Metal-free dehydration of glucose to 5-(Hydroxymethyl)furfural in ionic liquids with boric acid as a promoter

Article abstract of DOI:10.1002/chem.201002171

The dehydration of glucose and other hexose carbohydrates to 5-(hydroxymethyl)furfural (HMF) was investigated in imidazolium-based ionic liquids with boric acid as a promoter. A yield of up to 42% from glucose and as much as 66% from sucrose was obtained. The yield of HMF decreased as the concentration of boric acid exceeded one equivalent, most likely as a consequence of stronger fructose-borate chelate complexes being formed. Computational modeling with DFT calculations confirmed that the formation of 1:1 glucose-borate complexes facilitated the conversion pathway from glucose to fructose. Deuterium-labeling studies elucidated that the isomerization proceeded via an ene-diol mechanism, which is different to that of the enzyme-catalyzed isomerization of glucose to fructose. The introduced non-metal system containing boric acid provides a new direction in the search for catalyst systems allowing efficient HMF formation from biorenewable sources. Isomerization is the key: Dehydration of glucose from resources such as cellulose, starch or sucrose to 5-(hydroxymethyl)furfural (HMF) is promoted by boric acid in imidazolium-based ionic liquids. The yields of HMF decrease as the concentration of boric acid exceeds one equivalent as a consequence of stronger fructose-borate chelate complexes being formed.

Full text of DOI:10.1002/chem.201002171

DOI: 10.1002/chem.201002171
Metal-Free Dehydration of Glucose to 5-(Hydroxymethyl)furfural in Ionic
Liquids with Boric Acid as a Promoter
[
a]
[b]
[b]
Tim Stꢀhlberg, Sergio Rodriguez-Rodriguez, Peter Fristrup,* and
[
a]
Anders Riisager*
Abstract: The dehydration of glucose
and other hexose carbohydrates to 5-
chelate complexes being formed. Com-
putational modeling with DFT calcula-
tions confirmed that the formation of
1:1 glucose–borate complexes facilitat-
ed the conversion pathway from glu-
cose to fructose. Deuterium-labeling
studies elucidated that the isomeriza-
tion proceeded via an ene–diol mecha-
nism, which is different to that of the
enzyme-catalyzed isomerization of glu-
cose to fructose. The introduced non-
metal system containing boric acid pro-
vides a new direction in the search for
catalyst systems allowing efficient
HMF formation from biorenewable
sources.
(
hydroxymethyl)furfural (HMF) was
investigated in imidazolium-based ionic
liquids with boric acid as a promoter.
A yield of up to 42% from glucose and
as much as 66% from sucrose was ob-
tained. The yield of HMF decreased as
the concentration of boric acid exceed-
ed one equivalent, most likely as a con-
sequence of stronger fructose–borate
Keywords: boric acid · catalysis ·
furfural · glucose · ionic liquids
[5–10]
Introduction
or ionic liquids with or without an acid catalyst.
The
most important industrial method for fructose production is
by the enzymatic isomerization of glucose syrup, a process
The impending exhaustion of fossil resources and climate
change has prompted an intensified research for new and
better pathways for chemicals and fuels from renewable
sources. Particular attention has been set on 5-(hydroxyme-
thyl)furfural (HMF), formed by the triple dehydration of
hexoses, which is expected to play an important role in a
[11]
yielding only about 50% fructose.
Direct conversion of
the obtained aqueous glucose/fructose mixture would be
problematical since the dehydration in water suffers from
side reactions by which HMF is rehydrated to leuvulinic
[12]
acid and formic acid. In order to find an economical and
environmentally feasible industrial process for the produc-
tion of HMF, an efficient direct conversion from glucose
would therefore be most beneficial. Unlike the dehydration
of fructose, however, the dehydration of glucose to HMF
demands a special catalyst to attain acceptable yields. This
[1,2]
future carbohydrate-based economy (Scheme 1).
HMF is
primarily considered to be a starting material for its diacid
counterpart (FDA) which is a possible replacement of ter-
[3]
ephthalic acid as a monomer in plastics. Reduction of the
furan ring would yield compounds suitable as solvents or
[4]
[13]
fuels (Scheme 1).
has so far been achieved by lanthanide chlorides, stannous
[14]
[15–18]
The formation of HMF from fructose is achieved readily
at elevated temperatures in high-boiling molecular solvents
chloride and chromium chlorides,
with only the latter
providing yields that could be adequate for a cost effective
process development. The main difficulty in the dehydration
is polymerization of the sugars during the reaction. The
polymers formed, commonly known as humins, vary in size
and are either soluble or insoluble in water depending on
chain length.
[
a] T. Stꢀhlberg, Prof. A. Riisager
Centre for Catalysis and Sustainable Chemistry
Department of Chemistry, Technical University of Denmark
2
800 Kgs. Lyngby (Denmark)
E-mail: ar@kemi.dtu.dk
In this work we have focused on finding an alternative to
chromium chloride as catalyst for the direct dehydration of
glucose to HMF. It is well known that carbohydrates form
[
b] Dr. S. Rodriguez-Rodriguez, Prof. P. Fristrup
Department of Chemistry, Technical University of Denmark
2
800 Kgs. Lyngby (Denmark)
[
19–21]
stable chelate complexes with boric acid
shown in Scheme 2 and catalyze the isomerization of aldo-
of the form
E-mail: pf@kemi.dtu.dk
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/chem.201002171.
