
3
(
SZ/MS) and MS-FS catalysts. The morphology of the SZ/MS
dicates that acid sites in MS-FS are uniformly distributed. This
result is consistent with cluster-based DFT calculation results.
Considering the uniform distribution of the acid sites, the se-
lectivity for acrolein of MS-FS was expected to be superior to
that of SZ/MS.
catalyst was confirmed to be identical with that of MS-FS (Fig-
ure 6a–d); the use of the same morphology can diminish the
effect of pore structure. As shown in the 2D nergy dispersive
X-ray spectroscopy (EDS) mapping images of SZ/MS (Fig-
ꢀ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2014, 7, 2193 – 2201 2197

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
The low selectivity for acrolein is explained by the fact that
SZ/MS has a stronger acid strength than MS-FS. Too strong
Brønsted acid sites could result in a significant formation of
by-products because many side reactions are also catalyzed by
[13]
Brønsted acid sites during glycerol dehydration.
However,
the formation of a considerable amount of acetaldehyde (8.6%
in SZ/MS.) cannot be solely explained by the difference in acid
strength. It was reported that the formation of acetaldehyde
[10,15,34]
stems from the homogeneous decomposition of 3-HPA.
Therefore, the formed 3-HPA was preferentially desorbed from
the acid sites of the SZ/MS catalyst and then homogeneous
decomposition to acetaldehyde occurred. To explain such a dif-
ference, we anticipated that the ‘reversibility’ of the Brønsted
acid site was the reason for the difference between the MS-FS
and SZ/MS catalysts. Herein, the term reversibility is similar to
the ability of recovering protons from reaction intermediates.
Scheme 2 shows a proposed reaction mechanism account-
ing for the reversibility of each catalyst. In both catalysts, glyc-
erol is converted into 3-HPA through protonation and dehydra-
tion. After this, in the case of the easily reversible MS-FS cata-
lyst, a proton is readily transferred to an acid site to regenerate
the Brønsted acidity. Over regenerated sites, 3-HPA is readily
adsorbed on an acid site by hydrogen bonding and selectively
dehydrated to acrolein. In contrast, the reversibility of SZ/MS is
somewhat lower than that of MS-FS, and the recovery of
a proton for SZ/MS is more difficult than for MS-FS. When the
deprotonated sites are readily regenerated in SZ/MS, the reac-
tion of 3-HPA would follow the same route as that on the MS-
FS catalyst. On the other hand, when deprotonated sites are
not regenerated, 3-HPA can be desorbed from the active site
due to the absence of interactions and would decompose into
formaldehyde and acetaldehyde. These compounds can partic-
3
Figure 7. NH -TPD–MS profiles of MS-FS and SZ/MS.
Experimental results show that the SZ/MS catalyst has a low
glycerol conversion and selectivity for acrolein, which is a com-
pletely different to the result for the MS-FS catalyst (Table 3).
The deactivation behavior in SZ/MS was also distinct from that
of MS-FS. Even though both catalysts have almost identical
Table 3. Catalytic activity of MS-FS and SZ/MS catalysts in the dehydra-
tion of glycerol with regard to TOS. Amount of catalyst: 0.3 g; glycerol
ꢀ
1
feed rate: 2.0 mLh ; reaction temperature: 2508C.
Catalyst
Glycerol conversion [%]
Selectivity [%]
TOS=1–2 h TOS=9–10 h acrolein acetol acetaldehyde
SZ/MS
MS-FS
76.8
100.0
56.8
100.0
18.3
73.4
10.7
9.3
8.6
0.0
textural properties, glycerol conversion using SZ/MS decreased
with increasing time on stream (TOS). After the reactions, we
analyzed the amount of sulfur in both used catalysts because
sulfonic acid groups or sulfuric acid groups can be detached
from the silica support. The weight percentage of sulfur re-
mained unchanged in both catalysts (Table S1). This indicates
that the main reason for the deactivation of SZ/MS is not the
leaching of sulfur from the catalyst or pore blocking.
[10,12,15]
ipate in side reactions to form coke.
