Full Papers

DOI: 10.1002/cssc.201600755
This is the peer reviewed version of the following article: ChemCatChem 2016, 8, 21 – 33, which has been
published in final form at http://dx.doi.org/10.1002/cctc.201501269. This article may be used for non-commercial
purposes in accordance with Wiley Terms and Conditions for Self-Archiving.

Catalyst and Process Design for the Continuous
Manufacture of Rare Sugar Alcohols by Epimerization–
Hydrogenation of Aldoses
Giacomo M. Lari,[a] Olivier G. Grçninger,[a] Qiang Li,[b] Cecilia Mondelli,*[a] Nfflria López,*[b] and
Javier PØrez-Ramírez*[a]
Sugar alcohols are applied in the food, pharmaceutical, polymer, and fuel industries and are commonly obtained by reduction of the corresponding saccharides. In view of the rarity of
some sugar substrates, epimerization of a readily available
monosaccharide has been proposed as a solution, but an efficient catalytic system has not yet been identified. Herein, a molybdenum heteropolyacid-based catalyst is developed to transform glucose, arabinose, and xylose into less-abundant mannose, ribose, and lyxose, respectively. Adsorption of molybdic
acid onto activated carbon followed by ion exchange to the
cesium form limits leaching of the active phase, which greatly
improves the catalyst stability over 24 h on stream. The hydrogenation of mixtures of epimers is studied over ruthenium cat-

alysts, and it is found that the precursor to the desired polyol
is advantageously converted with faster kinetics. This is explained by density functional theory on the basis of its more
favorable adsorption on the metal surface and the lower
energy barrier for the addition of a hydrogen atom to the primary carbon atom. Finally, different designs for a continuous
process for the conversion of glucose into mannitol are studied, and it is uncovered that two reactors in series with one
containing the epimerization catalyst and the other containing
a mixture of the epimerization and hydrogenation catalysts increases the mannitol/sorbitol ratio to 1.5 from 1 for a single
mixed-bed reactor. This opens a prospective route to the efficient valorization of renewables to added-value chemicals.

Introduction
Sugar alcohols, particularly those with C5 and C6 backbones,
are an important class of carbohydrate derivatives. Pentose-derived polyols include arabitol, xylitol, and ribitol, whereas mannitol and sorbitol are the most relevant representatives of
hexose-derived alcohols, which comprise a higher number of
molecules owing to the presence of an additional stereocenter.
All these molecules have multiple uses in the fine chemical,
pharmaceutical, and food industries.[1] Specifically, they improve the nutritional profile of food preparations, as they are
often incompletely absorbed by the intestine,[2] marginally
alter blood sugar levels,[3] and appear to limit the formation of
body fat.[4] They are also added to toothpaste and chewing
gum formulations for their anticaries effect, as oral bacteria
cannot ferment them.[5] Additionally, they are relevant and ver[a] G. M. Lari, O. G. Grçninger, Dr. C. Mondelli, Prof. J. PØrez-Ramírez
Institute for Chemical and Bioengineering
Department of Chemistry and Applied Biosciences
ETH Zurich
Vladimir-Prelog-Weg 1, 8093 Zurich (Switzerland)
E-mail: cecilia.mondelli@chem.ethz.ch
jpr@chem.ethz.ch
[b] Q. Li, Prof. N. López
Institute of Chemical Research of Catalonia, ICIQ
The Barcelona Institute of Science and Technology
¨
Av. Paısos Catalans 16, 43007 Tarragona (Spain)
E-mail: nlopez@iciq.es
Supporting Information and the ORCID identification number(s) for the
author(s) of this article can be found under http://dx.doi.org/10.1002/
cssc.201600755.

ChemSusChem 2016, 9, 3407 – 3418

satile monomers for the manufacture of highly performing biodegradable and biocompatible materials[6] and renewable feedstocks for the preparation of chemicals[7] and fuels.[8] Notably,
arabitol has recently entered the list of the top-12 biobased
chemicals.[9]
Sugar alcohols are rarely isolated from natural products and
are more commonly obtained by hydrogenation of the corresponding carbohydrates. This has mostly been achieved by
means of enzymes,[10] although chemocatalytic technologies
have also been proposed.[11] The latter systems feature a lower
environmental and economic impact owing to the possibility
of processing reactants at higher concentrations and of operating over wider pH and temperature ranges.[12] Aldoses are the
substrates of choice,[13] as the formation of a hydroxyl group
from a primary carbonyl function does not generate a stereocenter and the reaction can proceed with high selectivity. Nevertheless, the very low availability of certain sugars (e.g., mannose and ribose) limits the feasibility of obtaining some of the
polyols at a large scale.[14] Two approaches have been proposed to overcome this problem. The first comprises the use
of a suitable ketose, as in the case of fructose for the preparation of mannitol.[15] Here, the drawback is the formation of approximately equimolar mixtures of mannitol and sorbitol,
which have to be separated by costly crystallization or chromatography techniques.[16] Indeed, attempts to increase the selectivity to mannitol by modification of the catalyst[17] or by the
addition of a chiral organic co-catalyst[18] have only led to moderate success. Alternatively, the abundant epimeric form of

3407

 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
a rare sugar could be chosen as the starting material, and the
Results and Discussion
desired polyol could be obtained by combining epimerization
and hydrogenation steps.[19] Remarkably, this strategy could be
Design of the epimerization catalyst
applied to sugars such as pentoses, for which the ketonic form
is infrequent. Raney nickel and Ru/C have been shown to be
Bulk and supported molybdenum-based catalysts were preeffective hydrogenation catalysts,[11] whereas the best heteropared to evaluate their activity and stability in sugar epimerizageneous epimerization materials reported so far have shown
tion (Scheme 1). The first group included phosphomolybdic
limitations with respect to the concentration of the
substrate owing to the requirement of using methanol as the medium (Sn-b)[20] or to their level of activity (layered niobium molybdate).[21] Besides, only few
catalytic data are available for solids prepared by immobilizing the highly active but water-soluble[22]
phosphomolybdic heteropolyacid (HPA) through adsorption onto activated charcoal[23] or by precipitation in the form of an insoluble salt.[24] Most relevantly, the combination of the two reactions has never
been studied in detail, and the application of industrially more amenable reactors in flow mode instead
of vessels in batch mode has not been explored.
Herein, we target the preparation of rare sugar alcohols by epimerization followed by hydrogenation. Scheme 1. Preparation of bulk and supported heteropolyacid catalysts.
In particular, we focus on the synthesis of mannitol
from glucose, ribitol from arabinose, and arabitol
from xylose. We initially confront various methods for the imacid in its protonic form and as its silver and cesium salts.[24]
mobilization of HPA by evaluating the epimerization activity of
The second set was obtained by adsorption of the acid or the
the silica-supported, carbon-supported, and ion-exchangesalt on silica, activated carbon,[23] or an ion-exchange resin
resin-supported materials obtained and by investigating their
(IER).[25] Compositional data (Table 1) confirmed the expected
compositional and structural changes upon use in continuousmolybdenum contents in the bulk solids. In the case of the
flow catalytic runs. After selecting an efficient and stable hysupported catalysts, high metal loadings were observed that
drogenation catalyst based on ruthenium, we investigate the
corresponded to up to 75 % uptake of the precursor applied in
relative hydrogenation rates of the starting sugars and their
the preparation. The Mo/P ratio was used to estimate the presepimers by combining kinetic tests and density functional
ervation of the heteropolyacid nature (Scheme 1) of the active
theory (DFT) simulations. Finally, we study the overall perphase upon synthesis. The typical value of phosphomolybdic
formance of a continuous process by comparing different aracid (12) was obtained in most cases. A notable exception was
rangements of the catalysts in single and series reactor configthe material prepared by using the IER, which lost part of the
urations.
phosphoric acid groups. This was ascribed to the limited stability of the heteropolyacid at pH > 3, which resulted in depletion
of its Keggin-type structure and the formation of molybdate
and phosphate ions.[26] Finally, the Mo/M ratio (M = Ag, Cs) was
considered to evaluate the efficiency of the proton exchange.