[22,23]
hexoses to ketohexoses in aqueous basic environment.
1456
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2011, 17, 1456 – 1464

FULL PAPER
In the work we demonstrate the effect of boric acid in the
dehydration with different carbohydrates and bring clarity
to the mechanism by the aid of computational modeling and
experiments with deuterated glucose.
Results and Discussion
Dehydration of glucose: Based on the results from our pre-
[13]
vious work, we focused on the glucose dehydration using
alkylmethylimidazolium chlorides as the reaction media. To
demonstrate the effect of the boric acid as a promoter we
chose an initial catalyst load of 1.0 equivalent with respect
to glucose in order to form a 1:1 sugar/boric acid complex.
At a temperature of 1208C the dehydration of glucose
(
(
(
10 wt%)
in
1-ethyl-3-methylimidazolium
chloride
[EMIm]Cl) and 1-butyl-3-methylimidazolium chloride
[BMIm]Cl) with boric acid showed that up to 40% yield of
HMF could be obtained (Figure 1). The yield was higher in
EMIm]Cl than in [BMIm]Cl as is the case with chromium
[
[15]
catalysts. The yield appeared to reach a maximum in both
reactions after 3 h where after it slowly declined as a conse-
quence of degradation or polymerization. In the case of
[
5
EMIm]Cl the highest HMF selectivity of approximately
0% was reached after one hour of reaction. Full conver-
sion was not reached until after 21 h. The maximum ob-
tained HMF selectivity in [BMIm]Cl was of around 30%
which also coincided with the highest yield of around 20%
after 3 h. Close to full conversion in [BMIm]Cl was also
reached after 21 h.
Scheme 1. Formation of HMF from hexoses and derivatization further to
compounds with important applications in a post-petrochemical world,
for example, solvents and fuel.
Scheme 2. Formation of boric acid–diol complex.
The ability to form stable complexes with fructose has fur-
ther been utilized as a method of enriching fructose in the
isomerization of glucose to fructose by the enzyme glucose
[24]
isomerase. Boric acid is also known to catalyze several re-
actions with important synthetic applications such as esterifi-
[25]
[26]
[27]
cations,
cyclization reactions
and decarboxylations.
So far no successful dehydrations of glucose to HMF with
the aid of boric acid have been reported. With the assump-
tion that the boric acid-sugar interaction also might stabilize
intermediates or transition states in the isomerization of glu-
cose to fructose or even the dehydration to HMF, we sur-
mised that boric acid could act as a catalyst for the direct
conversion of glucose to HMF in ionic liquids. Compared
with the technologies developed up to now, this would be
beneficial from both an environmental as well as an eco-
nomical point of view, since boric acid is a non-toxic, non-
metal and inexpensive compound that exists in great abun-
dance in nature, primarily as sodium borate.
Figure 1. Dehydration of glucose in [EMIm]Cl at 1208C. The reaction
contained ionic liquid (1.0 g), glucose (100 mg, 0.56 mmol) and boric acid
(34.3 mg, 0.56 mmol); * = HMF yield; ~
HMF selectivity.
A
more extensive screening was further made in
[EMIm]Cl with a variation of boric acid concentration from
0 to 4.0 equivalents. The HMF yield increased markedly
from 0.1 equivalents and reached a maximum between 0.8
and 1.0 equivalents, where after it gradually declined
(Figure 2). The observed selectivity drop at high boric acid
Chem. Eur. J. 2011, 17, 1456 – 1464
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
1457

A. Riisager, P. Fristrup et al.
correlation between HMF yield and alkyl chain length on
the imidazolium cation of the ionic liquids could be ob-
served. Interestingly, methylimidazolium chloride ([MIm]Cl,
entry 1) only exhibited about half the yield of [EMIm]Cl.
Since [MIm]Cl is a protic ionic liquid one might have pre-
sumed this to inhibit the formation of more stable fructose–
borate complexes as is the case with a higher proton concen-
[
21]
tration in aqueous environments.
halide containing ionic liquids, HMF formation was only ob-
served in 1-ethyl-3-methylimidazolium ethylsulfate
[EMIm][C H OSO ]), choline methylsulfate ([Choline]-
[MeOSO ]), 1-ethyl-3-methylimidazolium methylsulfate
In the examined non-
(
AHCTUNGTRENNUNG
ACHTUNGTRENNUNG
2
5
3
3
(
[EMIm] ACHTUNTGNERNUG[ CH OSO ]), with a yield of 6%, suggesting that
3 3
the sulfate ion was of some importance (entries 6, 7 and 12).
The highest yield for the high-boiling molecular solvents
was obtained in DMSO (entry 15) with 13%. All these re-
sults substantiate the crucial role of the chloride in the con-
version of hexoses to HMF in ionic liquids. The reaction was
also performed in ethylene glycol (entry 17) to investigate
what effect a competing diol functionality had on the cata-
lytic performance. This resulted in no conversion of glucose
to HMF, which was probably caused by stronger binding of
the borate to the bulk solvent compared with that of glu-
cose. Dehydration attempts were also made using NaBF4
and sodium tetraborate as boron source in [EMIm]Cl, but
resulted in HMF yields below two percent. This suggested
that not only was the chloride anion vital, but also the
nature of the boron source. The effect of water on the
[EMIm]Cl/B(OH)3 system was also tested. Up to five
weight percent of water in the reaction mixture induced no
detrimental effect on HMF formation, which is consistent
Figure 2. Dehydration of glucose in [EMIm]Cl at 1208C with different
concentrations of boric acid (0–2.24 mmol), ionic liquid (1.0 g), and glu-
cose (100 mg, 0.56 mmol); * = HMF yield; ~ HMF selectivity.
content could be a result of formation of more stable sugar–
boric acid complexes as well as an increase in humin forma-
tion. Increasing the temperature above 1208C did not result
in a higher HMF yield, indicating that the yield of 42% is
the maximum yield obtainable with this procedure.