Therefore, we
expect that this difference is the reason why the SZ/MS cata-
lyst exhibits different catalytic activities compared to the MS-
FS catalyst.
To verify the proposed scheme, we compared the recyclabili-
ty of acidic protons for each catalyst using DFT calculations. In
Scheme 2. Proposed mechanism concerning recyclability of a Brønsted acid site in (a) MS-FS and (a) SZ/MS catalysts
ꢀ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2014, 7, 2193 – 2201 2198

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
[
14]
the elementary steps of a typical dehydration, the reaction is
initiated by the migration of a proton from a Brønsted acid
site to the reactant. Then, the protonated reactant undergoes
dehydration and a charged intermediate is formed. Finally, the
proton is pulled out from the charged intermediate with the
Brønsted acidity being recovered. As already noted, we antici-
pate that the last step is probably important in the sequential
dehydration of glycerol.
lower protonation energy corresponds to a catalyst with
a stronger acidity. At the same time, however, the recovery of
a proton from a hydronium ion over the MS-FS catalyst is
more thermodynamically favored than the analogous process
for SZ/MS due to the higher stabilization energy of proton ex-
ꢀ
1
traction (ꢀ111.4 vs. ꢀ79.2 kJmol ). Therefore, the recyclability
of SZ/MS was inferior to that of MS-FS, which resulted in the
generation of deprotonated acid sites. Over the deprotonated
acid sites, the adsorption of 3-HPA was not favorable as the
In the calculation, a water molecule was used as a probe to
investigate the recyclability of the MS-FS and SZ/MS catalysts.
Figure 8 shows the optimized structures for the overall dehy-
ꢀ
1
calculated energy was only ꢀ1.9 kJmol , which was negligible
ꢀ
1
compared to ꢀ56.2 kJmol for the initial acid site (Figure S5).
In this situation, 3-HPA can be readily desorbed from the site
and decomposition to acetaldehyde and formaldehyde would
be expected to follow. Based on these computational calcula-
tions, we confirmed that the proposed reaction mechanism in
MS-FS and SZ/MS catalysts is reasonable. In addition, it cannot
be emphasized enough that not only appropriate acidity but
also the recyclability of acid sites is a crucial factor for the se-
lective and stable production of acrolein in the sequential de-
hydration of glycerol.
Conclusions
We designed and successfully synthesized a hierarchical-pore-
structured and propanesulfonic-acid-functionalized silica cata-
lyst (MS-FS) to stably produce acrolein from glycerol. As ex-
pected, the MS-FS catalyst showed an outstanding selectivity (
ꢁ
73%) and stability (50 h) during the sequential dehydration
Figure 8. (a,c) Adsorption geometries and (b) transition state for water-as-
sisted proton transfer reaction on (A) MS-FS and (B) SZ/MS catalysts.
of glycerol to acrolein. Compared to a microporous HZSM-5
zeolite, the hierarchically structured meso–macropores in MS-
FS were strongly resistant to the blocking of pore entrances by
coke deposition. Moreover, the formed coke was confirmed to
be more easily oxidized than the highly condensed coke
formed in the case of HZSM-5. As a consequence, acrolein was
stably produced using the MS-FS catalyst, even after a consider-
able amount of coke had been deposited.
dration process, which consists of the adsorption of water,
proton transfer from a Brønsted acid site to water, and the re-
covery of a proton from a hydronium ion. To describe proton
transfer and recovery, we simulated a water-assisted proton ex-
change. The adsorption energies of water on the MS-FS and
In addition, we found that the recyclability of deprotonated
Brønsted acid sites is a significant factor in the sequential de-
hydration of glycerol. If the recyclability of a catalyst is suffi-
cient, the regeneration of deprotonated Brønsted acid site
easily takes place. In this case, 3-HPA formed from the first de-
hydration of glycerol is favorably adsorbed onto the regenerat-
ed Brønsted acid site followed by dehydration into acrolein in
the subsequent step. However, in the case of a catalyst with
a low recyclability, the regeneration of the deprotonated acid
sites is retarded and the probability of 3-HPA being desorbed
from the active site increases, resulting in a high likelihood of
decomposition into acetaldehyde and formaldehyde. DFT cal-
culations revealed that the MS-FS catalyst had an improved re-
generation ability of deprotonated Brønsted acid sites com-
pared to that of sulfated zirconia supported on silica (SZ/MS).