Table 1. Characterization data and epimerization activity of the molybdate-based catalysts.[a]

Catalyst

Mo[b]
[wt %]

Mo/P[c]

Mo/M[c]

SBET[d]
[m2 gÀ1]

Vpore[e]
[mL gÀ1]

TOF[f]
[hÀ1]

SM[g]
[%]

Mo leaching[b]
[%]

MoO3
H-HPA
Ag-HPA
Cs-HPA
H-HPA/IER
H-HPA/SiO2
H-HPA/C
Ag-HPA/C
Cs-HPA/C

73.5
65.6
48.1
49.7
12.2
18.6
25.5
20.1
16.1

–
12.8
10.9
11.5
15.9
11.8
12.4
12.6
13.1

–
–
3.5
3.8
–
–
–
5.2
5.5

15
7
41
61
21
105
581
629
654

0.21
0.06
0.18
0.16
0.07
0.35
0.41
0.45
0.42

1.8
28.9
31.1
29.1
15.4
35.4
51.8
60.7
55.4

95.7
93.6
89.9
89.4
91.5
95.1
93.4
92.8
90.2

100
100
21
13
3
61
16
4
2

[a] Reaction conditions: T = 333 K, [glucose] = 0.28 m, glucose/Mo = 50, t = 30 min. [b] ICP-OES. [c] XRF. [d] BET method. [e] Volume of N2 adsorbed at p/p0 =
0.99. [f] Based on Mo. [g] Selectivity to mannose.

ChemSusChem 2016, 9, 3407 – 3418

www.chemsuschem.org

3408

 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
This was close to the theoretical value of 4 for the bulk AgHPA and Cs-HPA, whereas it was significantly higher for the
carbon-supported materials, indicating that part of the exchanged metal was lost (Table 1). This can be explained by the
ionic nature of the interaction between the support and the
active phase, as hinted by previous reports describing that
positively charged sites on the carbon surface act as cations
displacing the Ag + and Cs + species.[23] The porous properties
of the catalysts were determined by nitrogen sorption. The
data evidence that H-HPA possessed an even lower specific
surface area (SBET) than the reference material MoO3, whereas
the SBET values of Ag-HPA and Cs-HPA were markedly higher.[27]
Carbon-supported catalysts displayed high porosity (Vpore) and
total surface areas (SBET) greater than 500 m2 gÀ1. Nevertheless,
it should be noted that these values are substantially lower
than those of the bare carbon (Vpore = 0.61 mL gÀ1 and
SBET = 1100 m2 gÀ1). This points to significant pore blockage
upon deposition of large quantities of the heteropolyacid. The
same was observed in the case of the silica-supported material.
Incorporation of the active phase reduced the SBET of the bare
support by 20 m2 gÀ1 and its Vpore by 0.15 mL gÀ1. In the case of
the IER-supported solid, no differences were observed relative
to the pristine carrier. H-, Ag-, and Cs-HPA deposited on
carbon were amorphous, as no distinctive reflections were observed by X-ray diffraction (Figure S1 a in the Supporting Information). On the other hand, H-HPA/SiO2 and the bulk materials
showed typical reflections of the Keggin-type phosphomolybdic acid. The very sharp diffraction lines point to crystals
> 100 nm in size Taking into account the substantially different
loadings of the elements, energy-dispersive X-ray spectroscopy
(EDS) mapping of the Cs-HPA/C catalyst (Figure 1 a) displayed
homogeneous distributions of Cs, Mo, and P on the carbon
support, in line with a high dispersion of the active phase. Additionally, the absence of areas enriched in Mo or P pointed to
a preserved heteropolyacid structure.
The catalytic activities of the prepared materials were tested
in the batch epimerization of glucose to mannose. The heteropolyacid-based catalysts exhibited better performances than
the bulk molybdenum oxide (Table 1 and Table S1), as expected on the basis of previous evidence.[22] Interestingly, the exchange of protons by cesium or silver cations did not influence
the reaction rate significantly, in line with the fact that protons
do not partake in the reaction,[28] but the material became less
soluble and only 13–21 % of the molybdenum leached from
the solids into the reaction mixture.
The immobilization of molybdic acid onto solid supports led
to inferior (IER carrier), similar (SiO2 carrier), or doubled (C carrier) activity. The enhancement in the turnover frequency (TOF)
in the latter case (Table 1) is possibly due to a better dispersion
of the active phase. The amount of Mo leached varied over
a very broad range. Specifically, the maximum (61 %) was observed in the case of the silica-supported solid and the minimum (3 %) for H-HPA/IER. This indicates that ionic interactions,
which were expected to be the only means for immobilization
onto the IER, are most beneficial to anchor the active phase in
a stable manner, whereas weak dipolar interactions, such as
those acting in the case of silica, are ineffective. Both mechaChemSusChem 2016, 9, 3407 – 3418

www.chemsuschem.org

Figure 1. EDS mapping of Mo, Cs, P, and C for Cs-HPA/C in (a) fresh and
(b) used forms.

nisms likely play a role in the deposition of HPA on carbon,[23]
which rationalizes the intermediate leaching observed by
using this carrier. Immobilization of the ion-exchanged HPAs
produced catalysts showing activities and selectivities comparable to those shown by H-HPA/C, in line with the testing of
the bulk materials featuring different counterions. Still, these
solids were more stable, as metal depletion was similar to that
of the robust H-HPA/IER. Also in this case, the presence of a different counterion did not significantly alter the activity or selectivity. On the basis of its high activity and robustness, CsHPA/C was chosen to investigate the temperature dependence
of the reaction and the catalyst stability upon repeated use.
Regarding the first aspect, a negligible glucose conversion was
observed below 313 K, whereas it steadily increased between
333 and 373 K (Figure 2 a). The Mo leaching over 10 cycles was
higher than 80 % in the case of H-HPA/C but was limited to
only 35 % in the case of Cs-HPA/C (Figure 2 b). Notably, in both
cases, the Mo/P ratio did not change, which likely indicates
that the Keggin-type structure was maintained. On the basis of
these results, Cs-HPA/C was tested in continuous mode under
industrially relevant, that is, high-conversion, conditions (Figure 2 c). A minor activity loss was observed over the first 12 h
on stream, whereas deactivation became more pronounced in
the subsequent 12 h of the test. The major reason for the activity loss seems to be the substantial (45 %) depletion of the
Mo content, as determined by elemental analysis. It should be
noted that, despite significant deactivation, the stability of this
optimized catalyst was found to be greatly superior to that of