Dehydration in various solvents: To verify the necessity of
the chloride anion in HMF formation, a short survey of the
dehydration of glucose was performed in different media
such as high-boiling molecular solvents and various ionic liq-
uids together with 0.8 equivalents of boric acid. As shown in
Table 1, the best HMF yield was obtained in the chloride
containing ionic liquids where [EMIm]Cl still proved to be
[10]
with what previously has been reported. Nevertheless, in-
creasing the water concentration further resulted in a signifi-
cant decrease in HMF yield and above 30% no HMF was
formed. The fact that anhydrous or near anhydrous condi-
tions was essential for progression of the reaction suggests
that presence of water forms hydrated chloride ions which
are unable to take part in the reaction, in accordance with
[13]
the best (entries 1–5). Unlike with lanthanide catalysts, no
Table 1. Dehydration of glucose in various solvents with boric acid as
[
a]
catalyst.
Entry
Solvent
[MIm]Cl
[EMIm]Cl
[BMIm]Cl
Yield HMF/%
Selectivity HMF/%
[16]
the mechanism described by Binder and Raines.
1
2
3
4
5
6
7
8
9
A
C
H
T
U
N
G
T
R
E
N
N
U
N
G
19
41
14
32
26
6
6
0
0
0
0
6
7
2
13
0
0
20
43
30
47
41
57
8
0
0
0
0
25
11
23
37
0
ACHTUNGTRENNUNG
A
C
H
T
U
N
G
T
R
E
N
N
U
N
G
Dehydration of fructose: As mentioned above, fructose can
[
[
b]
c]
A
C
H
T
U
N
G
T
R
E
N
N
U
N
G
[HMIm]Cl
[OMIm]Cl
[EMIm]
[EMIm]
[EMIm]
[EMIm]
[EMIm][N(CN)
[BMIm][N(CN)
[choline][CH
be dehydrated directly to HMF in [EMIm]Cl without a cata-
A
C
H
T
U
N
G
T
R
E
N
N
U
N
G
[15]
lyst. To bring further clarity to whether the boric acid is
A
C
H
T
U
N
G
T
R
E
N
N
U
N
G
A
C
H
T
U
N
G
T
R
E
N
N
U
N
G
[CH
3
OSO
3
]
actually catalyzing the direct conversion to HMF, or merely
the isomerization of glucose to fructose, experiments with
fructose as starting compound were needed.
A
C
H
T
U
N
G
T
R
E
N
N
U
N
G
A
C
H
T
U
N
G
T
R
E
N
N
U
N
G
[C
2
H
5
OSO
3
]
A
C
H
T
U
N
G
T
R
E
N
N
U
N
G
A
C
H
T
U
N
G
T
R
E
N
N
U
N
G
[AlCl
4
]
A
C
H
T
U
N
G
T
R
E
N
N
U
N
G
A
C
H
T
U
N
G
T
R
E
N
N
U
N
G
[OAc]
10
11
12
13
14
15
16
17
A
C
H
T
U
N
G
T
R
E
N
N
U
N
G
2
2
]
]
In Figure 3 the results from experiments with fructose as
starting material together with various boric acid concentra-
tions are depicted (selectivity is excluded from the graph
since the conversion was between 98 and 100% in all reac-
tions). At low boric acid concentrations, that is, less than 0.2
equivalents, the obtained HMF yield was high around 80%.
Increasing the boric acid content further brought about a
gradual decrease in yield, which was even larger than that
for glucose. This clearly indicated that boric acid first and
foremost promoted the isomerization of glucose to fructose
and that a higher concentration inhibited further reaction
A
C
H
T
U
N
G
T
R
E
N
N
U
N
G
[
d]
A
C
H
T
U
N
G
T
R
E
N
N
U
N
G
A
C
H
T
U
N
G
T
R
E
N
N
U
N
G
3
OSO
3
]
[
e]
DMF
NMP
DMSO
[
f]
a-butyrolactone
ethylene glycol
0
[
2
a] Reaction conditions: 1.0 g solvent, 100 mg (0.56 mmol) glucose,
7.5 mg (0.44 mmol) boric acid, 1208C, 3 h. [b] [HMIm]=1-hexyl-3-meth-
ylimidazolium. [c] [OMIm]=1-octyl-3-methylimidazolium. [d] [Chol-
ine]=N,N,N-trimethylethanolammonium. [e] DMF=dimethylformamide.
[
f] NMP=N-methyl-2-pyrrolidone.