Therefore, the sequential dehydration is easily achieved in the
case of the MS-FS catalyst with a high selectivity for acrolein.
Furthermore, catalyst stability is also enhanced because the
formation of coke precursors generated from the degradation
of 3-HPA is inhibited at the regenerated acid sites.
ꢀ
1
SZ/MS surfaces were ꢀ39.8 and ꢀ49.6 kJmol , respectively.
The activation energy required for the transfer of a proton
ꢀ
1
from MS-FS was 111.4 kJmol , which was higher than that for
ꢀ
1
SZ/MS (79.2 kJmol , Figure 9). This result is consistent with ex-
perimental observations (NH -TPD–MS, Figure 7) because the
3
Figure 9. DFT calculations of the minimum energy path for water-assisted
proton exchange over MS-FS (*) and SZ/MS (
) surfaces.
&
ꢀ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2014, 7, 2193 – 2201 2199

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
ꢀ1
to 508C and increased again to 6008C (108C·min ) under a flow
of helium. The data was collected simultaneously.
Experimental Section
Preparation of catalysts
MS-FS was synthesized by a water-in-oil microemulsion process
using a hydrothermal reactor. A synthetic method for marigold-like
Catalytic activity test
[22]
silica (MS) was published by Park et al. Urea (0.6 g) and cetylpyri-
dinium bromide (1.0 g) were dissolved in water (30 mL). Tetraethyl
orthosilicate (2.5 g), pentanol (1.5 mL), and cyclohexane (30 mL)
were mixed. The two solutions were mixed for 30 min, and the
mixture was prehydrolyzed at 1208C for 2.5 h in an autoclave
while stirring. After the reaction, the sample was cooled to room
temperature and 3-mercaptopropyl trimethoxy silane (0.62 g) was
slowly added while stirring. The sample was prepared in the same
way as for prehydrolyzation, with the only difference being a 4 h
reaction time. The product was centrifuged three times using a mix-
ture of deionized water (15 mL) and acetone (15 mL). After that,
the resulting material was dried at room temperature and grinded
to a fine powder. To remove the surfactant, ethanol extraction was
conducted because the functional groups can be oxidized through
heat treatment or calcination. The sample (1 g) was suspended in
ethanol (250 mL). After stirring at 758C for 24 h, a precipitate was
obtained by centrifugation and dried at room temperature. Then,
the sample (1 g) was oxidized using hydrogen peroxide (80 g;
The catalytic activity test was conducted at 2508C in a fixed-bed
quartz reactor. Before the reaction, the catalyst was pretreated at
ꢀ1
2
508C for 30 min under a flow of N (30 mLmin ). For the long-
2
term stability test of MS-FS, the catalyst (0.45 g) was loaded and
a glycerol solution (1.2m) was fed into the reactor using a syringe
ꢀ
1
pump at
(
a
rate of 1.5 mLh
30 mLmin ). For the comparison test, the catalyst (0.3 g) was
loaded and a glycerol solution (1.2m) was fed into the reactor
under a flow of nitrogen
ꢀ
1
ꢀ
1
using a syringe pump at a rate of 2.0 mLh i to exactly observe
the difference in catalytic activities. When the glycerol solution was
injected, the temperature of the inlet line was heated to 2708C to
vaporize the reactant. After the gas-phase glycerol reacted, the
products were condensed in a cold trap and analyzed using a gas
chromatograph (Younlin ACME 6100) equipped with a flame ioniza-
tion detector (FID) and an HP-Innowax capillary column.
Computational details
3
4.5 wt%) at 608C for 24 h followed by washing with deionized
water and ethanol. Finally, the product was dried at 708C for 12 h.