3409

 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Table 2. Performance of Cs-HPA/C in the epimerization of arabinose and
xylose.[a]

Reaction

Xeq[b]
[%]

TOF[c]
[hÀ1]

Sepimer[d]
[%]

arabinose$ribose
xylose$lyxose

41.2
52.7

54.3
48.2

96.6
98.1

[a] Reaction conditions: T = 333 K, [arabinose, xylose] = 0.28 m, (arabinose,
xylose)/Mo = 50, t = 30 min. [b] Equilibrium conversion obtained at t =
16 h. [c] Based on Mo. [d] Selectivity to ribose or lyxose from arabinose
and xylose, respectively.

carbon has been reported to display the highest activity in this
reaction, although its selectivity to mannitol is limited by the
sustained formation of the epimeric alcohol sorbitol. A more
selective process has been attained by alloying this element
with other metals.[17] To choose the most suitable material, we
thus prepared a 5 wt % Ru/C catalyst as well as bimetallic systems comprising the same amount of Ru and 1 wt % Sn, Cr, La,
or Pd by incipient wetness. As shown in Table 3, the actual

Table 3. Characterization data of the ruthenium-based catalysts.
Catalyst

previously reported materials. The used catalyst was analyzed
by transmission electron microscopy (TEM)-EDS (Figure 1 b),
which showed that the fine distribution of the active phase
was preserved. The slightly decreased intensity of the Cs, Mo,
and P signals in the map of the used sample is in line with
leaching (Table 1). Finally, Cs-HPA/C was evaluated in the epimerization of arabinose to ribose and of xylose to lyxose. In
both cases, high selectivities were attained (Table 2), and the
turnover frequencies were similar to that obtained in the conversion of glucose, which confirmed the versatility of this catalyst with respect to the nature of the saccharide substrate.
Design of the hydrogenation catalysts
For the hydrogenation of glucose/mannose, xylose/lyxose, and
arabinose/ribose mixtures obtained by epimerization of the
first more naturally abundant sugar over molybdate-based materials, the catalyst selection relied on literature studies on the
reduction of fructose to mannose. Ruthenium supported on
ChemSusChem 2016, 9, 3407 – 3418

www.chemsuschem.org

M[b]
[wt %]

Vpore[c]
[mL gÀ1]

SBET[d]
[m2 gÀ1]

DCO[e]
[%]

Ru/C
RuSn/C
RuCr/C
RuLa/C
RuPd/C
Figure 2. (a) Glucose conversion (relative to the equilibrium) as a function of
temperature over Cs-HPA/C. (b) Variation in the molybdenum content and
the molar molybdenum-to-phosphorous ratio upon consecutive batch runs
over H-HPA/C and Cs-HPA/C at 333 K. (c) Glucose conversion (relative to the
equilibrium) versus time-on-stream during continuous experiments over CsHPA/C at 363 K.

Ru[a]
[wt %]

4.8
5.1
4.9
4.8
5.0

–
1.0
0.9
1.0
1.0

0.58
0.61
0.53
0.55
0.57

912
956
988
973
934

61
54
58
60
67

[a] ICP-OES. [b] XRF. [c] Volume of N2 adsorbed at p/p0 = 0.99. [d] BET
method. [e] Ru dispersion determined by CO pulse chemisorption.

metal loadings were determined to be close to the nominal
values. The porous properties of the support were not substantially modified upon deposition of the active phase. No reflections attributed to any of the metals were observed in the
XRD patterns of the solids (Figure S1 b), which was indicative
of high dispersion. The latter was determined for Ru to be in
the 54–67 % range on the basis of CO chemisorption. Additionally, TEM analysis of Ru/C (Figure 3 a) visualized very small metallic particles with diameters less than 5 nm.
Testing of these catalysts in the hydrogenation of fructose
surprisingly revealed that the presence of any of the second
metals did not improve the mannitol/sorbitol ratio (Figure 3 b)
and reduced the activity. The discrepancy between our data
and earlier evidence may originate from the different reaction
conditions applied.[17] On the basis of these results, further
studies were conducted over the simple Ru/C catalyst. Firstly,
the dependence of mannitol selectivity on the temperature
and hydrogen pressure was explored. Although very different
operating conditions were investigated (Figure 3 c), only slight
changes in the mannitol selectivity were observed, which was
the highest below 0.7 MPa and at 313–343 K. On the other
hand, the conversion increased monotonously with the tem-

3410

 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Table 4. Reaction rates for the hydrogenation of glucose/mannose,
xylose/lyxose, and arabinose/ribose on Ru/C.[a]

Sugar mixture (i/j)

TOFi
[hÀ1]

TOFj
[hÀ1]

glucose/mannose
xylose/lyxose
arabinose/ribose

13.4
10.9
19.1

25.1
23.6
48.1

[a] Reaction conditions: T = 333 K, [glucose, xylose, arabinose] = [mannose, lyxose, ribose] = 0.14 m, (glucose + mannose, xylose + lyxose, arabinose + ribose)/Ru = 100, t = 15 min, PH2 = 0.1 MPa.

Figure 3. (a) HAADF-STEM micrographs for the fresh (left) and used (right)
Ru/C catalyst. (b) Effect of metal modifiers on the fructose hydrogenation activity and selectivity to mannitol of Ru/C at 333 K. The TOF was calculated
on the basis of the exposed Ru sites. (c) Mannitol/sorbitol ratio upon fructose hydrogenation versus temperature and H2 pressure.

perature or the pressure (Figure S2). Subsequent testing of the
hydrogenation of the mixtures of epimers was conducted in
these ranges. Batch experiments performed at low conversion
levels proved that the hydrogenation of one of the two epi-

mers proceeded more readily for all mixtures (Table 4). In particular, mannose, lyxose, and ribose were reduced to the corresponding sugar alcohols 2–3 times faster than their isomers. To
further investigate this effect, kinetic tests were performed to
derive the apparent activation energies for the conversion of
the two hexoses. On the basis of the Arrhenius plots (Figure 4 a), the hydrogenation of mannose was found to be less
impeded than that of glucose by 6.5 kJ molÀ1.
This significant difference implies that, assuming first-order
reaction kinetics, the maximum theoretical mannitol/sorbitol
formation rate ratio is between 4 and 2.5 in the 293–373 K
range (Figure 4 b), as confirmed by the experimental data. This
hints at the possibility of overcoming, after adequate engineering of the catalyst and the process, the approximately 1:1
sugar ratio obtained upon epimerization. Ultimately, the twostep process would thus enable an overall higher selectivity
than that obtained by direct hydrogenation of the typically
rare ketose. The analogous differences in the reaction rate observed during the hydrogenation of mixtures of aldopentose
epimers likely find the same origin. Indeed, the stereochemical
configuration of the two carbon atoms in proximity to the carbonyl group is the same for glucose, arabinose, and xylose
(2R,3S) and for mannose, ribose, and lyxose (2S,3S), which exhibited slower and faster reaction rates, respectively.
The stability of the hydrogenation catalyst was evaluated in
24 h runs (Figure 5 a–c). The catalyst lost 15–35 % of its original
activity in the conversion of the three mixtures. This evidence
is in line with previous hydrogenation studies for which Ru/C
catalysts were used and for which the deactivation was ascribed to fouling or poisoning by adsorption of trace amounts

Figure 4. (a) Arrhenius plot and the corresponding apparent activation energies for the hydrogenation of glucose and mannose over Ru/C. (b) Mannitol/sorbitol ratio versus temperature. The red line represents the theoretical values obtained from the activation energies.

