Mechanistic insights on catalytic conversion fructose to furfural on beta zeolite via selective carbon-carbon bond cleavage

Article abstract of DOI:10.1016/j.mcat.2018.11.022

The mechanism for the formation of furfural by dehydration of D-fructose via selective carbon-carbon (C–C) bond cleavage was investigated over beta zeolite (Hβ). Several different pore size zeolites were employed to determine the shape selectivity of fructose isomers. Zeolitic pore sizes that smaller than the kinetic diameter of cyclic fructose also produced yield furfural (～22.5%), while the high yield (66.0%) of hydroxymethylfurfural (HMF) can be achieved over large pore zeolite (USY zeolite), which could accommodate the cyclic fructose. These results indicated that the furfural formation likely began with acyclic fructose. In situ ¹³C NMR and GC–MS studies, using labeled ¹³C-1-fructose as substrate, suggested that the conversion of fructose to furfural involved with splitting of the C5-C6 bond. Furthermore, the C1 compound from the cleavage of C–C bond was identified as formaldehyde, inferring that the selective scission of C–C bond was ascribe to the retro-aldol reaction. Interestingly, in situ NMR studies implied that the acyclic fructose mainly derived from pyranose forms. In addition, compared with glucose, fructose directly converted to furfural at the higher yield and reaction rate.

Full text of DOI:10.1016/j.mcat.2018.11.022

Molecular Catalysis 463 (2019) 130–139
Contents lists available at ScienceDirect
Molecular Catalysis
journal homepage: www.elsevier.com/locate/mcat
Mechanistic insights on catalytic conversion fructose to furfural on beta
zeolite via selective carbon-carbon bond cleavage
Yueqing Wanga,b, Xiaohai Yanga,b, Hongyan Zheng , Xianqing Li , Yulei Zhua,c,⁎, Yongwang Lia,c
d
c
a
State Key Laboratory of Coal Conversion, Institute of Coal Chemistry, Chinese Academy of Sciences, Taiyuan, 030001, PR China
b
University of Chinese Academy of Sciences, Beijing, 100049, PR China
Synfuels China Co. Ltd., Beijing, PR China
c
d
Department of Chemistry and Chemical Engineering, Taiyuan Institute of Technology, Taiyuan 030008, PR China
A R T I C L E I N F O
A B S T R A C T
Keywords:
Fructose
Furfural
Beta zeolite
Mechanism
In situ NMR
The mechanism for the formation of furfural by dehydration of D-fructose via selective carbon-carbon (CeC)
bond cleavage was investigated over beta zeolite (Hβ). Several different pore size zeolites were employed to
determine the shape selectivity of fructose isomers. Zeolitic pore sizes that smaller than the kinetic diameter of
cyclic fructose also produced yield furfural (∼22.5%), while the high yield (66.0%) of hydroxymethylfurfural
(
HMF) can be achieved over large pore zeolite (USY zeolite), which could accommodate the cyclic fructose.
13
These results indicated that the furfural formation likely began with acyclic fructose. In situ C NMR and
1
3
GCeMS studies, using labeled C-1-fructose as substrate, suggested that the conversion of fructose to furfural
involved with splitting of the C5-C6 bond. Furthermore, the C1 compound from the cleavage of CeC bond was
identified as formaldehyde, inferring that the selective scission of CeC bond was ascribe to the retro-aldol
reaction. Interestingly, in situ NMR studies implied that the acyclic fructose mainly derived from pyranose forms.
In addition, compared with glucose, fructose directly converted to furfural at the higher yield and reaction rate.
1. Introduction
the global demand for furfural is about 300,000–700,000 tons annually.
About 70% of furfural is used for the production of furfuryl alcohol, and
Research on new and renewable resource have been stimulated by
rising demand on fuels and chemicals due to global population growth
1]. As a non-fossil carbon energy resource, lignocellulosic biomass is
this demand is expected to rise [8]. Typically, furfural is the natural
dehydration product of xylose or arabinose. Although favorable yield of
furfural can be achieved in homogeneous catalytic system [9], hetero-
geneous catalysts has gained increasing attention in order to fully
comply with the targets of green chemistry and sustainable production
[10–15]. A large variety of heterogeneous catalysts with acidic char-
acter including zeolites [11], sulfonated resin [16], heteropolyacid [17]
and metal oxide [18], have been studied, and one of the most com-
monly studied solid acids is zeolites. Currently, furfural was mainly
produced from pentose and pentosan. However, it is still a great chal-
lenge to obtain furfural directly from hexose. Despite furfural could be
produced from hexose under pyrolysis and hydrothermal conditions,
the yield of furfural was less than 10–22 % [19–22]. The key devel-
opment in this area was made by Gürbüz and co-worker, by introducing
mordenite and beta zeolite as effective material for the direct conver-
sion of fructose (a ketohexoses containing six carbon atoms) to furfural
[10]. In addition, previous work [23] from our group also revealed that
the high yield (38.5%) of furfural was achieved successfully from cel-
lulose catalyzed by Hβ zeolite. Zeolite has the advantage of shape
[
promising because of its renewability, abundance and wide distribution
in nature, and could have minimal environmental impact if it is prop-
erly processed [2]. Both hydroxymethylfurfural (HMF) and furfural, as
the bridge between lignocellulosic biomass and high add-value fine
chemicals, have received renewed attention for production of biofuels
and biochemicals [3]. Among the products that can be obtained from
cellulose and hemicellulose, furfural is particularly promising, because
it can replace petroleum-based raw materials for the production of re-
sins, lubricants, adhesives and plastics [4]. Furfural is also a natural
precursor for the production of some valuable furan-based chemicals,
such as furfuryl alcohol, tetrahydrofurfuryl alcohol [5], tetra-
hydrofuran, 2-methylfuran and methyltetrahydrofuran [6]. Never-
theless, HMF is mainly derived from six-carbon sugar or cellulose via
dehydration and hydrolysis, and its volume is low due to the difficulties
in cost-effective production [7]. In contrast, the first commercial pro-
duction of furfural was started by Quaker Oats in 1921 [5]. Currently,
⁎
Corresponding author at: State Key Laboratory of Coal Conversion, Institute of Coal Chemistry, Chinese Academy of Sciences, Taiyuan, 030001, PR China.
E-mail address: zhuyulei@sxicc.ac.cn (Y. Zhu).
https://doi.org/10.1016/j.mcat.2018.11.022
Received 12 October 2018; Received in revised form 26 November 2018; Accepted 30 November 2018
2468-8231/ © 2018 Elsevier B.V. All rights reserved.