Zirconia supported on silica (7 wt%) was synthesized using an in-
cipient wetness impregnation method. Zirconium oxychloride was
dissolved in deionized water, and the marigold-like silica was im-
pregnated with the solution. The sample was then dried at 808C
overnight and calcined at 3008C for 2 h. The calcined sample
Plane-wave DFT calculations were performed using the Vienna Ab-
[35]
initio Simulation Package (VASP) code implementing the general-
ized gradient approximation (GGA) of Perdew-Burke-Ernzerhof
[
36]
(
PBE) exchange-correlation functional. As an all-electron descrip-
[37]
tion, the projector augmented wave method (PAW) was used.
The energy cut-off for the plane-wave basis set expansion was set
to 400 eV and the Brillouin zone was sampled using a 3ꢂ3ꢂ1 Mon-
khost–Pack k-point mesh. All structures were optimized until forces
(
0.5 g) was treated with sulfuric acid (10 mL; 0.5m, 95%) at room
temperature for 30 min and calcined again at 6508C for 3 h.
ꢀ1
on all atoms were converged to <0.03 eVꢁ . The electronic opti-
ꢀ
4
mization steps were converged to <2ꢂ10 eV.
The model surfaces of propanesulfonic acid functionalized silica
(MS-FS) and SZ/MS were constructed based on the model of Ro-
Catalyst characterization
[38]
zanska et al. The (111) surface of b-cristobalite was used as the
To record the morphology of the samples, we carried out electron
microscopy. HR-TEM and SEM images were obtained using a JEOL
JEM-3010 and SUPRA 55VP microscope, respectively. An analytical
high-angle annular dark-field scanning transmission electron micro-
scope (HAADF-STEM, Tenai F20-FEI, 200 kV) equipped with EDS
[39–41]
model to represent amorphous silica surface.
Physical proper-
ties, such as refractive index and bulk density, of b-cristobalite
[42,43]
fairly resemble those of amorphous silica.
The most preferred
structure was found through careful examination of possible struc-
tures, for example, location of propanesulfonic acid group and zir-
conium ion on the silica surface, and rotating angle between the
surface and propanesulfonic acid group. In our calculations, all
atomic positions were fully relaxed to move, and a sufficient dis-
tance between the two adjacent slabs was provided to avoid peri-
(
Tecnai 136-5-EDAX) was used for elemental mapping of the
sample. Nitrogen adsorption–desorption isotherms were recorded
on a Micrometrics ASAP-2010, and the pore size distribution was
determined from the branches of the isotherm using the Barrett–
Joyner–Halenda method. Elemental analysis (CHNS0932, LECO) was
conducted to determine the amount of sulfur in samples before
and after the reaction. XRD patterns were obtained at angles rang-
ing from 10–808 using a Rigaku D-MAX2500-PC powder X-ray dif-
odic interactions.
The climbing image-nudged elastic band (CI-NEB) method
[44,45]
was
used to determine transition states and energy barriers for the
proton-exchange ability of the catalysts. A 1ꢂ1ꢂ1 Monkhost–Pack
k-point mesh and a cut-off energy of 400 eV were used for these
calculations. Initial reaction trajectories consisted of three images
obtained through linear interpolation. To determine minimum
energy paths, these images were optimized until the forces be-
fractometer with CuK radiation (1.5406 ꢁ). The chemical structure
a
of the catalyst was confirmed using a FTIR spectrophotometer
(
Nicolet 6700, Thermo Scientific). TPO and NH -TPD results were
3
obtained using a Micromeritics Autochem II chemisorption ana-
lyzer with an on-line mass spectrometer (QGA, HIDEN ANALYTI-
CAL). In the case of TPO analysis, the sample was loaded onto the
reactor and heated up to 1008C under a helium flow to vaporize
the physically adsorbed molecules. After waiting for 1 h, the tem-
perature was cooled to 508C. The data was collected while the
ꢀ1
tween the images fell to <0.06 eVꢁ .
Acknowledgements
ꢀ1
temperature was increased to 6008C at a rate of 58Cmin under
This work was financially supported by a grant from the Industri-
al Source Technology Development Programs (2013-10033352) of
the Ministry of Trade, Industry and Energy (MOTIE) of Korea. This
research was also supported by the Supercomputing Center/
a flow of 10% O /He. In the case of NH -TPD analysis, the sample
2
3
was pretreated with 10.2% NH /He gas at 508C. The temperature
was increased to 1008C under a He flow to eliminate physisorbed
3
NH . After removing physisorbed NH , the temperature was cooled
3
3
ꢀ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2014, 7, 2193 – 2201 2200

CHEMSUSCHEM
FULL PAPERS
www.chemsuschem.org
Korea Institute of Science and Technology Information with su-
percomputing resources including technical support (KSC-2013-
C1-025). This work was also supported by the National Research
Foundation of Korea (NRF) grant funded by the Korea govern-
ment (MEST) (No.2013R1A2A2A01067164).
[19] B. Katryniok, S. Paul, M. Capron, V. Belliꢃre-Baca, P. Rey, F. Dumeignil,
ChemSusChem 2012, 5, 1298–1306.
[
20] E. Tsukuda, S. Sato, R. Takahashi, T. Sodesawa, Catal. Commun. 2007, 8,
349–1353.
1
[21] T. Q. Hoang, X. Zhu, T. Danuthai, L. L. Lobban, D. E. Resasco, R. G. Mallin-
son, Energy Fuels 2010, 24, 3804–3809.
[
22] D. S. Park, D. Yun, Y. Choi, T. Y. Kim, S. Oh, J. H. Cho, J. Yi, Chem. Eng. J.
013, 228, 889–895.
[23] M. A. Ecormier, K. Wilson, A. F. Lee, J. Catal. 2003, 215, 57–65.
2
Keywords: brønsted acid · coke resistivity · dehydration ·
glycerol · hierarchical pore structure
[
[
24] M. Brꢅndle, J. Sauer, J. Am. Chem. Soc. 1998, 120, 1556–1570.
25] W. M. Van Rhijn, D. E. D. Vos, B. F. Sels, W. D. Bossaert, P. A. Jacobs, Chem.
Commun. 1998, 317–318.
[
[
1] R. Slade, R. Gross, A. Bauen, Energy Environ. Sci. 2011, 4, 2645–2657.
2] M. I. Hoffert, K. Caldeira, G. Benford, D. R. Criswell, C. Green, H. Herzog,
A. K. Jain, H. S. Kheshgi, K. S. Lackner, J. S. Lewis, H. D. Lightfoot, W.
Manheimer, J. C. Mankins, M. E. Mauel, L. J. Perkins, M. E. Schlesinger, T.
Volk, T. M. L. Wigley, Science 2002, 298, 981–987.
[
[
[
26] E. Cano-Serrano, J. M. Campos-Martin, J. L. G. Fierro, Chem. Commun.
2
003, 246–247.
27] D. Margolese, J. A. Melero, S. C. Christiansen, B. F. Chmelka, G. D. Stucky,
Chem. Mater. 2000, 12, 2448–2459.
28] Y. K. Oh, L. Y. Hong, Y. Asthana, D. P. Kim, J. Ind. Eng. Chem. 2006, 12,
[3] C. H. Zhou, X. Xia, C. X. Lin, D. S. Tong, J. Beltramini, Chem. Soc. Rev.
911–917.
2011, 40, 5588–5617.
[
[
29] F. wang, J. L. Dubois, W. Ueda, J. Catal. 2009, 268, 260–267.
30] B. Katryniok, S. Paul, M. Capron, C. Lancelot, V. Belliꢃre-Baca, P. Rey, F.
Dumeignil, Green Chem. 2010, 12, 1922–1925.
[
4] A. J. Ragauskas, Ch. K. Williams, B. H. Davison, G. Britovsek, J. Cairney,
C. A. Eckert, W. J. Frederick, J. P. Hallett, D. J. Leak, C. L. Liotta, J. R. Mie-
lenz, R. Murphy, R. Templer, T. Tschaplinski, Science 2006, 311, 484–489.
5] G. W. Huber, A. Corma, Angew. Chem. Int. Ed. 2007, 46, 7184–7201;
Angew. Chem. 2007, 119, 7320–7338.
[
[
[
[
31] P. Lauriol-Garbey, S. Loridant, V. Belliꢃre-Baca, P. Rey, J. M. M. Millet,
[
Catal. Commun. 2011, 16, 170–174.
32] D. S. Park, B. K. Kwak, N. D. Kim, J. R. Park, J. H. Cho, S. Oh, J. Yi, Chem-
CatChem 2012, 4, 836–843.
33] F. Cavani, S. Guidetti, C. Trevisanut, E. Ghedini, M. Signoretto, Appl.
Catal. A 2011, 409–410, 267–278.
34] M. R. Nimlos, S. J. Blanksby, X. Qian, M. E. Himmel, D. K. Johnson, J. Phys.
Chem. A 2006, 110, 6145–6156.
[
[
6] A. Corma, S. Iborra, A. Velty, Chem. Rev. 2007, 107, 2411–2502.
7] A. Alhanash, E. F. Kozhevnikova, I. V. Kozhevnikov, Appl. Catal. A 2010,
3
78, 11–18.
8] B. Katryniok, S. Paul, V. Belliꢃre-Baca, P. Rey, F. Dumeignil, Green Chem.
010, 12, 2079–2098.
9] B. Katryniok, S. Paul, M. Capron, F. Dumeignil, ChemSusChem 2009, 2,
19–730.
10] A. Corma, G. W. Huber, L. Sauvanaud, P. O’Connor, J. Catal. 2008, 257,
63–171.
[
[
2
[
[
35] G. Kresse, J. Furthmuller, Phys. Rev. B 1996, 54, 11169–11186.
36] J. P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 1996, 77, 3865–
7
[
3868.
1
[
[
37] P. E. Blçchl, Phys. Rev. B 1994, 50, 17953–17979.
38] X. Rozanska, F. Delbecq, P. Sautet, Phys. Chem. Chem. Phys. 2010, 12,
[
[
11] L. Liu, X. P. Ye, J. J. Bozell, ChemSusChem 2012, 5, 1162–1180.
12] W. Suprun, M. Lutecki, T. Haber, H. Papp, J. Mol. Catal. A 2009, 309, 71–
14930–14940.
78.
[
[
[
[
[
[
[
39] F. Vignꢆ-Maeder, P. Sautet, J. Phys. Chem. B 1997, 101, 8197–8203.
40] J. Handzlik, J. Ogonowski, J. Phys. Chem. C 2012, 116, 5571–5584.
41] J. Handzlik, J. Phys. Chem. C 2007, 111, 9337–9348.
42] D. W. Sindorf, G. E. Maciel, J. Am. Chem. Soc. 1983, 105, 1487–1493.
43] I. S. Chuang, G. E. Maciel, J. Phys. Chem. B 1997, 101, 3052–3064.
44] G. Henkelman, H. Jonsson, J. Chem. Phys. 2000, 113, 9978–9985.
45] G. Henkelman, B. P. Uberuaga, H. Jonsson, J. Chem. Phys. 2000, 113,
[
[
[
[
[
[
13] X. Lin, Y. Lv, Y. Qu, G. Zhang, Y. Xi, D. L. Phillips, C. Liu, Phys. Chem.
Chem. Phys. 2013, 15, 20120–20133.
14] K. Kongpatpanich, T. Nanok, B. Boekfa, M. Probst, J. Limtrakul, Phys.
Chem. Chem. Phys. 2011, 13, 6462–6470.
15] S. H. Chai, H. P. Wang, Y. Liang, B. Q. Xu, Green Chem. 2007, 9, 1130–
1
136.
16] C. J. Jia, Y. Liu, W. Schmidt, A. H. Lu, F. Schꢄth, J. Catal. 2010, 269, 71–
9.
17] Y. Choi, D. S. Park, H. J. Yun, J. Baek, D. Yun, J. Yi, ChemSusChem 2012, 5,
460–2468.
18] Y. T. Kim, S. J. You, K. D. Jung, E. D. Park, Bull. Korean Chem. Soc. 2012,
3, 2369–2377.
9901–9904.
7
2
Received: February 13, 2014
3
Published online on July 8, 2014
ꢀ
2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2014, 7, 2193 – 2201 2201