ChemSusChem 2016, 9, 3407 – 3418

www.chemsuschem.org

3411

 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers

Figure 5. Total conversion and ratio of sugar alcohols for the hydrogenation
of (a) glucose/mannose, (b) ribose/arabinose, and (c) lyxose/xylose mixtures
over Ru/C at 333 K

of gluconic/mannonic acids formed upon exposure of the
starting saccharide solutions to air.[29] An increase in the sugar
alcohol ratio was observed in parallel to the decrease in activity. This is explained by the fact that, at higher conversion
levels, the concentration of the slower reacting sugar is greater

than that of the faster reacting one, which lowers the difference between their actual hydrogenation rates. Aiming at
a more robust process, the catalyst could be periodically
washed with a suitable solvent to restore its functionality or
the liquid feed could be isolated under an inert atmosphere.
Additionally, the sugar alcohol ratio was slightly altered in
favor of mannitol, ribitol, and arabitol with time on stream.
This is explained by the fact that the more reactive sugar is
consumed more rapidly than the more inert epimer. Accordingly, keeping a low conversion level ensures a higher amount
of mannose, ribose, and lyxose available as substrates, which
promotes the formation of the desired alcohol.
DFT was applied to unravel the molecular reason for the different reactivities of the sugars and epimers towards hydrogenation by considering the case of glucose and mannose. First,
the adsorption modes of these two saccharides on a metallic
Ru(0 0 0 1) surface were studied. In the liquid phase, their cyclic
a and b forms dominate over their open structures (denoted
as o-glucose and o-mannose). Specifically, a-glucose, b-glucose,
a-mannose, and b-mannose are more abundant (33, 66, 66,
and 33 %, respectively) than o-glucose and o-mannose, which
only account for less than 1 % each.[30] Nevertheless, both configurations were considered. The adsorption geometries obtained in the screening are reported in Figure S3, and the adsorption energies relative to the most stable configuration are
reported in Table S2. The calculated energy of water solvation
for b-glucose (À87.8 kJ molÀ1) is in reasonable agreement with
previous reports (À77.2 kJ molÀ1).[31]
Thereafter, activation energies for the hydrogenation reactions were computed. Those obtained for the linear forms
(Table S3 and Figure S4) are not able to explain the difference
in rate determined experimentally, probably because of their
negligible concentration in solution. The results for the ring
structures (Figure 6) show that the a forms have lower adsorption energies than the b forms for both glucose (À97.4 vs.
À92.6 kJ molÀ1) and mannose (À124.5 vs. À109.0 kJ molÀ1) with
energy differences of 4.8 and 15.5 kJ molÀ1, respectively. Moreover, they indicate that both a- and b-mannose adsorb on the
catalyst more strongly than a- and b-glucose. The different stereochemical configuration of the C2 in mannose (S) relative to
that in glucose (R) increases the number of sugar interactions

Figure 6. Adsorption structures and binding energies of the a and b forms of glucose and mannose on Ru(0 0 0 1).

ChemSusChem 2016, 9, 3407 – 3418

www.chemsuschem.org

3412

 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
with Ru through O atoms by one, which, as reported earlier, is
the primary contributor to adsorption.[32] After interaction with
the metal, the molecules can react with water, which leads to
o-glucose and o-mannose, or with adsorbed H atoms. For both
glucose and mannose, the transition states relative to ring
opening[33] and the following hydrogenation reaction were initially studied in the absence of water (Table 5). The energy barriers for ring opening (CÀO cleavage, see Scheme 2) are com-

Table 5. Activation barriers (Ea) for ring-opening and hydrogenation reactions for glucose and mannose in the presence or absence of H2O.

Species
a-glucose
b-glucose
a-mannose
b-mannose

CÀO

C1 + H

193.9
150.5
144.7
174.6

297.2
260.5
162.1
250.9

Ea [kJ molÀ1]
O5 + H
CÀO(H2O)
321.3
235.4
247.0
272.1

194.9
95.5
131.2
159.2

C+H
–
116.7
91.7
–

attain a more realistic representation of the catalytic system.
The results (Figure 7) indicate that the presence of the solvent
promotes ring opening by lowering the energy demand by
55.0 and 5.9 kJ molÀ1 for b-glucose and a-mannose, respectively. The beneficial impact of water is twofold: it provides an H
atom for O5 protonation and stabilizes the transition states by
hydrogen bonding (Scheme 2). In spite of a lower ring-opening
barrier, the activation energy for C1 hydrogenation is higher
for o-glucose than for o-mannose (116.7 vs. 91.7 kJ molÀ1).
Hence, the rate-determining steps are C1 hydrogenation for bglucose and ring opening for a-mannose.
To explain the experimental results, a kinetic model based
on Langmuir competitive adsorption and transition state theories was built.[34] The coverage of surface species was estimated
by Equation (1) and (2):
qiÀA ¼

K ads Á ½i À AŠ
iÀA
P
1 þ i K ads Á ½i À AŠ
iÀA

ð1Þ

for which
parably high for b-glucose and a-mannose (150.5 and
144.7 kJ molÀ1, respectively) and are lower than the energy barriers of the other two isomers, a-glucose (193.9 kJ molÀ1) and
b-mannose (174.6 kJ molÀ1). Regarding hydrogenation, two
sites are present that can accept H atoms (C1 and O5). For
both glucose and mannose, O5 hydrogenation is associated
with higher barriers than C1 hydrogenation. The energy
demand of the latter is lower for a- and b-mannose (162.1 and
205.9 kJ molÀ1) than for a- and b-glucose (297.2 and
260.5 kJ molÀ1), which points to the positive effect of the additional interaction between the surface and the C2 or C1 hydrogenation. On the basis of the values obtained, it appears that
the different configuration of C2 determines that the reactive
conformation is a in the case of mannose and b in the case of
glucose. In view of the distance between C2 and O5, the
impact of RuÀC2 bonding is not effective in lowering the O5
hydrogenation barrier. Additional theoretical simulations were
performed considering the presence of a water molecule to


K ads ¼ exp
iÀA

ÀDGads
iÀA
RT


ð2Þ

and Kads is the equilibrium constant for the adsorption of iÀA,
iÀA
DGads is the Gibbs energy of the adsorption, and [iÀA] is the
iÀA
concentration of glucose or mannose in the form i (a or b) in
solution. The expression for the hydrogenation rate is [Eq. (3)]:

r iÀA ¼ kiÀA Á qiÀA Á qH ¼ AiÀA

ÀE A
iÀA
RT


Á qiÀA Á qH

ð3Þ

in which kiÀA is the rate coefficient, AiÀA is the pre-exponential
factor for the hydrogenation reaction, Ea is the energy barrier
iÀA
for the rate-limiting process (i.e., ring opening or hydrogenation), and qH is the surface coverage of H (which is treated as
a constant; i.e., qH = 1). According to the analysis of the transition states, only b-glucose and a-mannose would be hydro-

Scheme 2. Reaction mechanisms for the hydrogenation of a-mannose to mannitol (top) and b-glucose to sorbitol (bottom).