1458
www.chemeurj.org
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2011, 17, 1456 – 1464

Ionic Liquids
FULL PAPER
As expected, sucrose, consisting of a linked fructose unit
and glucose unit gave the highest HMF yield of 66%. Mal-
tose, a glucose dimer only amounted to 33% which was on
the same level as cellulose and starch. Reaction times for
obtaining the maximum HMF yield varied from 24 h for
starch, to 8 h for cellulose and the disaccharides. For com-
parison, all samples in Figure 4 were dehydrated using 0.5
equivalents of boric acid. Increasing the boric acid amount
to one equivalent actually resulted in a decrease in HMF
yield for all carbohydrates, most likely due to the fact that
glucose is formed in situ from depolymerization and present
in a much lower concentration than in the glucose experi-
ments.
Figure 3. Dehydration of fructose in [EMIm]Cl at 1208C with different
concentrations of boric acid (0–1.96 mmol), ionic liquid (1.0 g), and fruc-
tose (100 mg, 0.56 mmol).
Computational study: The results obtained from the experi-
mental work warranted a supplementary study using molec-
ular modeling to elucidate the reaction mechanism in detail.
In particular, we were interested in determining how the
presence of boric acid affected the relative energies of the
various intermediates along the reaction pathway. From the
literature the most feasible mechanisms could be narrowed
due to strong binding of boric acid to fructose. When the
boric acid concentration was increased beyond 2.0 equiva-
lents, the HMF yield was even less than 10%. At boric acid
contents higher than what corresponds to two moles of
boric acid per mole fructose one would assume the complex
with the sugar bound to two borate molecules to predomi-
nate. These complexes are likely to be more stable than
mono-borate complexes of fructose, something which was
[28–33]
down to two different pathways
where the rate-deter-
mining reaction step consists of either, a) hydride transfer
from C2 to C1, or b) the formation of an ene–diol inter-
mediate (also known as Lobryde Bruyn–van Ekenstein iso-
[34]
merization). For the enzymatic isomerization by glucose
isomerase deuterium labeling at the C2 position of glucose
[24]
[33]
also observed in earlier work by Takasaki when enriching
fructose from glucose with glucose isomerase.
has pinpointed the route to go via a 1,2-hydride shift. Re-
cently, a detailed study of the chromium(II) chloride cata-
[35]
lyzed reaction was published by Pidko et al. In analogy to
the enzymatic pathway the authors proposed a dinuclear
chromium(II) complex which facilitates the rate-determining
hydride shift from C2 to C1. An alternative mechanism in-
volving several protonation and deprotonation steps can
also be envisioned, in which a high-energy intermediate
ene–diol is formed. The results from the calculations of dif-
ferent complexes and intermediates along the isomerization
pathways are illustrated in Figure 5.
Dehydration of various carbohydrates: In the light of the
good dehydration results obtained with glucose, we wanted
to expand our boric acid–ionic liquid reaction system to the
dehydration of other carbohydrates. The direct conversion
of polysaccharides such as cellulose and starch is highly de-
sirable for a cost competitive process to make HMF. In
Figure 4 our results on the various sugars or polymers of
glucose capable of forming HMF are summarized.
Reaction pathway without boric acid: Starting from b-glu-
copyranose (1a) protonation of the anomeric oxygen, fol-
lowed by ring-opening and depronation gives the corre-
sponding open-chain form of glucose (1b) with a relative
À1
energy of 20 kJmol . The protonated open-chain glucose is
À1
a high-energy intermediate (1c, 141 kJmol ). An alterna-
tive deprotonation route could also be envisioned but when
glucose was deprotonated in the 2-position there was a
proton shift from O3 during minimization thus resulting in a
structure with O3 deprotonated which is no longer en route
À1
to the ene–diol (relative energy 28 kJmol , not shown).
From the protonated form of glucose formation of the neu-
tral ene–diol intermediate is very favorable (1d,
À1
5
4 kJmol ). We have located the transition state for the
direct hydride shift TS1b–1e but found the energy to be pro-
À1
hibitively high (183 kJmol ). From the ene–diol intermedi-
Figure 4. Dehydration of various carbohydrates in [EMIm]Cl at 1208C
with 0.5 equivalents of boric acid (1.0 g ionic liquid and 100 mg carbohy-
ate additional protonation can lead to formation of proton-
À1
drate: N maltose, * sucrose,
&
starch, ! cellulose.
ated fructose (154 kJmol , not shown) although in this case
Chem. Eur. J. 2011, 17, 1456 – 1464
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
1459

A. Riisager, P. Fristrup et al.
Figure 5. Overview of the two reaction pathways for isomerization of glucose to fructose. The stabilizing effect of boric acid coordination along the reac-
tion pathway is clear.
the deprotonated form of fructose is significantly more fa-
vored (1e, 82 kJmol ). Subsequently, the formation of the
open-chain form of fructose (1 f) can take place, which is
the generation of the open-chain glucose with boron coordi-
nated to the 3- and 4-positions (2b). Since the ring-opening
probably is faster than boric acid repositioning we have
started the pathway for isomerization with the 3,4-boroglu-
cose (2a). It is clear that the chelating effect of the boric
acid results in a stabilization of the open-chain analogue
where the O-B-O angles are close to the optimum for a tet-
rahedral geometry (107.5 and 109.38, optimum 109.58) com-
pared with the closed form where they are either too small
3,4-boroglucose (104.4 and 108.88) or too large 4,6-boroglu-
cose (113.4 and 115.38). Protonation of O1 is more favorable
À1
significantly more stable than the open-chain glucose (rela-
À1
tive energy 5 kJmol ). Finally, a ring-closure yields the final
b-fructofuranose (1g), and it should be noted that the over-
À1
all isomerization is energetically unfavored by 10.4 kJmol .