Y. Wang et al.
Molecular Catalysis 463 (2019) 130–139
selectivity, and makes it possibility to control the pore size and to ob-
tain the optimal structure for intended end use. The unique pore
structure of Hβ favors the special selectivity of furfural formation from
fructose. It is worthy note that the cellulose is the largest fraction (40-
position was located between C1 and C2 depending on 2,3-enediol,
which derived from thermally uncatalyzed process in supercritical
water for fructose conversion [19] (as shown in ESI†, Scheme S1, route
b). Whereas, the breaking of CeC bond occurred between C5 and C6
from 1,2-enediol intermediate may explaining the limited formation of
furfural over mordenite [31] (as shown in ESI†, Scheme S1, route c).
Despite several assumptions have been proposed, it is not yet well
understood if furfural is formed directly from the fructose or HMF, or
via another intermediate over Hβ zeolite so far. In present paper, we
studied the mechanism of the conversion fructose to furfural over Hβ
zeolite on the basis of several aspects: (i) identifying fructose isomer for
furfural formation with zeolitic shape selectivity, (ii) identifying the
position of CeC bond scission using isotopic labeling method and (iii) in
situ monitoring the isomers for formation of furfural. In order to un-
derstand how the CeC bond scission in bio-based fructose occurs over
Hβ zeolite, in situ C NMR was employed to trace the isotopic carbon of
fructose mapped into which carbon of furfural. It was found that the
retro-aldol process was easily catalyzed by Hβ zeolite to produce the
C5 + C1 products. Therefore, accurate understanding of mechanism of
furfural formation via selective cleavage of CeC bond on Hβ zeolite at
molecular level could help the design of high selective catalyst to en-
hance furfural yield.
50%) of lignocellulosic biomass, secondly, the proportion of hemi-
cellulose is about 2530% [5]. However, the single production of fur-
fural product from lignocellulose residues would be wasteful, inefficient
and uneconomic, because pentosans only contribute a portion of total
composition of lignocellulose. Especially, industrial production of fur-
fural from agricultural waste often leaves behind a solid residue con-
taining the amorphous cellulose and lignin [8]. Converting hexose into
furfural over Hβ provides a new pathway for the production of valuable
chemicals from excessive cellulose and its derived monosaccharides,
instead of burning the solid residues in furfural manufacture plant.
However, the precise cleavage of special CeC bond for the pro-
duction of furfural is a challenging task because there are five CeC
bonds with similar bond dissociation enthalpy in fructose molecule.
Fructose was the natural precursor for production of C3 polyols or or-
ganic acid by retro-aldol reaction, and makes it more difficult for the
selective production of furfural by splitting special CeC bond [24,25].
Currently, a variety of catalysts and processes have been studied and
developed to convert hexose (fructose and glucose) to organic acid
13
(
levulinic acid, lactic acid) and polyols (ethylene, propylene glycol)
through catalyzed cleavage of CeC bond, which were nicely summar-
ized and compared in a recent review [26]. A small amount of furfural
can be achieved from hexose under the thermal supercritical water
without catalysts [22]. It is difficult to perform the special CeC bond
scission reaction via chemocatalysis at mild condition. However, due to
the unique three-dimensional intersecting channels with pore openings
size of 6.6 × 6.7 Å and appropriate acid density [27], Hβ zeolite was
proved as the promising material for the production of furfural from
fructose through selective activation of special CeC bond at mild re-
action conditions (140–170 °C). Fundamentally, it is of particular in-
terest to understand how the precise scission of the special CeC bond
can be achieved by Hβ zeolite catalyzed. While the mechanisms of
fructose transformation to HMF via heterogeneous and homogeneous
acid catalysts have recently been elucidated, understanding of the
furfural formation from fructose in Hβ zeolite is still limited. The for-
mation of furfural from fructose over Hβ zeolite is the complex multi-
step process with many possible side reactions, which not only asso-
ciated with the activation of CeO bonds but also the selective cleavage
of special CeC bond. It is known that the breaking of CeC bond is the
major reaction in petroleum refining processes, and the carbocation
mechanism was widely accepted pathway for catalytic cracking and
hydrocracking of heavier hydrocarbons [28]. However, the selective
cleavage of CeC bond in fructose, containing multiply hydroxyl and
carbonyl group, may proceed through different mechanisms. Two
possible schemes have been proposed for the formation of furfural
during the production of HMF from hexose. One possible pathway is
that the fructose was dehydrated to HMF, followed by the elimination
2
. Experimental section
2.1. Materials and instruments
Fructose (analytical grade, 99.5%), glucose (analytical grade, >
9.0%) 5-hydroxymethylfurfural (HMF, > 99.0%), furfural (≥ 99.5%)
9
and γ-butyrolactone (GBL, chromatographic grade, 99.9%) were ob-
tained from Shanghai Aladdin Co., Ltd., ( C-1)-fructose (99 atom %)
was purchased from Cambridge Isotope Laboratories (CIL). D O (99.8
2
atom %) was also obtained from CIL, and zeolites from Nankai
University Catalyst Co., Ltd.. All reagents were used as received without
13
further purification in this work.
Liquid phase 13C nuclear magnetic resonance (NMR) spectra were
obtained by using Bruker AVANCE 400 spectrometer equipped with a
variable temperature probe. The chemical shifts were referenced to an
external standard of D -dioxane. The C spectrum was obtained at
6
13
100 MHz, and the relaxation delay was 10 s.
2.2. Fructose dehydration and analysis
All experiments were carried out in a batch reactor with magnetic
stirrer. Typically, 1.7 wt% initial materials and a certain volume solvent
were added into stainless vessel, followed by the addition of the catalyst
Hβ (0.3 wt %). Before each run the vessel was sealed and flushed with
2
N to exclude air for three times. After the reaction is completed, the
vessel was immediately immerged in ice bath. The reaction products
were centrifuged for 5 min and then filtrated to obtain a clear solution.