ChemSusChem 2016, 9, 3407 – 3418

www.chemsuschem.org

3413

 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers

Figure 7. Energy profiles for the hydrogenation of b-glucose to sorbitol (red) and of a-mannose to mannitol (blue) and the corresponding structures calculated over the Ru(0 0 0 1) surface.

genated with lower energy barriers. The most energy-demanding steps are different for a-mannose (step 2) and b-glucose
(step 3), as shown in Table 6. The relative rates for glucose and
mannose hydrogenation can be written as follows, for which
CÀO refers to the ring-opening reaction and C+H to C1 hydro+
genation [Eq. (4)]:
kaÀmannose;CÀO Á qaÀmannose
r aÀmannose r aÀmannose;CÀO
%
¼
r bÀglucose
rbÀglucose;CþH
k bÀglucose;CþH Á qoÀglucose Á qH
kaÀmannose;CÀO Á qaÀmannose
%
kbÀglucose;CþH Á K CÀO
bÀglucose Á qbÀglucose
ads
a
ads
Ea
aÀmannose þ DGaÀmannose ÀE bÀglucose ÀDGbÀglucose
¼ exp À
RT

Table 6. Steps for the hydrogenation of a-mannose and b-glucose on
Ru(0 0 0 1).[a]

Step

a-Mannose

b-Glucose

1
2
3
4

a-mannose+*$a-mannose*
+
a-mannose*!o-mannose*[a]
o-mannose*+H*$mannitol* + *
+
mannitol*$mannitol + *

b-glucose+*$b-glucose*
+
b-glucose*$o-glucose*[a]
o-glucose*+H!sorbitol* + *
+
sorbitol*$sorbitol + *

[a] o-Mannose* and o-glucose* correspond to the linear structures of
mannose and glucose, respectively.

!

estimated to be approximately 11. Considering that the DFT
simulations were performed with several approximations, including the use of only one adsorption configuration for the
½a À mannoseŠ
Á CÀO
reactants and the use of a single molecule of water in the ringK bÀglucose Á ½b À glucoseŠ
! opening step and excluding solvation effects when the moleads
a
ads
Ea
1
cule was adsorbed, the qualitative agreement between this esaÀmannose þ DGaÀmannose ÀE bÀglucose ÀDGbÀglucose
¼ CÀO
exp À
RT
K bÀglucose
timate and the value derived experimentally (2.7) is remarkable. In this perspective, even the order of magnitude of the
ð4Þ
relative activity is meaningful. Notice that the error associated
with the difference in the computed and experimentally estiin which qo-glucose is the surface coverage of open glucose and
mated energy barriers corresponds to only 0.04 eV.
CÀO
K bÀglucose is the equilibrium constant of the b-glucose ringopening reaction. Given that mannose can be practically obtained from glucose by an equilibrium-limited epimerization
Process design
reaction with [glucose]/[mannose] % 2 [Eq. (5)]:
The identification of efficient epimerization and hydrogenation

 

catalysts suitable for use under flow conditions opens the door
½aÀmannoseŠ
2
1
¼ ½mannoseŠ = ½glucoseŠ % 1
ð5Þ
to realize a continuous two-step process for the conversion of
3
3
½bÀglucoseŠ
glucose into mannitol. To this end, the configuration of the
catalytic bed(s) and the operation temperature comprise the
Using the calculated adsorption and activation energies, the
main parameters to be tuned. Initially, the performance of
ratio of reaction rates between a-mannose and b-glucose was
a single bed constituted by a physical mixture of the two cataChemSusChem 2016, 9, 3407 – 3418

www.chemsuschem.org

3414

 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
lysts was evaluated. In this case, the temperature was set to
363 K, as the epimerization catalyst Cs-HPA/C cannot produce
a high concentration of mannose under milder conditions (Figure 2 a), as discussed before. At the initial stages of the reaction, the mannitol/sorbitol ratio was approximately 1 (Figure 8 a
and Table S4). This is a remarkable result, as the value is equivalent to that obtained in the reduction of fructose.[11] Still,

Figure 8. Glucose conversion and mannitol/sorbitol ratio by using different
reactor configurations: (a) a single reactor with a mixture of Cs-HPA/C and
Ru/C at 363 K, (b) two series reactors containing Cs-HPA/C at 363 K and Ru/C
at 333 K, respectively, and (c) two series reactors containing Cs-HPA/C at
363 K and a mixture of Cs-HPA/C and Ru/C at 333 K, respectively.

during the course of the reaction, the glucose conversion decreased by approximately 30 % and the mannitol/sorbitol ratio
was reduced to approximately 0.4. The two observations are
likely related, that is, the lower amount of mannitol formed as
a result of the deactivation of the epimerization catalyst
caused a reduction in the mannitol production rate, similar to
the case of the hydrogenation of equimolar mixtures of epimers (Figure 5). Moreover, the hydrogenation catalyst was operated at an excessively high temperature, which could trigger
additional deactivation phenomena (Figure 4 b). Hence, an improved process was achieved by separating the epimerization
ChemSusChem 2016, 9, 3407 – 3418

www.chemsuschem.org

and the hydrogenation catalysts into two distinct reactors. In
this case, the temperature applied to the latter could be lowered by 50 K, which theoretically should enable an increase in
the mannitol/sorbitol ratio by approximately 40 %. In this configuration, the polyols ratio was raised to 1.2 (Figure 8 b and
Table S4) and, owing to the milder conditions in the second reactor, remained significantly more stable than in the first scenario. The glucose conversion and the deactivation rate were
similar to those observed in the previous case. Still, further optimization of the process was envisaged on the basis of the following reasoning. Even if the hydrogenation catalyst is operated at its optimal temperature, concentration gradients will develop along the bed. Specifically, the portion of catalyst located closer to the reactor outlet will be contacted with a solution
containing much less mannose, as a great part of it will have
already been converted over the fraction of catalyst placed at
the beginning of the bed. Thus, the formation of mannitol will
decrease along the bed. Consequently, a third configuration
was explored, in which the first reactor was dedicated solely to
epimerization at high temperature (363 K) and the second reactor contained a mixture of Ru/C and Cs-HPA/C, operated at
a lower temperature (333 K). In this way, the mannose concentration, which decreases during the course of the hydrogenation reaction, could be restored continuously. Remarkably, the
mannitol/sorbitol ratio increased even further to approximately
1.6 (Figure 8 c and Table S4). Additionally, the glucose conversion was constant over 1 day, possibly thanks to the overall increased mass of the Cs-HPA/C catalyst. In line with that observed in the second process design, the mannitol selectivity
remained more stable over the same timeframe as well.
This continuous chemocatalytic technology stands as
a promising alternative to the batch biocatalytic preparation of
sugar alcohols. In fact, it allows for higher operational flexibility
and for reduced waste production, as sugar feeds featuring
a less constrained composition and a higher concentration can
be applied. Moreover, it is foreseen that the continuous-flow
reactors can be scaled up easily and that energy can be saved
through heat integration between units, that is, heat recovered
from the reactors can produce energy to operate the separation units. These elements are expected to substantially lower
the economic and environmental footprint of the production
of polyols.