Reaction pathway with boric acid: For each intermediate
we investigated all possible complexes with boron to deter-
mine the energetically most favorable coordination site. For
glucose we found the complex with boron coordinated in
the 4- and 6-position to be most stable. When comparing to
the isolated b-glucopyranose the relative energy of the 4,6-
À1
than in the absence of boric acid (2c, À16 kJmol ), which
is probably because it is facilitated by the negatively charged
boric acid. Subsequent proton transfers results in the forma-
tion of the ene–diol intermediate which also has boron coor-
À1
boroglucopyronose (2a’) is À60.5 kJmol . Although this
À1
comparison involves molecules of different charge the size
of the molecules is large enough to allow for reasonable sol-
vation energies. Interestingly, with boric acid present the fol-
lowing ring-opening is now energetically favored resulting in
dinated in the 3- and 4-position (2d, À29 kJmol ). Also in
the presence of boric acid we succeeded in locating the tran-
À1
sition state for the hydride shift (TS2b–2f, 75 kJmol ) but
also here the energy is significantly higher than for the
1460
www.chemeurj.org
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2011, 17, 1456 – 1464

Ionic Liquids
FULL PAPER
route
involving
successive
proton transfers. From the ene–
diol intermediate additional
proton transfers can result in
the formation of the protonat-
ed, open form of 3,4-borofruc-
tose (2e) with a relative energy
À1
of À2 kJmol . Deprotonation
can give a quite stable open
form of 3,4-borofructose (2 f,
À1
À53 kJmol ), from which
a
Scheme 3. Theoretical amount of deuterium incorporated in HMF by the two possible isomerization mecha-
nisms.
final ring-closure furnishes the
fructofuranose with boron coor-
dinated in the 3- and 4-position
2g) with a relative energy of 0 kJmol . Alternatively, a re-
arrangement of the boric acid moiety to the 2,3-position is
À1
(
percent deuterium was incorporated in the final HMF prod-
uct (see Supporting Information). This result was compati-
ble with an ene–diol mechanism and the small amount of
deuterium actually incorporated could be explained by H/D
exchange with the bulk. Since the reaction was performed
under anhydrous conditions, a limited number of protons
are available and consequently some of the deuterium
atoms expelled in the first step could be available for incor-
poration further on during the course of reaction. Interest-
ingly, the experiment showed that the isomerization of glu-
cose to fructose using the borate–ionic liquid system pro-
ceeded via a different mechanism than that reported for the
À1
very favorable (2g’, À74 kJmol ), something which has also
been confirmed by NMR studies where 2g’ appeared to be
the most prevalent monoborate complex of fructofura-
nose. Further on, one could imagine the structure being
further stabilized as a very unreactive 2,3,4,6-diborofructose
[21]
À1
complex (3g, À96 kJmol ). We believe this significant sta-
bilization of the diboron complexes of fructose (both open
and closed forms) is responsible for the observed decrease
in efficiency when the boric acid/glucose ratio surpasses 1.5
(
see Figure 2) and the strong inhibition when converting
[33]
fructose to HMF (see Figure 3). For the route with mono-
coordinated boric acid the overall transformation from b-
glucopyranose to b-fructofuranose is energetically favored
enzyme glucose isomerase which reacts via a 1,2-hydride-
shift mechanism.
Based on this experiment, the results from the DFT calcu-
À1
[16,36,37]
by 14 kJmol which explains the increased rate of isomeri-
lations and previous work,
we could propose a puta-
zation in the presence of boric acid. It can be seen that the
effect of boron is both to lower the energy of the ene–diol
intermediate relative to the glucose and also to increase the
exothermicity of the overall isomerization from glucose to
fructose.
tive mechanism for the complete reaction of glucose to
HMF promoted by boric acid in imidazolium chlorides. As
shown in Scheme 4, the glucose–borate complex 2a gives a
favorable transition to 2g via the ene–diol intermediate 2d.
When 2g loses its borate and forms free fructose it can pro-
ceed by two different pathways: either react further to HMF
(pathway a) or form the more stable 2,3-borate complex 2g’
(pathway b). The energy of 2g is higher than that of 2g’ be-
cause the boric acid is bound trans to the diol functionality
of fructose, whereas it in the latter case is bound cis. Evi-
dently, the first step of fructose dehydration to HMF is
made impossible in complex 2g’ since the oxygen on C2 is
bound to boron and consequently water cannot be eliminat-
ed. Once 2g’ is formed it can in principle react further with
another borate to form the diborate complex 3g, which is
even more stable than 2g’. This would explain the lower
HMF yield at a boric acid content above 1.0 equivalent
mentioned in the previous section. When the reaction is
complete the boric acid is bound up as these stable borate
esters.
Dehydration of 2-[D]-glucose: In order to finalize an overall
mechanism for the entire dehydration of glucose to HMF
one of the isomerization mechanisms mentioned above
would have to be ruled out. By reacting glucose deuterated
at the C2 position a substantially different ratio of isotope-
labeled products would be obtained. An ene–diol mecha-
nism would expel all the deuterium into the solvent and
form the ketone on the C2 position of fructose. The HMF
product would therefore in theory have no deuterium incor-
poration though some minor exchange with the solvent
might be expected. In contrast, the 1,2-hydride shift mecha-
nism would result in a fructose-species containing 100%
deuteration at the C1 position. Accordingly, further reaction
to HMF would theoretically result in a product mixture in
which 50% deuterium would be retained on the aldehyde
based on the established mechanism of HMF formation
The mechanism of HMF formation from fructose has
been debated over the years, where both a pathway with
[16,36,37]
[36–40]
from fructose.
lustrated in Scheme 3.
The possible reaction scenarios are il-
cyclic intermediates
nism
as well as an open-chain mecha-
[
41–45]
have been proposed. Through extensive experi-
[
36]
We reacted 2-[D]-glucose using the standard dehydration
procedure and studied the resulting HMF by NMR and GC/
MS methods. This experiment showed that less than five
mental work Antal et al. concluded that the cyclic mecha-
nism is the most plausible of the two and the cyclic inter-
[37]
mediate 8 was recently identified through an NMR study.
Chem. Eur. J. 2011, 17, 1456 – 1464
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
1461

A. Riisager, P. Fristrup et al.
HMF was formed. The chal-
lenge for the future develop-
ment of this system is therefore
to find a catalyst for the dehy-
dration
of
fructose
in
[
EMIm]Cl which does not have
a detrimental impact on the
boric acid promoted isomeriza-
tion of glucose to fructose. This
could be achieved by screening
the reaction together with dif-
ferent catalysts known for cata-
lyzing the dehydration of fruc-
tose in ionic liquids. The addi-
tional catalyst must naturally
meet the criteria of being green
and sustainable to obtain
a
system that would be advanta-
geous to the previous transition
metal systems. An alternative
to this approach could imply
boric acid derivatives which
possibly display a different be-
havior towards isomerization.
Another vital aspect of the
introduced reaction system is
the ionic liquid. The structure
of the cation has evidently a
momentous impact on the
Scheme 4. Putative mechanism for the dehydration of glucose to HMF in imidazolium chlorides with boric
acid as promoter.
In the mechanism we have depicted the chloride ion acting
as a nucleophile, but it could in principal also be acting as a
base as proposed by Binder and Raines.
HMF yield. As mentioned in the previous section, there is
no clear reactivity pattern regarding the chain length of the
alkyl group on the alkylmethylimidazolium ion. Extended
screening of other cations might therefore result in the dis-
covery of a cation that would enhance the HMF yield even
more or give a rational explanation for the superior perfor-
mance of the 1-ethyl-3-methylimidazolium ion.
[16]
Conclusion
The boric acid–ionic liquid reaction system presented here
is the first metal-free system that catalyzes the conversion of
glucose and its polymeric counterparts to HMF. Even
though the yields do not surmount what has been achieved
We believe that with extensive experimental work and
modeling on different boric acid based catalytic systems the
yields of HMF could be enhanced and be a strong competi-
tor to other catalysts in a future biopetrochemical industry.
[15]
so far with chromium catalysts,
our results introduce a
new class of promoter not based on transition metals, which
could prove more practical for scaled up applications where
utilization of metals might be questionable from an environ-
mental viewpoint.
Experimental Section
Materials and equipment: All chemicals were used as received. d-glucose
The main obstacle with the boric acid–ionic liquid system
is the formation of stronger fructose–boric acid complexes
such as 2g’ which in due course stops the isomerization by
binding up boric acid and at the same time blocking the
elimination of water in the first step of the conversion to
HMF. The idea of adding a competitor for fructose as a
complexation agent falls on the fact that glucose forms
weaker complexes than fructose. Such a complexation agent
would be an even stronger competitor to glucose and would
most likely have the consequence of inhibiting the isomeri-
zation of glucose to fructose, something which was evident
when performing the reaction in ethylene glycol where no
(
99.5%), 2-[D]-d-glucose (98 atom% D), dimethyl sulfoxide (98%), cel-
lulose (powder, ca 20 micron), dimethylformamide (99%), a-butyrolac-
tone (98%) and N-methyl-2-pyrrolidone (99%) were purchased from Al-
drich. Boric acid (puriss), starch (p.a.) and d-fructose (puriss) were pur-
chased from Riedel-de Haꢂn. Ethylene glycol (normapur) and d-maltose
(
p.a.) were purchased from Prolabo. Sucrose (99%) was purchased from
Alfa Aesar. [EMIm][N(CN) ] (98%), [choline][CH OSO ] (98%) and
[EMIm][C H OSO ] (98%) were purchased from Solvent Innovation,
2
A
H
U
G
E
N
N
3
3
AHCTUNGTRENNUNG
2
5
3
while all other ionic liquids were obtained from BASF (>95%). The de-
hydration experiments were performed under nitrogen atmosphere using
a Radley Carousel 12 Plus Basic System with temperature control (+/
À18C). All samples were analyzed by HPLC (Agilent 1200 series, Bio-
Rad Aminex HPX-87H, 300 mm ꢃ 7.8 mm pre-packed column, 0.005m
À1
H
2
SO
4
mobile phase, 608C, 0.6 mLmin ). The HMF yields and selectivi-
1462
www.chemeurj.org
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2011, 17, 1456 – 1464

Ionic Liquids
FULL PAPER
ties were based on conversion of glucose and confirmed by calibration of
standard solutions of the products and reactants involved.
[12] A. F. V. Grote, B. Tollens, Justus Liebigs Ann. Chem. 1875, 175,
181–204.
[
13] T. Stꢀhlberg, M. G. Sørensen, A. Riisager, Green Chem. 2010, 12,
General dehydration procedure: A 40 mL reaction tube was charged
with ionic liquid (1 g) and boric acid (34.6 mg, 0.56 mmol) and heated at
321–325.
1
(
008C for 1 h. Glucose (100 mg, 0.56 mmol) or different carbohydrate
0.56 mmol) was then added and the solution stirred for 3 h at 1208C.
After reaction the reaction tube was cooled in an ice bath and water
5 mL) added. The solids were filtered off and the filtrate analyzed by
[14] S. Hu, Z. Zhang, J. Song, Y. Zhou, B. Han, Green Chem. 2009, 11,
1746–1749.
[15] H. Zhao, J. E. Holladay, H. Brown, Z. C. Zhang, Science 2007, 316,
1597–1600.
(
HPLC. Alternatively filtrate was extracted with ethyl acetate (3ꢃ30 mL)
and the solvent then removed in vacuo.
[16] J. B. Binder, R. T. Raines, J. Am. Chem. Soc. 2009, 131, 1979–1985.
[17] G. Yong, Y. Zhang, J. Ying, Angew. Chem. 2008, 120, 9485–9488;
Angew. Chem. Int. Ed. 2008, 47, 9345–9348.
[18] C. Li, Z. Zhang, Z. K. Zhao, Tetrahedron Lett. 2009, 50, 5403–5405.
[19] J. Verchere, M. Hlaibi, Polyhedron 1987, 6, 1415–1420.
[20] S. Chapelle, J.-F. Verchere, Tetrahedron 1988, 44, 4469–4482.
[21] R. van den Berg, J. A. Peters, H. van Bekkum, Carbohydr. Res. 1994,
253, 1–12.
[22] J. F. Mendicino, J. Am. Chem. Soc. 1960, 82, 4975–4979.
[23] K. B. Hicks, E. V. Symanski, P. E. Pfeffer, Carbohydr. Res. 1983, 112,
37–50.
[24] Y. Takasaki, Agric. Biol. Chem. 1971, 35, 1371–1375.
[25] T. A. Houston, B. L. Wilkinson, J. T. Blanchfield, Org. Lett. 2004, 6,
679–681.
Computational method: Density functional theory (DFT) in combination
[
46–48]
with the B3LYP functional
from Schrodinger Inc.
as incorporated in Jaguar version 7.6
was used for the calculations. In the current
study we employed the LACVP* basis set in Jaguar which uses the Hay–
[
49]
[
50]
Wadt small-core ECP and basis set for boron and the 6-31G* basis set
for the remaining elements. Solvation energies were obtained using the
[
51,52]
PB-SCRF solvation model
employing parameters suitable for di-
chloromethane (dielectric constant 9.08 and probe radius 2.33237 ꢄ). In
this model the solvent is described as a continuum, where the molecule is
put into a reaction field consisting of surface charges on a solvent accessi-
ble surface constructed using a hypothetical spherical solvent probe mol-
[
53]
ecule with the indicated radius.
The wavefunction and the reaction
field charges are solved iteratively until self-consistency is reached. Physi-
cal data on ionic liquids in the literature are still scarce, for example it
was not possible to find a dielectric constant for [EMIm]Cl. Thus, we
have in the current study used parameters suitable for dichloromethane
since its dielectric constant (e=7.26) was close to the ones reported of
[26] W. Moneta, P. Baret, J.-L. Pierre, J. Chem. Soc. Chem. Commun.
1985, 899–901.
[27] P. A. Wehrli, V. Chu, J. Org. Chem. 1973, 38, 3436–3436.
[28] Y. J. Topper, D. Stetten, J. Biol. Chem. 1951, 189, 191–202.
[
[
29] D. W. Harris, M. S. Feather, Carbohydr. Res. 1973, 30, 359–365.
30] T. Yamauchi, K. Fukushima, R. Yanagihara, S. Osanai, S. Yoshika-
wa, Carbohydr. Res. 1990, 204, 233–239.
[
54,55]
several 1-ethyl-3-methylimidazolium based ionic liquids.
[
56, 57]
In previous projects
we have found that energies calculated with the
continuum solvent model gives a fair correspondence with experimental
ratios, but for a quantitative agreement we frequently also need to ac-
count for the vibrational component of the free energy. The free energy
adjustment was then added to the corresponding energy determined in
solvent, to arrive at a composite free energy that is our best estimate of
the free energy in solvent.
[31] K. J. Schray, I. A. Rose, Biochemistry 1971, 10, 1058–1062.
[32] M. Whitlow, A. J. Howard, B. C. Finzel, T. L. Poulos, E. Winborne,
G. L. Gilliland, Proteins Struct. Funct. Genet. 1991, 9, 153–173.
[33] H. Hꢆusler, H. Weber, A. E. Stꢇtz, J. Carbohydr. Chem. 2001, 20,
239–256.
[34] C. A. L. de Bruyn, W. A. van Ekenstein, Recl. Trav. Chim. Pays-Bas
Belg. 1897, 16, 274–281.
[
[
[
35] E. Pidko, V. Degirmenci, R. van Santen, E. Hensen, Angew. Chem.
010, 122, 2584–2588; Angew. Chem. Int. Ed. 2010, 49, 2530–2534.
36] M. J. Antal, W. S. L. Mok, G. N. Richards, Carbohydr. Res. 1990, 199,
1–109.
37] A. S. Amarasekara, L. D. Williams, C. C. Ebede, Carbohydr. Res.
008, 343, 3021–3024.
2
9
Acknowledgements
2
The work was supported by the Danish National Advanced Technology
Foundation in cooperation with Novozymes A/S. Special thanks to BASF
for providing the ionic liquids. P.F. thanks the Danish Council for Inde-
pendent Research & Technology and Production Sciences for financial
support.
[
[
38] J. U. Nef, Justus Liebigs Ann. Chem. 1910, 376, 1–119.
39] W. N. Haworth, E. L. Hirst, V. S. Nicholson, J. Chem. Soc. 1927,
1
513–1526.
[
[
40] W. N. Haworth, W. G. M. Jones, J. Chem. Soc. 1944, 667.
41] M. L. Wolfrom, E. G. Wallace, E. A. Metcalf, J. Am. Chem. Soc.
1
942, 64, 265–269.
42] M. L. Wolfrom, R. D. Schuetz, L. F. Cavalieri, J. Am. Chem. Soc.
948, 70, 514–517.
43] M. L. Wolfrom, R. D. Schuetz, L. F. Cavalieri, J. Am. Chem. Soc.
949, 71, 3518–3523.
[
[
[
1] T. Werpy, G. Petersen, US DOE 2004, No. DOE/GO-102004, 1992,
http://www.nrel.gov/docs/fy04osti/35523.pdf.
1
[
[
2] J. J. Bozell, G. R. Petersen, Green Chem. 2010, 12, 539–554.
3] A. Boisen, T. Christensen, W. Fu, Y. Gorbanev, T. Hansen, J. Jensen,
S. Klitgaard, S. Pedersen, A. Riisager, T. Stꢀhlberg, J. Woodley,
Chem. Eng. Res. Des. 2009, 87, 1318–1327.
1
[
[
[
[
[
[
44] E. F. L. J. Anet, J. Am. Chem. Soc. 1960, 82, 1502–1502.
45] E. Anet, Aust. J. Chem. 1961, 14, 295–301.
46] A. D. Becke, J. Chem. Phys. 1993, 98, 5648–5652.
47] A. D. Becke, J. Chem. Phys. 1993, 98, 1372–1377.
48] C. Lee, W. Yang, R. G. Parr, Phys. Rev. B 1988, 37, 785–789.
49] Schrçdinger, Jaguar 7.6, release 108 (Ed.: Schrçdinger), LLC, New
York, NY, 2009.
[
4] Y. Romꢅn-Leshkov, C. J. Barrett, Z. Y. Liu, J. A. Dumesic, Nature
2
007, 447, 982–985.
5] Y. Nakamura, S. Morikawa, Bull. Chem. Soc. Jpn. 1980, 53-53,
705–3706.
[
3
[
[
[
6] C. Fayet, J. Gelas, Carbohydr. Res. 1983, 122, 59–68.
7] R. M. Musau, R. M. Munavu, Biomass 1987, 13, 67–74.
8] C. Moreau, A. Finiels, L. Vanoye, J. Mol. Catal. A 2006, 253, 165–
[
[
50] P. J. Hay, W. R. Wadt, J. Chem. Phys. 1985, 82, 299–310.
51] D. J. Tannor, B. Marten, R. Murphy, R. A. Friesner, D. Sitkoff, A.
Nicholls, B. Honig, M. Ringnalda, W. A. Goddard, J. Am. Chem.
Soc. 1994, 116, 11875–11882.
1
69.
9] Q. Bao, K. Qiao, D. Tomida, C. Yokoyama, Catal. Commun. 2008, 9,
383–1388.
[
1
[52] B. Marten, K. Kim, C. Cortis, R. A. Friesner, R. B. Murphy, M. N.
Ringnalda, D. Sitkoff, B. Honig, J. Phys. Chem. 1996, 100, 11775–
11788.
[
[
10] S. Hu, Z. Zhang, Y. Zhou, B. Han, H. Fan, W. Li, J. Song, Y. Xie,
Green Chem. 2008, 10, 1280–1283.
11] R. L. Antrim, W. Colilla, B. J. Schnyder, Appl. Biochem. Bioeng.
[53] C. Cramer, Essentials of Computational Chemistry: Theories and
Models, Wiley, New York, 2002.
1
979, 2, 97–155.
Chem. Eur. J. 2011, 17, 1456 – 1464
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
1463

A. Riisager, P. Fristrup et al.
[
54] C. Wakai, A. Oleinikova, M. Ott, H. Weingꢆrtner, J. Phys. Chem. B
005, 109, 17028–17030.
[57] J. H. Dam, P. Fristrup, R. Madsen, J. Org. Chem. 2008, 73, 3228–
2
3235.
[
[
55] H. Weingꢆrtner, Z. Phys. Chem. 2006, 220, 1395–1405.
56] L. Keinicke, P. Fristrup, P.-O. Norrby, R. Madsen, J. Am. Chem. Soc.
Received: July 29, 2010
2
005, 127, 15756–15761.
Published online: January 10, 2011
1464
www.chemeurj.org
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2011, 17, 1456 – 1464