The samples were analyzed by GC (Agilent 7890) using an instrument
2
of -CH O group to generate furfural, based on the observation of 5-
hydroxymethylfurfural in pyrolysis process under 350 °C and 500 °C
respectively [29]. As a by-product in production process of HMF, an-
other widely accepted explanation is that the formation of furfural was
mainly ascribed to the retro-aldol condensation reaction, a familiar
process in carbohydrate chemistry. However, the cleavage position of
fructose was different depending upon various mechanisms and inter-
mediates. One possible case is that the fructose was firstly converted
into 3-keto fructose via proton-coupled hydride shift in the presence of
Lewis acid sites, then the retro-aldol condensation reaction was per-
formed between C1 and C2 position, in which the pentose was pro-
duced with eliminating formaldehyde and the pentose can be trans-
formed to furfural by further dehydration (as shown in ESI†, Scheme
S1, route a) [30]. The second case involves enediol intermediate. The
five carbon intermediate for production of furfural derived from the
different cleavage position of CeC bond. For example, the splitting
equipped
with
an
AB-INNOWAX
capillary
column
(
30 m × 0.32 mm × 0.5 μm). Standard solutions were used to obtain
the calibration curves to calculate the concentrations of compounds by
the external standard method. The content of sugar was determined by
HPLC (Agilent 1260) with
a Shodex SH-1821 capillary column
(
300 mm × 8 mm × 0.6 mm). Standard solutions were used to obtain
the calibration curves to calculate the concentrations of the compounds
by the external standard method.
The conversion of substrate and the yield of products were quanti-
fied according to the following equations:
mole of sugar (inlet) − mole of sugar (outlet)
mole of sugar (inlet)
Conversion (mole %) =
×
100%
1
31

Y. Wang et al.
Molecular Catalysis 463 (2019) 130–139
mole of one product produced
mole of theoretical product value
Table 1
Yield (mole %) =
× 100%
a
The formation of furfural from fructose in different temperature .
Yield (%)
Gas chromatography-mass spectrometry (GC–MS) was then used to
identify the labeled products. The sample was analyzed by an Agilent
G2579 A series GC system connected to an Agilent G1530 M quaternary
pole mass detector, using electrospray ionization (EI) 70 eV in the po-
sitive mode. A 30 m × 0.32 mm × 0.5 μm, AB-INNOWAX column was
used, and 1 ml/min helium was used as carrier gas. Products were
identified by comparison of their mass spectra and retention times with
corresponding standard spectra in NIST libraries.
Entry
Temp. (°C)
Conv.(%)
Furfural
HMF
Glucose
others
1
2
3
4
5
6
7
100
120
140
150
160
170
180
170
62.9
91.2
99.9
99.9
99.9
99.9
99.9
5.6
11.0
33.8
47.9
49.8
50.7
51.8
50.7
/
9.2
/
4.8
25.2
29.4
24.3
23.5
19.3
16.8
/
6.4
3.4
1.7
1.4
1.7
0.4
/
4.9
12.3
13.5
17.6
19.9
18.4
3.0
b
8
2.3. Procedure for in situ NMR studies
a
All experiments were carried out in a monophasic system, containing either
O in GBL were used in all cases, each
1
0 mg labeled (13C-1)-fructose and 4 mg Hβ zeolite were added in
1
.7 wt % feed, 0.3 wt% Hβ and 5 wt % H
2
0
.5 ml γ-butyrolactone (GBL) with 5 wt % water. The mixture solution
runs were hold for 50min. GBL = γ-butyrolactone; others: formic acid, levulinic
was sonicated after it was added into the coaxial NMR tube. The NMR
tube was heated to 170 °C with sand bath. The reaction was im-
mediately quenched by immersing the NMR tube in an ice-bath, and the
acid.
b
Start from HMF under the same condition.
1
3
C (rd = 10 s, NS = 256) NMR spectra was recorded using condition
material.
identical to the t = 0 min spectrum. Acetone was added to the coaxial
NMR tube as external standard for in situ quantitative analysis of
fructose isomers.
Since no obvious furfural was formed from HMF, it could be con-
cluded that both furfuran and HMF were directly derived from fructose
via competitive reaction roots rather than sequential reaction (Entry 8).
Glucose as another by-product could be ascribed to the isomerization of
fructose in the presence of Lewis acid sites.
2.4. NMR study of D-fructose tautomer distribution
A low concentration of 0.027 M (mole/L) was adopted in order to
avoid possible contribution of solute-solute interaction, and labeled
3
.2. The fructose isomers analysis for furfural formation
1
3
(
C-1)-D-fructose was fully dissolved in the mixture solvent (GBL/H
2
O)
in the course of C NMR measurements. When GBL contented 5 wt %
O was used as solvent, appropriate D O was added in coaxial NMR
tube by using the deuterium resonance of D O as the lock signal. The
The presence of a variety of isomers makes it difficult to analyze the
1
3
mechanism of fructose conversion. In order to elucidate the mechanism
for furfural formation, it is important to understand how the fructose
isomers reach equilibrium under room temperature. In solution, fruc-
tose adopts five structural conformations: β-pyranose, β-furanose, α-
furanose, α-pyranose and linear keto form as shown in Scheme 1, the
isomers could be transformed between each other via linear keto form.
To help understand the complicated structures of fructose, an abbre-
viated notation are adopted for the fructose isomers.
H
2
2
2
1
3
composition of fructose tautomer was analyzed by C NMR (The pulse
sequence was zgpg 30, the relaxation delay was 10 s, 128 scans.) until
to reach the equilibrium.
2
.5. Catalysts characterization
Furthermore, the furanose (five-membered ring) and pyranose (six-
membered ring) form can be denoted as fur and pyr. By this, for ex-
ample, the β-furanose of fructose is denoted as β-fur. Importantly, the
tautomer distribution have a significant influence on product se-
lectivity, the pyranose forms is prone to produce humins, while fur-
anose forms is more likely to produce HMF by eliminating three water
molecules [32–34]. Now we discuss the pathway of furfural formation
over Hβ zeolite by focusing not only on the products but also on the
FTIR spectroscopy was employed to distinguish the acyclic and
cyclic form of fructose. The FTIR spectra were recorded on a Bruker
VERTX 70 FTIR spectrophotometer, equipped with deuterium triglycine
sulfate (DTGS) detector. The power samples were mixed with KBr and
pressed to translucent disks. The spectra were recorded between
−1
−1
4
000 cm
and 400 cm
.
The X-ray diffraction (XRD) patterns were recorded on a BrukerAxs
D2 diffractometer with Cu Kα radiation (λ = 1.54 Å) at 30 kV and
0 mA with scanning angles (2θ) in the range of 5°-60°.
13
isomers. Despite the isomers could be distinguished by C NMR spec-
troscopy, the low isotopic ratio of 13C as compared to C make the
detection of minor isomers are time consuming.
12
1
3
. Results and discussion
Additionally, due to the ratio of isomers is changing as time goes on,
the 13C NMR spectroscopy can be combined with a site selective la-
beling technique that enables us to dynamically identify competing
isomers. To this end, the high-resolution 13C NMR measurement carried
out as function of time by taking advantage of site selectively labeled
3.1. Furfural formation over Hβ zeolite
Initially, furfural was produced from fructose and 5-hydro-
13
xymethylfurfural (HMF) over Hβ zeolite. The furfural yields for dif-
ferent temperature (Entry 1–7) were shown in Table 1. The yield of
furfural was increased with increasing of reaction temperature. Al-
though favourable yield of furfural was obtained at low temperature,
the ratio of yield between furfural and HMF is lower, compared with
those obtained at elevated temperature (Entry 1–2). In addition, fur-
fural was more stable than HMF at elevated temperature (160–180 °C).
Compared with low temperature, the degradation of fructose was partly
suppressed at high temperature because most fructose was converted to
furfural. Although high yield furfural could be obtained from fructose
over Hβ zeolite, the formation mechanism is not yet clear. In order to
further investigate the mechanism, the production of furfural was stu-
died via experiments under optimal condition using HMF as initial
C-1-fructose at room temperature. As previous shown, the open chain
and cyclic forms of fructose are all observed in γ-butyrolactone (GBL)
containing 5 wt % water: β-pyr (65.11 ppm), β-fur (63.89 ppm), α-fur
(64.11 ppm), α-pyr (64.70 ppm) and keto form (66.85 ppm), where the
parenthesized numbers are the 13C chemical shifts of C1 carbon atom of
interest (as shown in ESI†, Figure S1). According to the integrated peak
intensities, the proportions of isomer fructose are determined at dif-
ferent times up to equilibrium. The relevant results have been shown in
Fig. 1, the following tendencies are observed: (i) the six-membered
fructose of type β-pyr decrease with time, and correspondingly, the
other isomeric forms increase, and (ii) the β-pyr form was the major
form at equilibrium status, following the five-membered fructose of
type β-fur. These transformation tendencies could be accelerated with
132

Y. Wang et al.
Molecular Catalysis 463 (2019) 130–139
Scheme 1. The structure and isomerization of fructose isomers.
Table 2
a
The influence of isomers for furfural formation .
Yield (%)
Entry
1
2
Feedstock
Equilibrated solution
Solid fructose
Conv.(%)
99.9
99.9
Furfural
48.6
49.0
HMF
14.5
15.5
a
All experiments were carried out in GBL containing 5 wt % water at 170 °C.
Once the temperature was reached set value, the feeds were immediately added
in.
forms, but the experiments exhibit the adverse results. This suggested to
us that the furfural may be derived from the same precursor either
pyranose or furanose was predominate. According to present analysis
and previous reports, the mostly possible pathway is that the conver-
sion of fructose into furfural begins with keto form, while the equili-
brium rate between furanose, pyranose and keto forms was sufficiently
faster to replenish the keto form than the dehydration and cleavage
reaction of CeC bond kinetics. However, the evidence is yet insufficient
with respect to the isomers distribution investigation. To this end, the
further investigation should be carried out in view of zeolitic shape
selectivity.
Fig. 1. The isomerization of fructose with time; β-pyr = β-pyranose, β-fur = β-
furanose, α-fur = α-furanose, α-pyr = α-pyranose, keto = linear fructose.
increasing of temperature, even other isomers such as β-fur could pre-
dominate at high temperature [35]. In order to further investigate the
relationship between isomers distribution and furfural formation, some
experiments were designed as follows. According the equilibrium rule
mentioned above, a certain amount of fructose solution containing
3.3. The shape selectivity to tautomer
1
.7 wt % (maximal dissolution in GBL containing 5 wt % H
2
O) was
The shape selectivity is classically defined as being caused by the
effects of mass transport and transition state [36,37]. Therefore, the
unique pore structure of zeolitic materials endows them with special
selectivity for reactants or transition state at molecular level. Specific
pore window size of zeolites limits certain reactant molecules, so shape
selectivity restricts certain reactant and influences the course of reac-
tion. Zeolites with medium pore size (ZSM-22, ZSM-5, MCM-22), large
pore zeolites (beta) and cavity zeolite (USY) are employed to get insight
into the shape selectivity on the fructose conversion pathway and me-
chanism. The kinetic diameter for the products and reactant was esti-
mated from literature value to determine whether the reactions occur
inside the pores or on the external surface [27]. The kinetic diameters
of cyclic fructose, furfural and HMF were estimated as 8.6 Å, 5.5 Å and
6.2 Å, respectively [27,38]. Additionally, the pore window sizes of
zeolites calculated with atomic radii are smaller than those adjusted
with Norman radii. The Norman radii is related to the diffusion of
equilibrated for 20 h at room temperature. Once the reaction tem-
perature reached the set value (170 °C), the fructose solution was im-
mediately added into the reactor, and hold for 50 min. Meanwhile,
another experiment was carried out under the same condition for
comparison using solid fructose as initial material. The distribution rule
of fructose isomers allows us to assume that the pre-equilibrium pro-
cedure make more β-fur form to participate for furfural formation,
while the reaction directly using solid fructose could be regarded as the
major β-pyr form.
However, as shown in Table 2, regardless of the isomers was equi-
librated or not, both the yield of furfural and HMF was no obvious
distinction. These results indicated that the yield of furfural was not
limited to the amount of cyclic fructose isomers. If equilibrium rate
between five isomers was slower than the dehydration, the distribution
of products should be depended on the ratio of pyranose and furanose
133

Y. Wang et al.
Molecular Catalysis 463 (2019) 130–139
Table 3
a
The shape selectivity of zeolitic catalysts .
Yield (%)
Entry
Zeoliteb
Pore size (Å)
Maximal pore sizec
Conv. (%)d
Furfural
HMF
Others
1
2
3
4
5
6
7
ZSM-22
ZSM-5
MCM-22
Beta
USY
A-15
4.5 × 5.2
5.1 × 5.5 5.3 × 5.6
4.0 × 5.5 4.1 × 5.1
6.6 × 6.7 5.6 × 5.6
7.4 × 7.4
/
/
5.9
6.2 and 6.3
6.2 and 5.8
7.4 and 6.3
8.1
/
/
96.6
96.4
99.9
99.9
94.2
99.9
82.3
22.5
21.2
22.5
51.8
10.3
trace
trace
33.4
36.3
33.1
19.3
66.0
37.8
38.6
4.0
5.5
6.5
19.9
1.7
22.8
6.9
γ-Al
2
O
3
+A-15
a
Pore window size of zeolites used in this study from the international zeolite association [39].
The XRD patterns of zeolites used in present work have been shown in ESI†, Figure S2.
Adjusted by Norman radii.
b
c
d
All experiments were carried out in GBL containing 5 wt % water, 1.7 wt % fructose and 0.3 wt % catalysts. Each runs was hold for 50 min at 170 °C. Others:
formic acid, levulinic acid; A-15: Amberlyst 15.
molecules whose diameter is larger than the size of crystallographic
pore [27]. Therefore, the pore sizes adjusted by Norman radii are used
to compare the zeolite pore sizes with kinetic diameter of molecules in
this work. The results have been listed in Table 3, and medium pore
zeolites produced favorable yields of furfural (Entry 1–3).
Jungho Jae and co-workers pointed out that the cyclic glucose could
be excluded out pore of ZSM-5 [27], as such the cyclic fructose is sig-
nificantly larger than the maximum pore size of ZSM-22, ZSM-5 and
MCM-22; it would not be expected to diffuse into the zeolite pore.
Therefore, the favorable yield of furfural from medium pore zeolites
indicated that the formation of furfural is likely derived from linear
fructose, which has the smaller cross section diameter and flexible to
diffuse through the pores of zeolites [38]. The yield of HMF was slight
higher than furfural from medium pore zeolites, indicating that the
more cyclic fructose molecule were converted on the external surface of
zeolite than the inside of pore due to increased diffusion resistance.
These results were further confirmed by using imporous Amberlyst 15
acyclic fructose favors the formation of furfural within the Hβ pores. In
addition, only trace furfural was detected over γ-Al and Amberlyst-
2 3
O
15 catalyst, indicating that the appropriate pore size other than the
properties of Brønsted and Lewis acid play the key role for the forma-
tion of furfural (Entry 7). However, it is not clear which CeC bond is
split over Hβ zeolite, and the selective isotopic labeling technique was
used to address this problem. Further 13C NMR and GCeMS analysis are
carried out using labeled fructose in followed section.
.4. In situ 13C NMR analysis for furfural formation
3
The first carbon (C-1) labeled fructose was examined to determine
which carbon atom (C-1 or C-6) is converted to the carbonyl carbon in
13
furfural. Fig. 3 shows the in situ C NMR spectrums as a function of
reaction time for the conversion fructose to furfural with Hβ in GBL
containing 5 wt % H O at 170 °C. The NMR peaks (64.05, 64.1, 64.9,
2
65.3 and 66.8 ppm) of C-1 carbons were identified as the five main
+
2 3
Al O catalyst (Entry 7), from which favourable yield HMF was
anomeric forms of fructose in the baseline spectrum taken at t = 0 min
in Fig. 3b, or the β-fur, α-fur, α-pyr, β-pyr and acyclic forms as reported
by Thomas Barclay [45].
achieved. Importantly, the highest yield of furfural was obtained over
the large pore size of zeolite (beta) without cavities, while the USY
zeolite containing large pore and 1.3 nm cavities results in high yield of
HMF. However, some previous reports [40–42] suggested that cyclic
fructose could diffuse in the large pore of Hβ. In order to further in-
vestigate the mechanism of furfural formation, the FTIR characteriza-
tion was employed, because carbonyl vibration peak of acyclic fructose
was observed in carbonyl region but this peak was absence in ring
structure fructose. Considering that carbonyl peak of γ-butyrolactone is
also located in the carbonyl region, the dioxane was chosen as replace
Although weak isotopic NMR peaks were visible in the carbonyl
region, which was not the GBL ester group (178.6 ppm), the fructose
13
resonance peaks was dominate on the C NMR spectrum recorded at
t = 5 min (Fig. 3b and c). Compared with other characteristic peaks of
furfural and HMF, which are too weak to be visible in the Fig. 3a, the
overlapping peaks at 178.5 and 178.2 ppm corresponding to the alde-
hyde moiety of furfural and HMF respectively in carbonyl region is
13
greatly enhanced due to C enrichment. Moreover, the intensity of
−1
solvent. As shown in Fig. 2A, the vibration peak at 1768 cm
was
both peaks at 178.2 and 178.5 ppm was increased and the intensity of
peaks assigned to the fructose correspondingly decreased with in-
creasing of time. The peaks ascribed to cyclic intimidate, which is
proposed as a precursor for the formation of HMF [35], was not ob-
served in our experiments. As the reaction time prolongs to 100 min,
the NMR peaks in the carbonyl region become apparent and the peaks
in anomeric carbon region are invisible except peaks at 64.8 ppm
observed, assigned to the carbonyl group of acyclic fructose, in the
-
fructose/Hβ system. Moreover, the very weak vibration peak at 870 cm
1
-1
and 930 cm were also observed, which could be assigned to ring
structure fructose [43,44]. In contrast, as shown in the Fig. 2B, only
-
1
-1
bands at 870 cm and 891 cm were observed with disappearing of
-
1
peak at 1768 cm in the fructose/USY system.
These results indicated that the cyclic fructose could actually diffuse
in the pores of Hβ zeolite, but fraction of acyclic fructose was enriched
in the channel of Hβ. Interestingly, furfural was the major product with
acyclic fructose, while the HMF was dominant without open-ring
fructose on cavities zeolite (USY). That is, the distinction of channels
between USY and Hβ played an important role in the selectivity of
products. Compared with Hβ, the cavities endow USY enough space to
accommodate the cyclic fructose, which feasible to produce HMF.
Whereas the intersecting straightly three-dimensional channels of Hβ
make acyclic tautomer to be enriched in it, and thus the furfural was
dominant product. On the basis of these experimental results from FTIR
characterizations, it is concluded that the high yield of furfural was
achieved mainly due to the shape selectivity of Hβ to tautomer, and the
(
anomeric carbon of linear form). This result manifested that the al-
dehyde carbon of furfural may be originated from the C1 of fructose.
However, it is difficult to identify the chemical shift of furfural and
HMF, because of the chemical shift was strong affected by the chemical
environment such as polarity or solvent composition. In our experi-
ments, the chemical shift of ester group ascribed to GBL was shifted
0
.6 ppm to low-field due to the formation of water in the process of
dehydration. The chemical shift of furfural and HMF was also suffered
from the same effect. Additionally, the resonance peaks could be sig-
nificant enhanced by little isotopic products. Therefore, the further
experiments should be carried out to distinguish the resonance peak of
furfural. Both furfural and HMF at a ratio of 1:1 by weight were added
in the mixture of GBL and 5 wt % water. The NMR spectrum was
134

Y. Wang et al.
Molecular Catalysis 463 (2019) 130–139
135

Y. Wang et al.
Molecular Catalysis 463 (2019) 130–139
136

Y. Wang et al.
Molecular Catalysis 463 (2019) 130–139
Fig. 4. The content of fructose isomers as a function of time, Y value represent the ratio of samples and external standard peak area, pyr = pyranose, fur = furanose.
HMF and partial acyclic fructose. As proposed in the previous reports
[
45,46], the conversion between each isomers of fructose was a re-
versible equilibrium reaction, and the pyranose forms of fructose ra-
pidly was converted to furanose forms via linear keto intermediate. The
reverse process occurs in a similar way. In our experiments, the pyr-
anose forms were consumed faster than furanose, indicating that the
acyclic fructose was mainly derived from pyranose forms. Compared
with blank experiments where furanose forms were dominant without
catalyst, the furfural was a major product catalyzed by Hβ. It indicated
that the equilibrium has been changed over Hβ, and the cyclic fructose
was transformed into acyclic (as shown in Fig. 4a and ESI†, Figure S4).
In situ observations provide direct evidence that the unique pore
structure endows Hβ zeolite with special shape selectivity to facilitate
the formation of keto fructose. Based on the analysis above, both the
formation of furfural and HMF were proved to start from different
isomers, the furfural mainly derived from linear fructose, while the
HMF was originated from furanose forms. The unique pore structure of
Hβ facilitates the conversion of both cyclic pyranose and furanose forms
to acyclic form. This may explain why furfural was the major product.
Moreover, the gas chromatography-mass spectrometry (GCeMS)
experiments were performed to further determine either C-1 or C-6
becomes the carbonyl group in furfural during the catalyzed dehydra-
tion of fructose over Hβ.
As shown in Fig. 5, the furfural molecule with a MW of 97.1 was
observed when using C-1 labeled fructose as initial material, while
unlabeled fructose produced an ion of m/z 96.1 normal molecular
weight. This is consistent with the observation from the NMR studies.
This result further confirms that the C-1 of fructose is converted to the
aldehyde moiety of furfural. Furthermore, the C1 compound from CeC
bond scission was identified as formaldehyde based on the MS spectrum
analysis (as seen in ESI†, Figure S5). Based on isotopic 13C NMR, shape
selectivity and GCeMS analysis, it could be proposed that the formation
of furfural began with acyclic fructose, and the elimination of the C6
moiety in fructose dominated over the elimination of C1 moiety due to
the unique pore structure of Hβ. Then, the C5 intermediate was further
converted to furfural inside pore of Hβ. Additionally, both the route a
and b (as shown in ESI†, Scheme S1) were ruled out by in situ isotopic
NMR and GCeMS studies. Moreover, the cleavage of C5-C6 bond could
be ascribed to the retro-aldol reaction.
Fig. 5. The mass spectrum: (a) 13C-1-fructose and (b) fructose dehydration
catalyzed by Hβ.
recorded, after furfural was added at ratio 2:1 and recorded again under
the same condition. Compared the ratio of peak area before and after
the addition of furfural, the intensity of peak at 178.5 ppm was ob-
viously increased, and the peaks at 178.5 ppm and 178.2 ppm could be
distinguished. The former was assigned to aldehyde group of furfural,
and the peak at 178.2 ppm should be ascribed to carbonyl carbon atom
of HMF (as shown in ESI†, Figure S3). Therefore, it is clarified that the
aldehyde moiety carbon atom was derived from C1 of fructose, and the
conversion of fructose to furfural involves splitting of C5-C6 bond in
linear fructose inside pore of Hβ.
Additionally, the distribution of five isomers was dynamically
monitored with increasing of reaction time, based on the integration of
anomeric carbon NMR peaks. The distribution of isomers was recorded
during initial reaction period when the temperature was reached to set
value. A certain amount of unlabeled acetone was added as external
3
standard for quantitative analysis, and the group −CH (30.3 ppm) was
chosen as standard. As shown in Fig. 4a and b, the contents of acyclic
fructose gradually decline and furfural increased during the initial
period, then content of acyclic fructose gradually increased and other
cyclic forms decreased. Moreover, as Fig. 4b shown, the content of
furfural was increased with increasing of acyclic fructose, and the
equilibrium was achieved at 40 min when the content of furfural and
linear fructose was almost unchanged. Blank experiments were carried
out under the same condition without catalyst (as shown in ESI†, Figure
S4), the equilibrium of isomers was reached within 10 min and main-
tained until 50 min. However, this process was slow with Hβ (Fig. 4a),
the decreasing of furanose forms might be caused by the formation of
3.5. Mechanism of furfural formation
In situ 13C NMR and GC–MS studies using labeled isotope fructose
has ruled out the route a and b. Although the route c could explain the
experiments results, previous reports [47,48] suggested that it was not
feasible to form the initial key intermediate (1,2-enediol) over Hβ,
which contained Lewis and Brønsted acid sites. The 1,2-enediol was
widely observed during the isomerization between glucose and fructose
137

Y. Wang et al.
Molecular Catalysis 463 (2019) 130–139
active sites in this work. Furthermore, the effects of acid sites were also
3
+
investigated using ion-exchanged method. Tervalent ion such as Fe
was selected to tune the acidity of Hβ, and the high silicon (SiO
Al = 40) Hβ was also employed to evaluated the effect of acid sites.
2
/
2 3
O
As shown in Table 4, the density of Lewis acid sites was significantly
decline with introducing of Fe3+ (as shown in ESI, method section).
Consequently, the yield of furfural decline from 50.7% to 27.5% with
decrease of Lewis acid sites, while the small change of Brønsted acid has
slight influence on the yield of furfural (Entry 2).
These results indicated that the Lewis acid sites may function as
active sites for scission of CeC bond, while the dehydration reaction
was performed on Brønsted acid sites. Therefore, it was concluded that
fructose was converted to furfural via the opening of ring of pyranose
fructose, followed by the β-elimination between OH at C3 and H at C4,
as shown in Scheme 2 below. According to the retro-aldol rule or
double bond rule, the double bond between C3-C4 facilitates the
splitting of C5-C6 bond. Additionally, the enol may be the more pre-
ponderant form than its keto because of the electron-withdrawing
properties of carbonyl at C2.
The C5 intermediate was formed by eliminating the C6 as for-
maldehyde, then it was converted to five-membered ring by hemiacetal
reaction between OH at C5 and C = O at C2. Finally, the furfural was
produced by eliminating last two water molecules. Moreover, the for-
mation of furfural via HMF has been ruled out by using HMF as sub-
strate.
Fig. 6. Furfural yields from glucose and fructose vs reaction time (fructose_Y
and glucose_Y are the yield of furfural from fructose and glucose respectively,
fructose_C and glucose_C represent conversion.).
under alkaline condition. Moreover, Davide Carnevali and his co-
worker developed a gas-phase process for conversion fructose to fur-
fural (∼22%) in oxidizing environments. They proposed that the fur-
fural was produced by eliminating formic acid in the first step [49].
However, this possible pathway was also ruled out by our experiments
stem from the fact that the C1 compound was identified as for-
maldehyde by GC–MS. Some studies suggested that glucose could be
converted to furfural via 3-deoxyglucose intermediate [21,50,51], and
glucose (aldose, isomeride of fructose) was detected in our all experi-
ments (Table 1). Therefore, the fructose could be transformed into
furfural through isomerization, dehydration and retro-aldol reactions.
In order to further confirm this proposal, experiments were carried out
using glucose and fructose as substrate, respectively, at 170 °C for dif-
ferent time.
As shown in Fig. 6, the yield of furfural obtained from glucose is
lower than those from fructose, and maximal yield of furfural occurred
at 15 min for fructose when it was converted completely, but the
maximal yield for glucose appeared after more than 30 min and the
conversion was still not complete. These results apparently implied that
the fructose did not undergo 3-deoxyglucose pathway. Therefore,
fructose and glucose were converted to furfural via different pathway
4. Conclusion
The mechanism of fructose conversion to furfural has been in-
vestigated using in situ NMR with isotope 13C labeling. Hβ zeolite ex-
hibits unique shape selectivity for fructose isomers, and the acyclic
isomer was likely an effective precursor for furfural formation. The in
situ 13C NMR studies point out that fructose was converted to furfural
by splitting of C5-C6 bond. In situ monitoring of the acyclic and other
cyclic isomers was using 13C NMR indicated that the acyclic isomer was
mainly derived from β-pyr forms and partially from fur isomers, and
more furfural was produced with increasing of acyclic fructose.
Moreover, the C1 compound, which was removed from terminal (C6) of
fructose, was identified as formaldehyde by GCeMS. These results
implied that the selective scission of CeC bond could be ascribed to the
retro-aldol reaction. Furthermore, fructose was different from glucose
in mechanism of furfural formation, and the fructose undergoes a
pathway with high reaction rate and yield. Based our investigation,
furfural is mainly derived on active sites from inside pores, and acyclic
fructose was mainly transformed from pyr isomer. Therefore, tuning the
external/internal active sites will contribute to the selectivity of fur-
fural. However, the formation of HMF was inevitable because of the
presence of furanose fructose.
(
as shown in ESI†, Table S1). Our previous studies [52] point out the
octahedrally framework aluminum may function as an important active
sites for splitting CeC bond. The formation of furfural was possible
ascribe to the synergistic effect between acid sites (Lewis and Brønsted
acid sites) and octahedrally framework aluminum. The selective scis-
sion of CeC bond most likely occurred via the retro-aldol reaction. The
retro-aldol reaction is common in the glycolysis catalyzed by aldolase A
enzymes, in which the fructose-1,6-bisphosphate was cleaved into di-
hydroxyacetone and glyceraldehyde-3-phosphate. The same reaction
was realized over chemical catalyst with unique pore structure and
Acknowledgements
The authors gratefully thank the financial supports of the National
Table 4
The effect of acid sites .
a
Yield (%)
Acid density (mmol/g)c
b
Entry
Catal.
SiO
2
/Al
O
2 3
Conv.(%)
Furfural
HMF
Others
Brønsted
Lewis
B/L
1
2
3
Hβ
25
40
46
99.9
99.9
99.9
50.7
45.6
27.5
23.5
22.8
20.0
17.6
15.9
7.1
0.28
0.18
0.14
0.17
0.16
0.06
1.65
1.12
2.33
Hβ
d
Fe/Hβ
a
All experiments were performed at 160 °C for 50 min, 1.7 wt% feedstock, 0.3 wt% catalyst, others = formic acid and levulinic acid.
Determined by ICP analysis.
The acid density was determined after desorption at 200 °C.
b
c
d
2 2 3
The Fe/Hβ was derived from Hβ (SiO /Al O = 25) by ion-exchanger method, the value of Fe/Al is 1.
138

Y. Wang et al.
Molecular Catalysis 463 (2019) 130–139
Scheme 2. The possible pathway for formation of furfural.
Natural Science Foundation of China (No. 21875276), Shanxi Science
and Technology Innovation Team of Biofuels & Chemicals, Institute of
Coal Chemistry, Chinese Academy of Sciences, Taiyuan, P R China. The
authors also thank Yingxiong Wang Researcher of Institute of Coal
Chemistry for NMR technical assistance.
[20] P.R. Patwardhan, J.A. Satrio, R.C. Brown, B.H. Shanks, J. Anal. Appl. Pyrolysis 86
(2009) 323–330.
[
[
21] F. Jin, H. Enomoto, Energy Environ. Sci. 4 (2011) 382–397.
22] T.M. Aida, K. Tajima, M. Watanabe, Y. Saito, K. Kuroda, T. Nonaka, H. Hattori,
R.L. Smith, K. Arai, J. Supercrit. Fluids 42 (2007) 110–119.
[23] J. Cui, J. Tan, T. Deng, X. Cui, Y. Zhu, Y. Li, Green Chem. 18 (2016) 1619–1624.
[24] M.S. Holm, S. Saravanamurugan, E. Taarning, Science 328 (2010) 602–605.
[25] N. Ji, T. Zhang, M. Zheng, A. Wang, H. Wang, X. Wang, J.G. Chen, Angew. Chem.
Int. Ed. 47 (2008) 8510–8513.
Appendix A. Supplementary data
[
[
26] W. Deng, Q. Zhang, Y. Wang, J. Energy Chem. 24 (2015) 595–607.
27] J. Jae, G.A. Tompsett, A.J. Foster, K.D. Hammond, S.M. Auerbach, R.F. Lobo,
G.W. Huber, J. Catal. 279 (2011) 257–268.
Supplementary material related to this article can be found, in the
online version, at doi:https://doi.org/10.1016/j.mcat.2018.11.022.
[28] J. Weitkamp, ChemCatChem 4 (2012) 292–306.
[29] K. Katō, Agric. Biol. Chem. 31 (1967) 657–663.
[30] R.S. Assary, L.A. Curtiss, Energy Fuels 26 (2012) 1344–1352.
[31] C. Moreau, P. Faugeras, P. Rivalier, P. Ros, G. Avignon, Appl. Catal. A Gen. 145
References
(1996) 211–224.
[
1] J.A. Melero, J. Iglesias, A. Garcia, Energy Environ. Sci. 5 (2012) 7393–7420.
[32] M. Bicker, D. Kaiser, L. Ott, H. Vogel, J. Supercrit. Fluids 36 (2005) 118–126.
[33] G.R. Akien, L. Qi, I.T. Horvath, Chem. Commun. (Camb.) 48 (2012) 5850–5852.
[34] Y. Qiao, C.M. Pedersen, Y. Wang, X. Hou, ACS Sustain. Chem. Eng. 2 (2014)
2576–2581.
[35] H. Kimura, M. Nakahara, N. Matubayasi, J. Phys. Chem. A 117 (2013) 2102–2113.
[36] B. Smit, Chem. Rev. 108 (2008) 4125–4184.
[2] R. Mariscal, P.M. Torres, M. Ojeda, I. Sádaba, M.L. Granados, Energy Environ. Sci. 9
(2016) 1144–1189.
[
[
[
[
3] J.J. Bozell, G.R. Petersen, Green Chem. 12 (2010) 539–554.
4] L.C. Nhien, N.V.D. Long, S. Kim, M. Lee, Biotechnol. Biofuels 10 (2017) 81.
5] X. Li, P. Jia, T. Wang, ACS Catal. 6 (2016) 7621–7640.
6] J.P. Lange, E. van der Heide, J. van Buijtenen, R. Price, ChemSusChem 5 (2012)
[37] A. Poursaeidesfahani, M.F. de Lange, F. Khodadadian, D. Dubbeldam, M. Rigutto,
N. Nair, T.J.H. Vlugt, J. Catal. 353 (2017) 54–62.
1
50–166.
[
[
7] C. Moreaua, M.N. Belgacemb, A. Gandini, Top. Catal. 27 (2004) 11–30.
8] A.S. Mamman, J.M. Lee, Y.C. Kim, I.T. Hwang, N.J. Park, Y.K. Hwang, J.S. Chang,
J.S. Hwang, Biofuels, Bioprod. Bioref. 2 (2008) 438–454.
[38] K.L. Ratikorn Netrabukkana, Gregory L. Rorrer, Ind. Eng. Chem. Res. 35 (1996)
458–464.
[39] http://www.iza-structure.org/databases/.
[
9] A.S. Dias, M. Pillinger, A.A. Valente, Appl. Catal. A Gen. 285 (2005) 126–131.
[40] M. Leon, T.D. Swift, V. Nikolakis, D.G. Vlachos, Langmuir 29 (2013) 6597–6605.
[41] T.D. Swift, H. Nguyen, Z. Erdman, J.S. Kruger, V. Nikolakis, D.G. Vlachos, J. Catal.
333 (2016) 149–161.
[
10] E.I. Gurbuz, J.M. Gallo, D.M. Alonso, S.G. Wettstein, W.Y. Lim, J.A. Dumesic,
Angew. Chem. Int. Ed. 52 (2013) 1270–1274.
[
[
11] S. Lima, M. Pillinger, A. Valente, Catal. Commun. 9 (2008) 2144–2148.
12] C.G. Sancho, I. Sadaba, R.M. Tost, J.M. Robles, J.S. Gonzalez, M.L. Granados,
P.M. Torres, ChemSusChem 6 (2013) 635–642.
[42] M. Holtkamp, T. Waluga, S. Scholl, J. Non-Equilib. Thermodyn. 37 (2012) 247–272.
[43] V.A. Yaylayan, A.A. Ismail, S. Mandeville, Carbohydr. Res. 248 (1993) 355–360.
[44] V.A. Yaylayan, A.A. Ismail, Carbohydr. Res. 276 (1995) 253–265.
[45] T. Barclay, M.G. Markovic, M.R. Johnston, P. Cooper, N. Petrovsky, Carbohydr. Res.
347 (2012) 136–141.
[
[
[
[
[
[
[
13] I.A. Telleria, J. Requies, M.B. Güemez, P.L. Arias, Appl. Catal. B: Environ. 145
(
2014) 34–42.
14] I.A. Telleria, J. Requies, M.B. Güemez, P.L. Arias, Appl. Catal. B: Environ. 115-116
2012) 169–178.
15] I. Sádaba, M. Ojeda, R. Mariscal, M.L. Granados, Appl. Catal. B: Environ. 150-151
2014) 421–431.
[46] T.D. Swift, C. Bagia, V. Choudhary, G. Peklaris, V. Nikolakis, D.G. Vlachos, ACS
(
Catal. 4 (2013) 259–267.
[47] Y.R. Leshkov, M. Moliner, J.A. Labinger, M.E. Davis, Angew. Chem. Int. Ed. 49
(2010) 8954–8957.
[48] J.M. Carraher, C.N. Fleitman, J.P. Tessonnier, ACS Catal. 5 (2015) 3162–3173.
[49] D. Carnevali, O. Guévremont, M.G. Rigamonti, M. Stucchi, F. Cavani, G.S. Patience,
ACS Sustain. Chem. Eng. 6 (2018) 5580–5587.
(
16] E. Lam, E. Majid, A.C. Leung, J.H. Chong, K.A. Mahmoud, J.H. Luong,
ChemSusChem 4 (2011) 535–541.
17] A.S. Dias, S. Lima, M. Pillinger, A.A. Valente, Carbohydr. Res. 341 (2006)
2
946–2953.
18] A. Dias, S. Lima, D. Carriazo, V. Rives, M. Pillinger, A. Valente, J. Catal. 244 (2006)
30–237.
[50] L.W. Bao, H. Luo, L.Z. Kong, S.G. Li, W. Wei, Sun Y.H, ChemSusChem 10 (2017)
3040–3043.
2
[51] M. Wang, C. Liu, X. Xu, Q. Li, J. Anal. Appl. Pyrolysis 120 (2016) 464–473.
[52] Y. Wang, G. Ding, X. Yang, H. Zheng, Y. Zhu, Y. Li, Appl. Catal. B: Environ 235
(2018) 150–157.
19] T.M. Aida, Y. Sato, M. Watanabe, K. Tajima, T. Nonaka, H. Hattori, K. Arai, J.
Supercrit. Fluids 40 (2007) 381–388.
139