Conclusions
We successfully demonstrated a chemocatalytic continuous
process for the preparation of sugar alcohols from the epimers
of their corresponding aldoses. Targeting the preparation of
mannitol from glucose, ribitol from arabinose, and arabinol
from xylose, an improved epimerization catalyst was initially
developed. In this regard, activated carbon was uncovered as
a support offering higher stability to the immobilized phosphomolybdic acid phase. Ion exchange of the heteropolyacid
from the protonic form to the cesium form limited its large solubility even further, which led to partial retention of the original catalyst performance even after 24 h on stream. Thereafter,
the hydrogenation of mixtures of epimers over a ruthenium

3415

 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
catalyst was studied. Interestingly, the reaction rate was found
to be dependent on the stereochemical configuration of the
sugar. Density functional theory simulations showed that the
latter strongly influences the adsorption and activation energies for the hydrogenation of the carbon atom. Finally, the
process design was optimized through the selection of the
most efficient reactor configuration. The use of two series flow
reactors brought considerable advantages with respect to
a single reactor. In particular, the mannitol yield was maximized if the first reactor contained the epimerization catalyst
and the second reactor contained a mixture of the epimerization and hydrogenation solids. In addition to improved selectivity, the possibility of optimizing the operation temperature
in each reactor led to a remarkable increase in catalyst stability.
Overall, this process enables a rather straightforward method
for the preparation of rare polyols starting from widely available sugars. The use of a fully chemocatalytic technology, in
contrast to fermentation, shall grant reduced waste production, continuous operation, easier scalability, and, possibly, improved ecological and economic metrics.

using an EDAX Orbis Micro-XRF analyzer equipped with a Rh
source operated at a voltage of 35 kV and a current of 500 mA.
Powder X-ray diffraction (XRD) was performed by using a PANalytical X’Pert PRO-MPD diffractometer with Ni-filtered CuKa radiation
(l = 0.1541 nm), acquiring data in the 2q = 5–708 range with a step
size of 0.058 and a counting time of 8 s per step. N2 sorption at
77 K was conducted by using a Micromeritics TriFlex analyzer. Prior
to the measurements, the samples were degassed at 573 K under
vacuum for 3 h. High-angle annular dark field scanning transmission electron microscopy (HAADF-STEM) images and energy dispersive spectroscopy (EDS) elemental maps were acquired by using
a FEI Talos instrument operated at 200 kV. Powdered samples were
deposited on Cu grids. HAADF images were collected before and
after the EDS measurements to corroborate the absence of morphological changes. CO chemisorption was performed by using
a Micromeritics AutoChem II 2920 chemisorption analyzer. The
samples were heated at 393 K under a He flow (50 mL minÀ1) for
60 min and were then reduced at 523 K under a flow of 5 vol % H2/
He (20 mL minÀ1) for 30 min. Afterwards, 1 mL of 1 vol % CO/He
was pulsed over the catalyst bed at 308 K every 4 min. The Ru dispersion was calculated on the basis of the amount of CO chemisorbed, considering an adsorption stoichiometry of 1.

Experimental Section

Catalytic testing

Catalyst preparation
Hydrated phosphomolybdic acid (Sigma–Aldrich, > 99 %, denoted
H-HPA) and MoO3 (Sigma–Aldrich, 99.5 %) were used as received.
Silver and cesium phosphomolybdates were prepared by coprecipitation of H-HPA (0.01 m aqueous solution) with AgNO3 (ABCR,
99.9 %) and CsNO3 (Acros Organics, 99.99 %), respectively, which
were added in 50 % molar excess. The solids were separated by filtration, washed with deionized water ( % 100 mL per gram of dried
material), and dried (338 K, 16 h). They were labeled Ag-HPA and
Cs-HPA. Supported HPA catalysts were prepared by using silica (Sipernat 120, code SiO2), an anionic ion-exchange resin
(Amberlyst A26, Sigma–Aldrich, code IER), and granulated activated
carbon (Norit Cabot RX 1.5 Extra, 0.2–0.4 mm sieve fraction,
code C). The supports (2.0 g) were added to a solution of H-HPA
(0.5 g) in water (50 mL), and the suspension was magnetically
stirred for 2 h. Thereafter, the solids were separated by filtration,
washed with deionized water ( % 100 mL per gram of dried support), and dried (338 K, 16 h). They were labeled H-HPA/SiO2, HHPA/IER, and H-HPA/C. Two additional C-supported samples (AgHPA/C and Cs-HPA/C) were produced through the same procedure
by adding AgNO3 or CsNO3 to the suspension. A supported 5 wt %
Ru/C catalyst was synthesized by impregnation of the C support
with an aqueous solution of RuCl3·H2O (ABCR, 99.9 %), followed by
drying (338 K, 16 h) and reduction in hydrogen flow (20 vol % H2/
N2, 100 mL minÀ1, 723 K). Bimetallic 5 wt % Ru/1 wt % M/C catalysts
(M = Sn, Cr, La, or Pd) were prepared through the same method by
additionally using SnSO4, Cr(NO3)3·9 H2O (Sigma–Aldrich, 99 %),
LaCl3 (ABCR, 99.9 %), or PdCl2 (ABCR, 99.9 %) as the precursor of the
second metal upon impregnation.

Catalyst characterization
The Mo and Ru contents in the catalysts were determined by inductively coupled plasma optical emission spectroscopy (ICP-OES)
by using a Horiba Ultra 2 instrument equipped with a photomultiplier tube detector. The Sn, Cr, La, and Pd contents and the Mo/P
ratio were determined by X-ray fluorescence (XRF) spectroscopy by
ChemSusChem 2016, 9, 3407 – 3418

www.chemsuschem.org

Batch epimerization experiments were performed under autogenous pressure in 15 mL, thick-walled glass vials (Ace, pressure
tubes, front seal) dipped in an oil bath heated at 303–353 K. The
vials were loaded with approximately 10 mL of a 0.28 m aqueous
solution of glucose, arabinose, or xylose. Then, the appropriate
amount of molybdate-based catalyst was added to achieve a substrate/Mo molar ratio of 50. The mixture was allowed to react
under vigorous stirring for 30 min. Thereafter, the reaction was
quenched by using an ice bath, and the catalyst removed by
means of a Chromafil Xtra 0.25 mm syringe filter. Batch hydrogenation experiments were performed under 1.0 MPa hydrogen
(Messer, 99.999 %) in 10 mL autoclaves (Endeavor) at 303–393 K.
The vials were loaded with 5 mL of a 0.28 m fructose (Sigma–
Aldrich, > 99 %), a 0.14 m glucose (Sigma–Aldrich, 99.5 %)–0.14 m
mannose (Sigma–Aldrich, > 99 %), a 0.14 m arabinose (ABCR, 99 %)–
0.14 m ribose (ABCR, 98 %), or a 0.14 m lyxose (ABCR, 99 %)–0.14 m
xylose (Sigma–Aldrich, > 99 %) solution in deionized water. Then,
the appropriate amount of Ru/C or modified Ru/C catalyst was
added to achieve a substrate/Ru molar ratio of 100 and hydrogen
was introduced. The mixture was heated to the desired reaction
temperature and allowed to react under 1000 rpm stirring for 1 h.
Then, the reaction was quenched, and the catalyst was removed
by using a Chromafil Xtra 0.25 mm syringe filter.
Continuous catalytic tests were performed by using a homemade
flow reactor setup comprising: one, an HPLC pump (Gilson-306);
two, a mass flow controller; three, stainless-steel tubular reactors
(Swagelok SS-T4-S-035, o.d. = 6 mm, i.d. = 4.6 mm) placed in tubular ovens; four, a backpressure regulator (Swagelok, LH2981001). In
the case of epimerization reactions, the reactor was loaded with
the catalyst (0.2–0.3 g, sieve fraction = 0.2–0.4 mm) and diluted
with quartz (0.5 g, sieve fraction = 0.2–0.4 mm). After heating at
the desired temperature, a liquid feed (0.5 mL minÀ1, 0.28 m glucose, arabinose, or xylose aqueous solution) was admitted, and the
system was pressurized at 1.0 MPa. For hydrogenations of glucose/
mannose, arabinose/ribose, and xylose/lyxose mixtures, the reactor
was loaded with the catalyst (0.2–0.3 g, sieve fraction = 0.2–
0.4 mm, diluted with 0.5 g quartz of the same sieve fraction). The
combination of the epimerization and hydrogenation steps was

3416

 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
studied by using a mixed catalyst bed or two separate catalytic
beds operated at different temperatures. In the former case, the reactor was loaded with a mixture of the two catalysts (0.2 g Ru/C,
0.2 g Cs-HPA/C, diluted with 0.4 g quartz). In the latter case, the
two reactors were loaded with Cs-HPA/C (0.4 g, diluted with 0.4 g
quartz) and Ru/C (0.2 g, diluted with 0.6 g quartz), respectively. Alternatively, the second reactor was filled by using a mixture of the
two catalysts (0.2 g Ru/C, 0.2 g Cs-HPA/C, diluted with 0.4 g
quartz). The hydrogenation and combined epimerization–
hydrogenation reactions were started by following the same protocol as that for the epimerization reaction but adding a gas feed of
50 mLH2 minÀ1. In all tests, liquid samples were periodically collected at the outlet of the reactor.

states, and then their energies were calculated with a denser mesh
than 30 ŠÀ1. The metal atoms were fixed during these processes.
An extensive search of potential adsorption conformations was
performed by following the adsorption rules previously developed
in our group. Solvation contributions for the molecules in the
liquid phase were investigated through the multigrid-based
(MGCM) methodology.[42] The climbing image nudged elastic band
(CI-NEB)[43] and improved dimer method (IDM)[44] were employed to
locate the transition states in the potential surfaces. All transition
states were confirmed with only one imaginary value in the frequency analysis. The most relevant structures were uploaded to
the ioChem-BD database.[45]

Reaction products were isolated by high-performance liquid chromatography (HPLC) in an Agilent 1260 system equipped with
a Biorad Aminex HPX-87C column heated at 338 K and a refractive
index detector set at 303 K by using water (0.450 mL minÀ1) as the
eluent. Quantification was attained on the basis of the absolute
peak areas. Calibration curves were measured in the 0.05–0.3 m
range for the sugars and the corresponding sugar alcohols sorbitol
(Sigma–Aldrich, > 98 %), mannitol (Fluka, 99 %), arabitol (Acros Organics, 99 %), xylitol (ABCR, 99 %), and ribitol (Sigma–Aldrich,
> 99 %). The conversion of substrate i (Xi) and selectivity to the
product k (Sk) were calculated as follows [Eqs. (6) and (7)]:

Acknowledgements



ni;1
X i ¼ 1À
ni;0
Sk ¼


ð6Þ

nk;1
ni;0 À ni;1

ð7Þ

in which n refers to the moles of i or k and 0/1 to the reaction beginning/end. The conversion of the mixtures of sugars was obtained by summing the conversions of the two aldose epimers.
The epimerization equilibrium conversions for glucose, ribose, and
arabinose were determined as the conversion observed after 16 h
of reaction by using Cs-HPA as the catalyst and were 35.0, 41.2,
and 52.7 %, respectively, in good agreement with the literature
data.[35]

Computational details
DFT was employed to study the ruthenium-catalyzed reduction of
glucose and mannose by using the Vienna Ab Initio Simulation
Package (VASP).[36] The exchange and correlation energies were obtained by using the PBE functional.[37] As glucose, mannose, and
other intermediates as well as their hydrogenation products sorbitol and mannitol are large molecules, van der Waals correction was
performed by using the Grimme’s DFT-D2 method[38] with the C6
parameters developed in our group.[39] The inner electrons were
represented by projector augmented wave (PAW) pseudopotentials[40] with cutoff energies of 450 eV. The calculated lattice parameter for Ru was 2.712 Š (c/a = 1.581), which agrees well with the
known experimental value of 2.706 Š (c/a = 1.584).[41] Gas molecules were calculated in a box of 20 ” 20 ” 20 Š3. The Ru/C catalyst
was modeled by a four-layer ruthenium slab with a p(4 ” 4) supercell, for which the two upmost layers were fully relaxed and the remaining atoms at the bottom were fixed to mimic the bulk. A 20 Š
thick vacuum region was set between each slab to avoid their interaction. Given that adsorption was performed on one side of the
slab, dipole correction was applied to eliminate the spurious contributions arising from the system’s asymmetry. Gamma point sampling was used to obtain the adsorption geometries and transition
ChemSusChem 2016, 9, 3407 – 3418

www.chemsuschem.org

This work was sponsored by the Swiss National Science Foundation (Project Number 200020–159760). Dr. S. Mitchell and E. Vorobjeva are thanked for the microscopic analyses. N.L. and Q.L.
are grateful to MINECO (CTQ2015–68770-R) for financial support
and to the Barcelona Supercomputing Centre (BSC-RES) for providing the computational resources.
Keywords: epimerization · flow chemistry · heterogeneous
catalysis · hydrogenation · sugar alcohols
[1] a) Y.-C. Park, E. J. Oh, J.-H. Jo, Y.-S. Jin, J.-H. Seo, Curr. Opin. Biotechnol.
2016, 37, 105 – 113; b) P. Mäki-Arvela, B. Holmbom, T. Salmi, D. Y. Murzin,
Catal. Rev. 2007, 49, 197 – 340.
[2] A. M. Langkilde, H. Andersson, T. F. Schweizer, P. Würsch, Eur. J. Clin.
Nutr. 1994, 48, 768 – 775.
[3] T. M. S. Wolever, A. Piekarz, M. Holland, K. Younker, Can. J. Diabetes
2002, 26, 356 – 362.
[4] T. Matsuo, Y. Baba, M. Hashiguchi, K. Takeshita, K. Izumori, H. Suzuki, J.
Clin. Biochem. Nutr. 2001, 30, 55 – 65.
[5] M. Fontana, C. Gonzµlez-Cabezas, Adv. Dent. Res. 2012, 24, 123 – 128.
[6] F. Zamora, K. Hakkou, A. Alla, J. L. Espartero, S. MuÇoz-Guerra, J. A.
Galbis, J. Polym. Sci. Part A 2005, 43, 6394 – 6410.
[7] E. L. Kunkes, D. A. Simonetti, R. M. West, J. C. Serrano-Ruiz, C. A. Gärtner,
J. A. Dumesic, Science 2008, 322, 417 – 421.
[8] a) J. C. Serrano-Ruiz, J. A. Dumesic, Energy Environ. Sci. 2011, 4, 83 – 89;
b) M. Rose, R. Palkovits, ChemSusChem 2012, 5, 167 – 176.
[9] T. Werpy, G. R. Petersen, Top Value Added Chemicals from Biomass:
Vol. I – Results of Screening for Potential Candidates from Sugars and Synthesis Gas, DOE/GO-102004-1992,Department of Energy, Washington,
DC, 2004.
[10] O. Akinterinwa, R. Khankal, P. C. Cirino, Curr. Opin. Biotechnol. 2008, 19,
461 – 467.
[11] a) C. Luo, S. Wang, H. Liu, Angew. Chem. Int. Ed. 2007, 46, 7636 – 7639;
Angew. Chem. 2007, 119, 7780 – 7783; b) J. V. Karabinos, A. T. Ballun, J.
Am. Chem. Soc. 1953, 75, 4501 – 4502.
[12] M. Morales, P. Y. Dapsens, I. Giovinazzo, J. Witte, C. Mondelli, S. Papadokonstantakis, K. Hungerbühler, J. PØrez-Ramírez, Energy Environ. Sci.
2015, 8, 558 – 567.
[13] A. K. Sirkar, US Pat. 4380678, 1983.
[14] T. Ona, T. Sonoda, K. Itoh, M. Shibata, Holzforschung 1997, 51, 396 – 404.
[15] J. Wisnlak, R. Simon, Ind. Eng. Chem. Prod. Res. Dev. 1979, 18, 50 – 57.
[16] a) R. A. Hales, US Pat. 3484492, 1969; b) H. J. Bart, R. C. Messenbçck,
C. H. Byers, A. Prior, J. Wolfgang, Chem. Eng. Process. 1996, 35, 459 – 471.
[17] A. W. Heinen, J. A. Peters, H. van Bekkum, Carbohydr. Res. 2000, 328,
449 – 457.
[18] M. A. Veksler, Y. I. Petrov, N. G. Geling, E. I. Klabunovskii, Bull. Acad. Sci.
USSR Div. Chem. Sci. (Engl. Transl.) 1974, 23, 45 – 49.
[19] W. M. Kruse, US Pat. 4029878, 1977.
[20] R. Bermejo-Deval, R. Gounder, M. E. Davis, ACS Catal. 2012, 2, 2705 –
2713.

3417

 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
[21] A. Takagaki, S. Furusato, R. Kikuchi, S. T. Oyama, ChemSusChem 2015, 8,
3769 – 3772.
[22] F. Ju, D. Van der Velde, E. Nikolla, ACS Catal. 2014, 4, 1358 – 1364.
[23] M. A. Schwegler, P. Vinke, M. van der Eijk, H. van Bekkum, Appl. Catal. A
1992, 80, 41 – 57.
[24] R. Sheikh, M.-S. Choi, J.-S. Im, Y.-H. Park, J. Ind. Eng. Chem. 2013, 19,
1413 – 1419.
[25] T. Baba, Y. Ono, Appl. Catal. 1986, 22, 321 – 324.
[26] a) J. A. R. van Veen, O. Sudmeijer, C. A. Emeis, H. de Wit, J. Chem. Soc.
Dalton Trans. 1986, 1825 – 1831; b) Z. Zhu, R. Tain, C. Rhodes, Can. J.
Chem. 2003, 81, 1044 – 1050.
[27] M. Langpape, J. M. M. Millet, U. S. Ozkan, M. Boudeulle, J. Catal. 1999,
181, 80 – 90.
ˇ
ˇ
[28] V. Bílik, L. Petrus, V. Farkas, Chem. Zvesti 1975, 29, 690 – 693.
[29] a) Z. Yan, L. Lin, S. Liu, Energy Fuels 2009, 23, 3853 – 3858; b) M. Besson,
P. Gallezot, Catal. Today 2003, 81, 547 – 559.
[30] Y. Zhu, J. Zajicek, A. S. Serianni, J. Org. Chem. 2001, 66, 6244 – 6251.
[31] C. O. da Silva, B. Mennucci, T. Vreven, J. Org. Chem. 2004, 69, 8161 –
8164.
[32] R. García-Muelas, N. López, J. Phys. Chem. C 2014, 118, 17531 – 17537.
[33] W. Plazinski, A. Plazinska, M. Drach, Phys. Chem. Chem. Phys. 2015, 17,
21622 – 21629.
[34] I. Chorkendorff, J. W. Niemantsverdriet, Concepts of Modern Catalysis
and Kinetics, Wiley-VCH, Weinheim, 2006.
[35] S. J. Angyal, Angew. Chem. Int. Ed. Engl. 1969, 8, 157 – 166; Angew.
Chem. 1969, 81, 172 – 182.

ChemSusChem 2016, 9, 3407 – 3418

www.chemsuschem.org

[36] a) G. Kresse, J. Furthmüller, Comput. Mater. Sci. 1996, 6, 15 – 50; b) G.
Kresse, J. Furthmüller, J. Phys. Rev. B 1996, 54, 11169 – 11186.
[37] J. P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 1996, 77, 3865 –
3868.
ˇ
[38] a) S. Grimme, J. Comput. Chem. 2006, 27, 1787 – 1799; b) T. Bucko, J.
Hafner, S. Lebegue, J. G. Angyµn, J. Phys. Chem. A 2010, 114, 11814 –
11824.
´
[39] N. Almora-Barrios, G. Carchini, P. Błonski, N. López, J. Chem. Theory
Comput. 2014, 10, 5002 – 5009.
[40] a) P. E. Blçchl, Phys. Rev. B 1994, 50, 17953 – 17979; b) G. Kresse, D. Joubert, Phys. Rev. B 1999, 59, 1758 – 1775.
[41] D. Lide, CRC Handbook of Chemistry and Physics, 84th ed., CRC, Boca
Raton, FL, 2003–2004, pp. 12(19 – 21).
[42] M. Garcia-RatØs, N. López, J. Chem. Theory Comput. 2016, 12, 1331 –
1341.
[43] a) G. Henkelman, H. Jónsson, J. Chem. Phys. 2000, 113, 9978 – 9985; b) G.
Henkelman, B. P. Uberuaga, H. Jónsson, J. Chem. Phys. 2000, 113, 9901 –
9904.
[44] A. Heyden, A. T. Bell, F. J. Keil, J. Chem. Phys. 2005, 123, 224101 – 224114.
[45] M. lvarez-Moreno, C. de Graaf, N. López, F. Maseras, J. M. Poblet, C. Bo,
J. Chem. Inf. Model. 2015, 55, 95 – 103.

Received: June 7, 2016
Revised: September 2, 2016
Published online on October 14, 2016

3418

 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim