Conversion of hemicellulose sugars catalyzed by formic acid: Kinetics of the dehydration of D -xylose, L -arabinose, and D -glucose

Article abstract of DOI:10.1002/cssc.201403328

The pre-treatment of lignocellulosic biomass produces a liquid stream of hemicellulose-based sugars, which can be further converted to high-value chemicals. Formosolv pulping and the Milox process use formic acid as the fractionating agent, which can be used as the catalyst for the valorisation of hemicellulose sugars to platform chemicals. The objective of this study was to investigate the reaction kinetics of major components in the hemicelluloses fraction of biomass, that is, D-xylose, L-arabinose and D-glucose. The kinetics experiments for each sugar were performed at temperatures between 130 and 170 °C in various formic acid concentrations (10-64wt %). The implications of these kinetic models on the selectivity of each sugar to the desired products are discussed. The models were used to predict the reaction kinetics of solutions that resemble the liquid stream obtained from the fractionation process of biomass using formic acid.

Full text of DOI:10.1002/cssc.201403328

DOI: 10.1002/cssc.201403328
Full Papers
Conversion of Hemicellulose Sugars Catalyzed by Formic
Acid: Kinetics of the Dehydration of d-Xylose, l-Arabinose,
and d-Glucose
Karla Dussan,[a] Buana Girisuta,*[b] Marystela Lopes,[a] James J. Leahy,[a] and
Michael H. B. Hayes[a]
The pre-treatment of lignocellulosic biomass produces a liquid
stream of hemicellulose-based sugars, which can be further
converted to high-value chemicals. Formosolv pulping and the
Milox process use formic acid as the fractionating agent, which
can be used as the catalyst for the valorisation of hemicellu-
lose sugars to platform chemicals. The objective of this study
was to investigate the reaction kinetics of major components
in the hemicelluloses fraction of biomass, that is, d-xylose, l-
arabinose and d-glucose. The kinetics experiments for each
sugar were performed at temperatures between 130 and
1708C in various formic acid concentrations (10–64 wt%). The
implications of these kinetic models on the selectivity of each
sugar to the desired products are discussed. The models were
used to predict the reaction kinetics of solutions that resemble
the liquid stream obtained from the fractionation process of
biomass using formic acid.
Introduction
Furanic compounds, for example, furfural and 5-hydroxyme-
thylfurfural (HMF), as well as organic acids, for example, levu-
linic and formic acids, are important renewable bulk chemicals
derived from lignocellulosic materials, such as agricultural
wastes and energy crops. Approximately 60–62% of the world-
wide production of furfural is used to produce furfuryl alcohol,
an important polymer precursor that is obtained mainly
through the hydrogenation of furfural in the gas phase over
Cu catalysts at mild temperatures.[1] Other furfural derivatives,
such as 2-methylfuran and 2-methyltetrahydrofuran, have been
designated as fuel additives.[2] Lange et al.[2] studied the pro-
duction of ethyl furfuryl ether derived from furfuryl alcohol
and its utilisation as a gasoline additive because of its high
octane number and high energy density (28 MJkgꢀ1). The car-
bonyl and hydroxyl groups of HMF can be transformed
through oxidation and esterification reactions to various bulk
and fine chemicals for a wide range of applications, such as
2,5-furandicarboxylic acid, 2,5-diformylfuran and others. Addi-
tionally, base-catalysed aldol condensation reactions of furfural
and HMF with other aldehydes or ketones at moderate tem-
peratures (50–1208C) are considered as a promising route for
the production of C8–C13 alkanes through a subsequent reduc-
tion of the condensed molecules.[3,4] Levulinic acid (LA) has
been recognised as a versatile building block and highly valua-
ble precursor of biofuels. LA derivatives, such as ethyl and n-
butyl levulinates, have been blended with fossil-based diesel
successfully up to 20%.[5] LA can also be hydrogenated selec-
tively to g-valerolactone, which can be further transformed to
alkane fuel mixtures.[6]
Furfural, HMF and LA can be produced by a complex reac-
tion network of the hemicelluloses fraction of biomass in
acidic aqueous media (Scheme 1). Initially, the glycosidic link-
ages of the hemicellulose are protonated and subsequently hy-
drolysed to release pentoses (i.e., xylose and arabinose) and
hexoses (i.e., glucose, mannose and galactose). The conversion
of d-xylose to furfural is initiated by the protonation of the hy-
droxyl groups of d-xylose in its pyranose form. Some investiga-
tions have suggested that once one molecule of water is elimi-
nated from the d-xylose molecule, an intermediate dehydrofur-
anose is formed, which would dehydrate further to form furfu-
ral.[7,8] The intermediate dehydrofuranose may react with vari-
ous functional groups present in the reaction medium, for
example, the aldehyde group in furfural or the nucleophilic
groups in lignin-derived compounds, to lead to non-valuable
byproducts.[9] Likewise, furfural is known to undergo resinifica-
tion and decomposition reactions through the hydrolytic cleav-
age of its saturated ring. These undesired side reactions are
promoted by acid catalysts and high temperatures[10] and
eventually lead to low selectivity for furfural production in
aqueous acidic media. Although l-arabinose is not especially
abundant in most common crops and agricultural residues
(~15% of total hemicelluloses), its content in bark fractions of
pine wood (Pinus sylvestris, Picea abies) and spruce wood
(Picea abies) can be much higher than in other lignocellulosic
biomass. The studies of Hosia et al. and Eskilsson and Hartler
(cited by Hayes[11]) reported that l-arabinose corresponded to
approximately 50–70% of the total pentoses found in the
stem bark of these wood species. Heartwood from species
[a] Dr. K. Dussan, M. Lopes, Dr. J. J. Leahy, Prof. M. H. B. Hayes
Chemical and Environmental Sciences Department
University of Limerick
Castletroy, Co. Limerick (Ireland)
[b] Dr. B. Girisuta
Institute of Chemical and Engineering Sciences
1 Pesek Road, Jurong Island, 627833 (Singapore)
E-mail: buana_girisuta@ices.a-star.edu.sg
ChemSusChem 0000, 00, 0 – 0
1
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
These are not the final page numbers! ÞÞ

Full Papers
Scheme 1. Reaction network of the acid-catalysed hydrolysis of hemicellulose in lignocellulosic biomass.
such as Larix lariciana also has relatively high amounts of arabi-
nose (~40% of total pentoses).[12] The furfural production from
these wood species can be mostly attributed to the dehydra-
tion of arabinose to furfural. Therefore, comprehensive studies
on the conversion of d-xylose and l-arabinose to furfural in
acidic media are essential for the development of lignocellulo-
sic biorefineries that comprise a wide range of feedstocks,
which include wood and wood residues.
by Lamminpꢁꢁ includes the formation of an unknown inter-
mediate from d-xylose before the formation of furfural, which
reacts subsequently with this intermediate to form the con-
densation products.[20] Kupiainen et al.[22] investigated the con-
version of glucose catalysed by formic acid (<20 wt%) at high
temperatures (>1808C). They found that HMF was formed via
an intermediate compound through first-order reaction kinet-
ics.
The total amount of hexoses from the hydrolysis reaction of
hemicellulose is generally low in grasses and hardwood materi-
als; however, depending on the type of lignocellulosic material
and the conditions of hydrolysis reaction, the amount of glu-
cose release can account for up to 40% of the glucan content
in cellulose with high formic acid concentrations at high tem-
peratures.[13] As a result of the low activity of hexoses towards
the enolisation reaction in water, all hexoses show generally
low selectivities to HMF[14] and require high temperatures and
strong acid catalysts to obtain significant yields of HMF.[15] HMF
can be hydrated easily in the aqueous phase at low tempera-
tures to form LA and formic acid.[16] However, undesired side
reactions of HMF to the condensation products also occur
easily, which decrease the yields of LA and formic acid.
Biomass fractionation processes that use formic acid as the
fractionation solvent (i.e., Formosolv pulping and the Milox
process) have been implemented not only as technologies for
paper production but also as a pre-treatment process for the
biochemical conversion of cellulosic pulps.[13,22–24] The hemicel-
lulose sugars released from these types of fractionation pro-
cesses can be further valorised in the formic acid solvent that
also acts as the catalyst to form platform chemicals such as fur-
fural, HMF and LA.[25] The aim of this study was to evaluate the
kinetics of the dehydration reactions of d-xylose, l-arabinose
and d-glucose in formic acid solution (10–64 wt%) at tempera-
tures between 130 and 1708C. Additionally, the effects of the
initial concentrations of substrates and products on the selec-
tivity of the dehydration reactions to the desired products
were also investigated and taken into account to validate the
reaction mechanisms. Finally, the kinetic models developed for
each sugar were combined and used to model the reaction ki-
netics of solutions that resemble the liquors obtained from the
fractionation process using formic acid.
Although numerous kinetic studies have been performed in
the past, most of these studies simplify the conversion reac-
tions of hemicellulose sugars to pseudo-first-order kinetics.[17–19]
Weingarten et al.[20] found that the kinetics of the d-xylose de-
hydration reaction catalysed by hydrochloric acid (0.1m) in a bi-
phasic system (water/methyl isobutyl ketone) followed
a second-order reaction mechanism in which d-xylose and fur-
fural reacted to form the decomposition products. Lamminpꢁꢁ Results and Discussion
et al.[21] proposed a more complex reaction mechanism for the
Reaction mechanisms
kinetics of the xylose dehydration reaction in an aqueous
medium catalysed by formic acid (<30 wt%) at temperatures
between 160 and 2008C. The reaction mechanism presented
Despite the number of studies on xylose dehydration, a reac-
tion mechanism for the kinetics determination has not yet
&
ChemSusChem 0000, 00, 0 – 0
www.chemsuschem.org
2
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
been collectively agreed. Three of the most common reaction
mechanisms found in the literature for the conversion of pen-
toses are described in Scheme 2. The most significant differ-
ence in these mechanisms is related to the side reactions that
affect the selectivity of the dehydration reactions of d-xylose
(XYL) or l-arabinose (ARA). The first mechanism involves first-
order reaction rates for the conversion of XYL/ARA to furfural
and to condensation products (COND). The second mechanism
implies the conversion of XYL/ARA to furfural via an intermedi-
ate species, as proposed from mechanistic and molecular mod-
elling studies.[7,8] This intermediate compound (INT) reacts with
furfural to form condensation products following a second-
order reaction rate. The third mechanism describes the forma-
tion of condensation products between furfural and XYL/ARA.
These three reaction mechanisms propose that furfural would
undergo a self-degradation reaction to form resinification
products (RES).
Equations (1)–(11) also represent the rate equations for the
modelling of the kinetics of ARA, in which XYL can be replaced
with the corresponding concentration of ARA, and the reaction
rate constants will correspond to the arabinose system (kA1, kA2,
kA3).
Following a similar approach, several reaction mechanisms
were investigated for the conversion of d-glucose (GLC) to
HMF, which is converted subsequently to LA and formic acid
(Scheme 3). Mechanism 1 follows a reaction network in which
the formations of humin-type byproducts from GLC and HMF
follow first-order reaction rates. Mechanism 2 considers the
conversion of GLC to an intermediate compound, which is con-
verted to HMF and reacts subsequently with HMF following
a second-order reaction rate to form the condensation byprod-
ucts. Mechanism 3 includes a second-order reaction in which
GLC and HMF react to form the condensation byproducts. The
intermediate compounds considered in Mechanism 2 for XYL,
ARA and GLC were not isolated or identified in the analytical
systems used in this study.
The reaction rate equations used for the modelling of the ki-
netics that follow the mechanisms described in Scheme 2 for
d-xylose are presented as follows:
The reaction rate equations used for the modelling of the ki-
netics that follow the mechanisms described in Scheme 3 for
d-glucose are presented as follows:
Formation of resinification products from furfural (FUR)
[Eq. (1)]:
Mechanism 1:
d½RESꢁ
ð1Þ
¼ kR½FURꢁ
d½GLCꢁ
dt
ð12Þ
ð13Þ
ð14Þ
ð15Þ
¼ ꢀðkG1 þ kG3Þ½GLCꢁ
dt
Mechanism 1 [Eqs. (2)–(4)]:
d½HMFꢁ
dt
¼ kG1½GLCꢁ ꢀ ðkG2 þ kG4Þ½HMFꢁ
d½XYLꢁ
ð2Þ
ð3Þ
ð4Þ
¼ ꢀðkX1 þ kX2Þ½XYLꢁ
dt
d½LAꢁ
¼ kG2½HMFꢁ
dt
d½FURꢁ
¼ kX1½XYLꢁ ꢀ kR½FURꢁ
dt
d½CONDꢁ
dt
¼ kG3½GLCꢁ þ kG4½HMFꢁ
d½CONDꢁ
¼ kX2½XYLꢁ
dt
Mechanism 2:
d½GLCꢁ
Mechanism 2 [Eqs. (5)–(8)]:
ð16Þ
ð17Þ
ð18Þ
ð19Þ
ð20Þ
¼ kG1½GLCꢁ
dt
d½XYLꢁ
¼ ꢀkX1½XYLꢁ
dt
ð5Þ
ð6Þ
ð7Þ
ð8Þ
d½INTꢁ
dt
¼ kG1½GLCꢁ ꢀ kG2½INT ꢂ Gꢁ ꢀ kG3½INT ꢂ Gꢁ½HMFꢁ
d½INTꢁ
d½HMFꢁ
dt
¼ kX1½XYLꢁ ꢀ kX2½INTꢁ½FURꢁ ꢀ kX3½INTꢁ
dt
¼ kG2½INT ꢂ Gꢁ ꢀ kG3½INT ꢂ Gꢁ½HMFꢁ ꢀ kG4½HMFꢁ
d½FURꢁ
d½LAꢁ
¼ kX3½INTꢁ ꢀ kX2½INTꢁ½FURꢁ ꢀ kR½FURꢁ
dt
¼ kG4½HMFꢁ
dt
d½CONDꢁ
¼ kX2½INTꢁ½FURꢁ
dt
d½CONDꢁ
dt
¼ kG3½INT ꢂ Gꢁ½HMFꢁ
Mechanism 3 [Eqs. (9)–(11)]:
Mechanism 3:
d½XYLꢁ
d½GLCꢁ
ð9Þ
ð10Þ
ð11Þ
ð21Þ
ð22Þ
ð23Þ
¼ ꢀkX1½XYLꢁ ꢀ kX2½XYLꢁ½FURꢁ
dt
¼ ꢀkG1½GLCꢁ ꢀ kG3½GLCꢁ½HMFꢁ
dt
d½HMFꢁ
dt
d½FURꢁ
¼ kG1½GLCꢁ ꢀ kG2½HMFꢁ ꢀ kG3½GLCꢁ½HMFꢁ
¼ kX1½XYLꢁ ꢀ kX2½XYLꢁ½FURꢁ ꢀ kR½FURꢁ
dt
d½LAꢁ
d½CONDꢁ
¼ kX2½XYLꢁ½FURꢁ
dt
¼ kG2½HMFꢁ
dt
ChemSusChem 0000, 00, 0 – 0
www.chemsuschem.org
3
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
water in the mixture. The degra-
dation of furfural in deionised
water was similar to that ob-
served in the dilute solutions of
formic acid. The presence of the
dissociated
hydronium
ion
(0.020–0.045m) in the solutions
of formic acid (30–62 wt%) used
in this study barely enhanced
the degradation rate of furfural.
This observation is in line with
previous studies on the degrada-
tion of furfural in dilute solutions
of formic acid (<30 wt%).[21]
However, if we used significantly
high formic acid concentrations
(>60 wt%), the effects of water
as a solvent on the rate constant
of the degradation of furfural
must be taken into account.
The kinetics of the furfural
degradation reaction was mod-
elled following a first-order reac-
tion rate in which the reaction
rate constant was expressed as
a function of temperature fol-
lowing the Arrhenius equation
[Eq. (25)]:
ꢀ
ꢁ
ꢀEA;R
kR ¼ AR exp
ð25Þ
RT
in which AR corresponds to the
frequency factor (minꢀ1), EA,R is
the activation energy (kJmolꢀ1),
and R is the universal gas con-
stant. The kinetic parameters
Scheme 2. Reaction mechanisms for the kinetic study of the acid-catalysed dehydration of d-xylose (XYL) and l-
arabinose (ARA) and the degradation of furfural (FUR).
d½CONDꢁ
ð24Þ
¼ kG3½GLCꢁ½HMFꢁ
dt
Furfural degradation in concentrated formic acid
A series of nine experiments was performed to assess the deg-
radation rates of furfural at various concentrations of formic
acid (27–60 wt%). The degradation of furfural was barely af-
fected by the formic acid concentration. For example, the deg-
radation of furfural (100 mm) at 1708C after 100 min of reac-
tion time was 14.2, 15.2 and 12.0 mol% in formic acid solutions
of 30, 47 and 62 wt%, respectively. Additional experiments
were performed in concentrated formic acid (98 wt%) and de-
ionised water to validate whether the furfural degradation re-
action was independent of the concentration of formic acid.
These experimental results are presented in Figure 1.
If concentrated formic acid (98 wt%) was used as the react-
ing solvent, only 5% of the furfural was degraded after
160 min. This indicated that furfural is stable in concentrated
formic acid solution, that is, with a very low concentration of
Figure 1. The degradation of the furfural (CFUR,i =100 mm) at 1708C in vari-
~
&
*
ous reacting solvents: , deionised water; , 29.5 wt% HCOOH; , 46.5 wt%
HCOOH; 3, 62.1 wt% HCOOH; N, concentrated formic acid (98 wt%
HCOOH).
&
ChemSusChem 0000, 00, 0 – 0
www.chemsuschem.org
4
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
Scheme 3. Reaction mechanisms for the kinetic study of the acid-catalysed dehydration of d-glucose (GLC) and the conversion of HMF to LA and formic acid.
Table 1. Kinetic parameters for the furfural degradation in concentrated
formic acid.
Rate
Pre-exponential factor Activation energy
Fit to experi-
mental data R2
constant Ln(AR) [minꢀ1
]
EA,R [kJmolꢀ1
]
kR 15.70ꢃ1.26
81.70ꢃ4.45
0.923
that fit the experimental values satisfactorily (R2 =0.923) are
presented in Table 1 and Figure 2.
The activation energy agreed with the values obtained from
previous studies on the kinetics of xylose dehydration and fur-
fural destruction. Lamminpꢁꢁ et al.[21] reported an activation
energy of 75.5 kJmolꢀ1 for the hydrothermal degradation of
furfural at temperatures between 130 and 1808C in formic acid
solutions (7 and 30 wt%). Other reports gave activation ener-
gies for furfural degradation between 67 and 131 kJmolꢀ1 if
catalysed by hydrochloric acid[17,20,26] and between 60 and
106 kJmolꢀ1 if catalysed by sulfuric acid.[17,26]
Figure 2. Comparisons between the experimental data and the kinetic
model (solid lines) of furfural degradation reaction (CFUR,i =100 mm).
ChemSusChem 0000, 00, 0 – 0
www.chemsuschem.org
5
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
Dehydration of d-xylose to furfural
the mean value of the experimental data and n is the total
number of data.
Kinetic modelling
The values of GF% determined for the three reaction mech-
anisms show significant differences, especially the GF% for fur-
fural concentration. The kinetic models obtained from Mecha-
nisms 1 and 2 gave higher GF% for furfural concentration (85.0
and 91.1%, respectively) than that from Mechanism 3 (76.3%).
However, there was no significant difference for the GF%
values for xylose concentration (88.3–89.4%). Based on this
statistical analysis, Mechanism 2 describes the dehydration re-
action of d-xylose to furfural most suitably. To support this
statistical analysis, additional experiments were performed at
different initial concentrations of d-xylose and furfural. These
experiments were performed at 150 and 1708C using 40 and
64 wt% HCOOH as the catalyst. The experimental data and the
prediction results obtained from the three sets of kinetic pa-
rameters given in Table 2 are shown in Figure 3. The kinetic
model derived from Mechanism 1 failed to estimate the rate of
furfural formation at a low initial concentration of d-xylose
(Figure 3a). If furfural was added at the beginning of the reac-
tion (Figure 3c and d), the kinetic model derived from Mecha-
nism 1 overestimated the formation of furfural because this re-
action mechanism does not include the undesired side reac-
tions between xylose and furfural, which could occur from the
beginning of the reaction. The kinetic models derived from
Mechanisms 2 and 3 can predict the formation of furfural more
accurately in every scenario shown in Figure 3. However, the
concentrations of d-xylose predicted by the kinetic model de-
rived from Mechanism 3 were lower than the experimental
data that were obtained in the presence of furfural (Figure 3c
and d). In this case, the conversion of d-xylose was overesti-
mated because the kinetic model derived from Mechanism 3
assumed that d-xylose was also consumed from the beginning
of reaction following a second-order reaction between d-
xylose and furfural. The fact that the conversion of d-xylose
was not affected in the presence of significant amounts of fur-
fural at the beginning of the reaction supports the hypothesis
that furfural only reacts with the intermediate derived from d-
xylose to form the condensation products. Based on this analy-
sis, Mechanism 2 was chosen as the most appropriate model
to describe the dehydration reaction of d-xylose to furfural in
a solution of formic acid as the catalyst.
The concentrations of d-xylose and furfural (200 data points in
total), which were obtained from nine different sets of experi-
mental conditions that combine three temperatures (130, 150,
1708C) and three formic acid concentrations (30, 50, 64 wt%),
were used to obtain the kinetic models based on the three
proposed reaction mechanisms presented in Scheme 2. The ki-
netic parameters for all the experimental data were obtained
through an estimation routine using the fminsearch optimisa-
tion function in Matlab. In this routine, the matrix determinant
of the residual errors for d-xylose and furfural concentrations
was used as the estimation criterion.[27] The kinetic model for
the furfural degradation reaction was also included in the esti-
mation routine. A modified Arrhenius equation that includes
the effects of temperature T and the effective acid concentra-
tion [H+] was used to fit the kinetic parameters [Eq. (26)]:
ꢀ
ꢁ
ꢀEA;i
mi
ki ¼ Ai exp
½Hþꢁ
ð26Þ
RT
The kinetic parameters in Equation (26) are the frequency
factor Ai [minꢀ1], the activation energy EA,i [kJmolꢀ1], and the
reaction order of [H+]m (dimensionless). All the kinetic parame-
i
ters determined for the three reaction mechanisms are listed
in Table 2. The goodness-of-fit (GF%) between the experimen-
Table 2. Kinetic parameters for the acid-catalysed dehydration of d-
xylose.
GF%[a]
Parameters
Rate constants
kX1
kX2
kX3
Ln(Ao,i) [minꢀ1 mꢀm
]
40.2ꢃ1.2 42.2ꢃ2.1
136.0ꢃ3.0 140.5ꢃ5.4
–
–
–
i
[XYL]=89.4%
[FUR]=85.0%
1
2
3
EA,i [kJmolꢀ1
]
mi
1.9ꢃ0.1
2.3ꢃ0.2
Ln(Ao,i) [minꢀ1 mꢀm
42.6ꢃ0.9 26.9ꢃ4.2[b] 27.3ꢃ3.2
140.3ꢃ2.4 82.9ꢃ11.1 91.5ꢃ8.5
i
]
[XYL]=89.4%
[FUR]=91.1%
EA,i [kJmolꢀ1
]
mi
2.1ꢃ0.1
1.5ꢃ0.5
1.1ꢃ0.4
Ln(Ao,i) [minꢀ1 mꢀm
41.2ꢃ1.3 41.7ꢃ3.4[b]
137.6ꢃ3.3 128.7ꢃ9.0
–
–
–
i
]
[XYL]=88.3%
[FUR]=76.3%
EA,i [kJmolꢀ1
]
mi
2.0ꢃ0.2
2.4ꢃ0.4
[a] Goodness-of-fit%. [b] minꢀ1 mꢀ1ꢀm
.
i
tal and modelled data for the concentrations of d-xylose and
furfural are also reported in Table 2 and were calculated as fol-
lows [Eq. (27)]:
Implications of the kinetics
The experimental data and the predicted profiles of d-xylose
conversion and furfural yields from the kinetics experiments
are presented in Figure 4. Furfural yields of between 40 and
55 mol% were obtained at temperatures between 130 and
1708C if initial d-xylose concentrations between 200 and
270 mm were used. The highest temperature used in this
study (1708C) gave the highest furfural yields (50–55 mol%).
An increase of the formic acid concentration improved the fur-
fural yield slightly, for example, an increase of the formic acid
concentration from 30.0 to 63.5 wt% at 1708C only enhanced
the furfural yield by 5%.
0
1
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
n
À
Á
P
2
^
Ci ꢀ C
B
B
C
C
C
A
i¼1
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
GF% ¼ 1 ꢀ
ꢄ 100
ð27Þ
B
n
P
@
2
ꢀ
ðCi ꢀ CÞ
i¼1
in which Ci corresponds to the concentration of d-xylose or
ˆ
furfural obtained from the experiments, Ci is the concentration
¯
of each compound predicted using the kinetic models, Ci is
&
ChemSusChem 0000, 00, 0 – 0
www.chemsuschem.org
6
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
&
*
Figure 3. Concentrations of d-xylose ( ) and of furfural ( ) with different initial d-xylose/furfural compositions and profile predictions from Mechanisms 1
(dashed line), 2 (solid black line) and 3 (solid grey line).
The activation energy of the conversion of d-xylose reported
in this study (140.3 kJmolꢀ1) was in agreement with values re-
ported by Lavarack et al.[17] (111.2–118.9 kJmolꢀ1), Nabarlatz
et al.[28] (122.5 kJmolꢀ1), Morinelly et al.[26] (106–132 kJmolꢀ1),
Weingarten et al.[20] (123.9 kJmolꢀ1) and Lamminpꢁꢁ et al.[21]
(153 kJmolꢀ1). The reaction orders of [H+] indicated that the
formic acid concentration affects the conversion rate of d-
xylose and the formation rate of furfural significantly. This is
presented in Figure 4, which shows that faster conversions of
d-xylose were observed at higher concentrations of formic
acid, particularly at low temperatures. However, the selectivity
of furfural formation was not increased extensively with the in-
crease of the acid concentration because of competing side re-
actions, which were also promoted by the acid to a similar
extent. The semi-empirical models demonstrated that the reac-
tion orders of the acid concentration for the rate constants kX1
and kX2 varied between 1.5 and 2.1; nevertheless, the mecha-
nistic pathway associated with this is yet unclear and requires
further investigation.
Miscanthus x giganteus (~10:1 ratio). A total of 180 experimen-
tal data points that consist of l-arabinose and furfural concen-
trations obtained from these experiments were used to deter-
mine the kinetic parameters for the kinetic models according
to the reaction mechanisms shown in Scheme 2 and the modi-
fied Arrhenius equation [Eq. (26)]. The kinetic parameters of
the l-arabinose conversion to furfural are listed in Table 3. The
values of GF% for l-arabinose in all three reaction mechanisms
Table 3. Kinetic parameters for the acid-catalysed dehydration of l-arabi-
nose.
GF%[a]
Parameters
Rate constants
kA2 kA3
kA1
Ln(Ao,i) [minꢀ1 mꢀm
]
36.9ꢃ1.4 32.1ꢃ3.6
–
–
–
i
[ARA]=86.7%
[FUR]=83.7%
1
2
3
EA,i [kJmolꢀ1
]
130.2ꢃ4.0 125.8ꢃ11.7
mi
1.6ꢃ0.1
0.8ꢃ0.2
Ln(Ao,i) [minꢀ1 mꢀm
36.3ꢃ1.2 20.1ꢃ7.9[b] 30.4ꢃ4.5
129.4ꢃ3.6 88.6ꢃ25.8 100.0ꢃ14.6
i
]
[ARA]=86.7%
[FUR]=85.4%
EA,i [kJmolꢀ1
]
mi
1.3ꢃ0.1
0.5ꢃ0.5
1.4ꢃ0.5
Ln(Ao,i) [minꢀ1 mꢀm
35.5ꢃ1.4 18.6ꢃ4.5[b]
127.6ꢃ4.4 94.2ꢃ16.3
–
–
–
i
]
[ARA]=86.4%
[FUR]=76.5%
EA,i [kJmolꢀ1
]
mi
1.3ꢃ0.1
0
Dehydration of l-arabinose to furfural
[a] Goodness-of-fit%. [b] minꢀ1 mꢀ1ꢀm
.
i
Kinetic modelling
The kinetics of the l-arabinose dehydration reaction was stud-
ied following a similar protocol as used for d-xylose. Nine sets
of experiments were performed at 130, 150 and 1708C in solu-
tions of formic acid (10.8, 42.8 and 61.8 wt%). The initial l-ara-
binose concentration (~20 mm) was selected according to the
proportion of arabinose and xylose in an energy crop such as
are higher than 86%; however, only Mechanisms 1 and 2 gave
a value of GF% higher than 80% for the furfural concentration.
To select the most suitable reaction mechanism from these
two options, additional experiments were performed at differ-
ent initial concentrations of l-arabinose and furfural at 1708C
ChemSusChem 0000, 00, 0 – 0
www.chemsuschem.org
7
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
~
&
*
Figure 4. Experimental and modelled (solid lines) d-xylose conversion and furfural yield at 130 ( ), 150 ( ) and 1708C ( ) with CXYL,o =200–270 mm and differ-
ent formic acid concentrations: a) 30 wt% HCOOH, b) 49 wt% HCOOH, c) 64 wt% HCOOH
in formic acid solutions (60–62 wt%). The experimental and
predicted data for these additional experiments are shown in
Figure 5. There were no significant differences between the l-
arabinose concentrations predicted by the kinetic models and
the experimental data. That means that the presence of furfu-
ral had no significant effect on the conversion rate of l-arabi-
nose. However, differences between the experimental data and
the predicted values for furfural concentrations were evident.
In Mechanisms 2 and 3, furfural also reacted with l-arabinose
or the intermediate from l-arabinose following a second-order
reaction rate. As a result, the furfural concentrations calculated
using these reaction mechanisms were much lower than the
experimental data (Figure 5b–d). This behaviour was more evi-
dent if furfural was added at the beginning of the reaction
(Figure 5c and d).
Based on these experimental results and the statistical analy-
sis, a simple first-order reaction mechanism with only one side
reaction (i.e., Mechanism 1) describes the dehydration reaction
of l-arabinose to furfural satisfactorily and was selected for fur-
ther study.
Implications of the kinetics
The experimental data and the predicted profiles of l-arabi-
nose conversion and furfural yields at various reaction condi-
tions are presented in Figure 6. Maximum furfural yields be-
tween 57 and 62 mol% were observed at 1708C and formic
&
ChemSusChem 0000, 00, 0 – 0
www.chemsuschem.org
8
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
&
*
Figure 5. Concentrations of l-arabinose ( ) and of furfural ( ) with different initial compositions and profile predictions from Mechanisms 1 (dashed line), 2
(solid black line) and 3 (solid grey line).
acid concentrations between 42.8 and 61.8 wt%. As a result of
the low initial concentration of l-arabinose used in these ex-
periments (20 mm), the maximum furfural yields obtained from
l-arabinose were higher than those obtained from d-xylose. In
line with the experimental results using d-xylose, the highest
temperature used in this study (1708C) gave the highest furfu-
ral yields.
were reported by Lavarack et al.[17] (135.1 kJmolꢀ1), Nabarlatz
et al.[28] (125.2 kJmolꢀ1), Danon et al.[29] (121.4 kJmolꢀ1) and Car-
valheiro et al.[30] (114–117 kJmolꢀ1). The activation energies of
the auto-catalysed dehydration reaction of l-arabinose report-
ed by Gairola and Smirnova[31] (109 kJmolꢀ1) and Zheng
et al.[32] (130–135 kJmolꢀ1) were also in line with this study. The
activation energy of l-arabinose conversion to the condensa-
tion products (125.8 kJmolꢀ1) was also in line with that report-
ed by Gairola and Smirnova[31] (123 kJmolꢀ1) and Danon
et al.[29] (120.2 kJmolꢀ1).
The difference in activation energies between the furfural
formation reaction and the undesired condensation reactions
reported in this study is only 4 kJmolꢀ1. As a result, the reac-
tion temperature could not play a significant role to change
the selectivity of the dehydration reaction of l-arabinose to
furfural. However, an increase of the concentration of formic
acid would not only enhance the conversion rate of l-arabi-
nose, but more importantly would enhance the furfural yield
from l-arabinose significantly. As an example, the furfural yield
increased from 10 to 60 mol% on increasing the formic acid
concentration from 10 to 62 wt% at 1308C. This result can be
explained by comparing the reaction orders of [H+] for the for-
mation reactions of furfural and of condensation products, that
is, 1.6 and 0.8, respectively. These results indicate that the reac-
tion rate of furfural formation is affected more profoundly by
[H+], and an increase of the formic acid concentration would
enhance the yields of furfural eventually.
Comparison of the dehydration reactions of d-xylose and
l-arabinose
The combined effects of temperature and acid concentration
on the maximum furfural yields obtained from d-xylose
(200 mm) and l-arabinose (20 mm) are depicted in Figure 7.
The selectivity of the d-xylose conversion to furfural was im-
proved mainly by reaction temperature. At reaction tempera-
tures above 2358C and using formic acid concentrations be-
tween 25 and 55 wt%, the maximum furfural yields from d-
xylose were around 68 mol%. If dilute formic acid solutions (~
14 wt%) are used, reaction temperatures above 1708C are es-
sential to maintain furfural yields above 50 mol%. However,
furfural yields greater than 50 mol% can be obtained readily at
reaction temperatures above 1508C if more concentrated
formic acid solutions (>64.4 wt%) are used.
The activation energy for the dehydration reaction of l-ara-
binose reported in this study (130.2 kJmolꢀ1) agreed with the
values reported in other studies on the degradation of l-arabi-
nose catalysed by mineral acids (HCl or H2SO4). These studies
ChemSusChem 0000, 00, 0 – 0
www.chemsuschem.org
9
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
~
&
*
Figure 6. Experimental and modelled (solid lines) l-arabinose conversion and furfural yield at 130 ( ), 150 ( ) and 1708C ( ) with CARA,o =20 mm and different
formic acid concentrations: a) 10 wt% HCOOH, b) 43 wt% HCOOH, c) 62 wt% HCOOH
Figure 7. Maximum furfural yields as a function of temperature and formic acid concentration: a) d-xylose (200 mm) and b) l-arabinose (20 mm).
&
ChemSusChem 0000, 00, 0 – 0
www.chemsuschem.org
10
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
Furfural yields greater than 50 mol% can be obtained from
l-arabinose at reaction temperatures lower than 1508C if
formic acid concentrations above 30 wt% are used as the cata-
lyst. As opposed to d-xylose, a gradual increase in the reaction
temperature up to 2408C does not increase the furfural yields
from l-arabinose gradually. The highest furfural yield
(~63.6 wt%) from l-arabinose can be obtained only at formic
acid concentrations higher than 55 wt% and temperatures
above 1628C. The concentration of formic acid plays a more
important role in the selectivity of the dehydration reaction of
l-arabinose to furfural than that of d-xylose (Figure 7). Al-
though the initial concentration of l-arabinose (20 mm) was
lower than that of d-xylose (200 mm), the highest furfural
yields predicted from l-arabinose (62.1–63.6%) were lower
than the furfural yields from d-xylose (65.0–67.7%).
The estimated loss of pentoses [mol%] through condensa-
tion or resinification reactions as a function of temperature in
solutions of 13.8 and 50.6 wt% HCOOH is shown in Figure 8.
These losses were estimated by simultaneously solving
a system of differential equations that describe the concentra-
tions of pentoses and furfural as a function of time, tempera-
ture and acid concentration using initial concentrations of
xylose and arabinose of 200 and 20 mm, respectively.
loss of pentoses through either condensation or resinification
reactions was calculated as shown in Equation (28):
vSR ꢄ nSR
Loss through side reactions ¼
ꢄ 100
ð28Þ
n
C5,o
in which nSR corresponds to the amount of d-xylose or l-arabi-
nose required to form 1 mol of condensation or resinification
products, that is, 2 mol XYL (1 mol of INT-X and 1 mol of FUR)
forms 1 mol COND, 1 mol ARA forms 1 mol COND, and 1 mol
of pentose forms 1 mol of furfural, which ultimately forms
1 mol of RES, nRX refers to the number of moles of condensa-
tion or resinification products formed and nC ,o corresponds to
5
the initial number of moles of xylose or arabinose.
An increase of the reaction temperature generally inhibits
the formation of resinification products from furfural (Figure 8a
and b). A more concentrated acidic medium is beneficial to
suppress the formation of resinification products, that is, the
loss of d-xylose in the solution of 50.6 wt% HCOOH was only
between 3–5%, compared to 9–13% loss in the solution of
13.8 wt% HCOOH. In a more concentrated solution of HCOOH
(50.6 wt%), the formation of resinification products from l-ara-
binose was higher than that from d-xylose because of the
lower rate of furfural formation from l-arabinose (1.2–
168.6 molFUR molARA,oꢀ1 minꢀ1) than from d-xylose (1.8–
187.8 molFUR molXYL,oꢀ1 minꢀ1). As a result, the maximum yield of
The amounts of condensation and resinification products
presented in Figure 8 were evaluated at a definite reaction
time when the maximum furfural formation was observed. The
Figure 8. Effect of the temperature on the loss of d-xylose (200 mm) and of l-arabinose (20 mm) through condensation and resinification products with 13.8
(white bar) and 50.6 wt% HCOOH (grey bar).
ChemSusChem 0000, 00, 0 – 0
www.chemsuschem.org
11
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
furfural from l-arabinose could
only be obtained at longer reac-
tion times, which allowed the
furfural to form more resinifica-
tion products than in the case of
d-xylose. No significant differen-
ces were observed for the reac-
tions in a solution of 13.8 wt%
HCOOH.
The loss of d-xylose or l-arabi-
nose through condensation reac-
tions showed a distinct trend as
a function of temperature and
acid concentration (Figure 8c
and d). For d-xylose, an increase
in temperature led to a lower
amount of condensation prod-
Scheme 4. Distribution of isomers in the aqueous dilution of d-xylose and l-arabinose at 20 and 258C, respective-
ly.[35,36]
ucts. The formation reaction of the condensation products,
which can be also referred to as humins or humic products,
may be categorised as a polymerisation reaction. Consequent-
ly, it is expected that higher temperatures would inhibit the
formation of these byproducts by preventing bond formation
between the xylose-derived intermediate and furfural. The loss
of the d-xylose was 7–14% higher if the acid concentration
was increased from 13.8 to 50.6 wt% at temperatures between
150 and 2408C. This indicates that the condensation reaction
between the xylose intermediate and furfural was catalysed
significantly by formic acid as observed by the high reaction
orders of [H+] of the condensation reaction (mX2 =1.5). Howev-
er, l-arabinose showed the opposite trends to d-xylose. The in-
crease in temperature led to a greater loss of l-arabinose
through condensation reactions. This suggests that these con-
densation products were formed through a direct thermal deg-
radation reaction of l-arabinose rather than from the polymeri-
sation between l-arabinose and furfural, which oc-
whereas for l-arabinose the a-pyranose form is predominant
(~57%) with significant amounts of a- and b-furanoses
(~12.5%) at 258C.[35,36] The presence of a more stable tautomer
(a-pyranose) in l-arabinose may hinder the dehydration of this
molecule, and consequently l-arabinose would be more vul-
nerable to thermal degradation at temperatures over 1508C.
As a result, the selectivity towards furfural decreased as ob-
served in this study.
Dehydration of d-glucose to HMF and LA
Kinetic modelling
The calculation routine described previously was followed, and
the three sets of kinetic parameters, which fit the kinetic
models of the glucose dehydration reaction shown in
Scheme 3, are presented in Table 4 as well as the GF% for the
curred with d-xylose. As shown in the previous sec-
Table 4. Kinetic parameters for the acid-catalysed dehydration of d-glucose.
tion, the formic acid concentration had a beneficial
effect on the selectivity of the dehydration of l-ara-
binose to furfural. The loss of l-arabinose was 9–
13% lower if the acid concentration increased from
13.8 to 50.6 wt% at temperatures between 150
and 2408C.
The differences in pentoses reactivity found in
this study are in agreement with previous studies
that have compared the reactivity of pentoses in
different types of reactions, such as enzymatic con-
version and acid or thermal degradations.[33,34]
Garret and Dvorchik[33] compared the HCl-catalysed
degradation of various aldopentoses (ribose, d-
xylose, lyxose, l-arabinose) in the aqueous phase.
Arabinose was the least reactive of the pentoses
GF%[a]
Parameters
Rate constants
kG1
kG2
kG3
kG4
28.5ꢃ6.3
[GLC]=78.6% Ln(Ao,i) [minꢀ1 mꢀm
]
36.4ꢃ5.5 25.8ꢃ1.8 27.1ꢃ5.8
i
1
2
3
[HMF]=68.9% EA,i [kJmolꢀ1
[LA]=81.3% mi
]
132.2ꢃ17.2 90.7ꢃ5.6 106.0ꢃ17.4 112.6ꢃ22.1
1.6ꢃ0.4
1.4ꢃ0.1
1.0ꢃ0.4
0.7ꢃ0.3
[GLC]=77.1% Ln(Ao,i) [minꢀ1 mꢀm
36.9ꢃ3.8 26.2ꢃ1.0 15.8ꢃ1.4[b]
132.4ꢃ11.4 95.2ꢃ3.1 71.6ꢃ4.4
22.5ꢃ2.3
i
]
[HMF]=73.5% EA,i [kJmolꢀ1
[LA]=84.3% mi
]
79.3ꢃ7.3
1.4ꢃ0.3
0.7ꢃ0.1
0.4ꢃ0.1
1.3ꢃ0.2
[GLC]=77.1% Ln(Ao,i) [minꢀ1 mꢀm
34.6ꢃ3.5 22.9ꢃ1.8 20.5ꢃ10.1[b]
126.1ꢃ10.6 81.0ꢃ5.7 90.2ꢃ32.3
–
–
–
i
]
[HMF]=57.0% EA,i [kJmolꢀ1
[LA]=86.1% mi
]
1.4ꢃ0.3
1.3ꢃ0.1
1.0ꢃ0.7
[a] Goodness-of-fit%. [b] minꢀ1 mꢀ1ꢀm
.
i
they used, and the authors attributed this to the particular
steric position of the hydroxyl groups in arabinose and their
role during the dehydration reaction. Notably, the tautomeric
distributions of l-arabinose and d-xylose are remarkably differ-
ent (Scheme 4). The predominant form of d-xylose in aqueous
solutions is known to be mostly the b-pyranose form (~65%),
concentrations of d-glucose, HMF and LA. The GF% for d-glu-
cose concentrations obtained from the three reaction mecha-
nisms did not differ significantly (77–78%). Mechanism 3 gave
the poorest fit (57%) for HMF concentrations; however, it gave
the highest GF% for LA concentrations (86%). Mechanism 2,
which involves the formation of an intermediate compound,
&
ChemSusChem 0000, 00, 0 – 0
www.chemsuschem.org
12
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
fitted the experimental data of the three compounds involved
in the dehydration reaction of d-glucose satisfactorily.
Implications of the kinetics
To support the selection of this reaction mechanism and the
validation of the kinetic parameters, additional experiments
with various initial concentrations of d-glucose and HMF were
performed at temperatures between 165 and 1708C in formic
acid solutions of 34–35 and 60–61 wt%. The experimental data
obtained from these validation experiments as well as the pre-
dicted data calculated from the three sets of kinetic parame-
ters given in Table 4 are presented in Figure 9.
The concentration profiles of d-glucose, HMF and LA obtained
from the kinetic experiments along with the predicted values
obtained from the kinetic model derived from Mechanism 1
are presented in Figure 10. Low conversions of d-glucose (<
40%) were observed at a reaction temperature of 1308C, al-
though the reaction times were extended to 12 h. The HMF
yield reached a maximum as soon as a significant amount of
d-glucose was converted. Higher temperatures led to higher
yields of HMF, and the highest HMF yield of 12 mol% was ob-
tained at 1708C. An increase of the concentration of formic
acid also had a positive effect on the maximum yield of HMF.
As a result of the slow reaction rates of d-glucose at 1308C,
the yields of LA did not exceed 20 mol% after 12 h. The high-
est LA yield of 40 mol% was obtained at 1708C.
The activation energy for the dehydration reaction of d-glu-
cose to HMF obtained in this study (132 kJmolꢀ1) is lower than
that reported in previous studies. In their kinetics studies, Giri-
suta et al.[37,38] and Weingarten et al.[39] reported activation en-
ergies between 152 and 160 kJmolꢀ1 for the dehydration of d-
glucose to HMF using mineral acids (H2SO4 or HCl) as the cata-
lyst. This means that the catalytic activity of formic acid is less
sensitive to changes in the reaction temperature than that of
mineral acids. Interestingly, the activation energy for the for-
mation reaction of condensation products from d-glucose ob-
tained in this study (106 kJmolꢀ1) is much lower than the re-
ported values from previous studies (150–160 kJmolꢀ1). This
relatively lower value is beneficial to suppress the formation of
condensation products by shifting the reaction temperature to
Although the second-order reaction rate proposed in Mech-
anisms 2 and 3 can describe the kinetic data satisfactorily,
these kinetic models could not fit the concentration of HMF if
the initial concentration of d-glucose was varied (Figure 9). The
kinetic models obtained from Mechanisms 2 and 3 over- and
underestimated the HMF concentrations, respectively, if the ini-
tial d-glucose concentration was much lower (12 mm; Fig-
ure 9a) and much higher (210 mm; Figure 9b) than the initial
concentration used in the kinetic experiments (~100 mm). Con-
sequently, the predicted LA concentrations were higher (Mech-
anism 2) or lower (Mechanism 3) than the experimental data.
These two reaction mechanisms also overestimated the LA
concentration if equimolar amounts of d-glucose and HMF
were reacted at 1708C in a solution of 60 wt% HCOOH (Fig-
ure 9d). However, Mechanism 1 predicted each of the experi-
mental scenarios presented in Figure 9 accurately and was se-
lected as the most suitable reaction mechanism for further
study.
~
&
*
Figure 9. Concentrations of d-glucose ( ), HMF ( ) and LA ( ) with different initial ratios of d-glucose/HMF and profile predictions from Mechanisms 1
(dashed line), 2 (solid black line) and 3 (solid grey line).
ChemSusChem 0000, 00, 0 – 0
www.chemsuschem.org
13
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
~
&
*
Figure 10. Experimental and modelled (solid lines) d-glucose conversion and HMF and LA yields at 130 ( ), 150 ( ) and 1708C ( ) with CGLC,o =82–103 mm
and different formic acid concentrations a) 12.8, b) 34.2 and c) 60.8 wt% HCOOH.
a higher value. This analysis is also supported by the experi-
mental data, which show that the selectivity to HMF was im-
proved at the highest reaction temperature used in this study.
The activation energies obtained in this study for the conver-
sion of HMF to the desired products (LA and formic acid) and
to the undesired condensation products were in agreement
with the values reported in previous studies.
and the conversion of HMF to LA (m2 =1.4) were higher than
the reaction orders for the conversion of glucose and HMF to
the condensation products (m3 =1.0 and m4 =0.7, respective-
ly). Therefore, an increased acid concentration would favour
the selectivity of the dehydration reaction of d-glucose to-
wards valuable products, that is, HMF, LA and formic acid.
The maximum yields of HMF and LA as a function of reac-
tion temperature (130–2408C) and formic acid concentration
(14–64 wt%) are shown in Figure 11. HMF yields as high as
In this study, the reaction orders of the effective acid con-
centration for the conversion of d-glucose to HMF (m1 =1.6)
Figure 11. Maximum molar yields of a) HMF and b) LA as a function of temperature and formic acid concentration (CGLC,o =200 mm).
&
ChemSusChem 0000, 00, 0 – 0
www.chemsuschem.org
14
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
18 mol% can be obtained at reaction temperatures over
2128C and formic acid concentrations over 28 wt% (Fig-
ure 11a). However, reaction temperatures lower than 1708C
would give a maximum yield of HMF of 11 mol%, which can
only be achieved using the most concentrated solution of
formic acid (64 wt%). The HMF yields obtained in this study
were substantially higher than those obtained using mineral
acids (0.05–0.5m H2SO4 or HCl) as the catalyst, which only gave
a maximum HMF yield between 2–6 mol% at 170–1758C.[38,39]
The formation of LA was particularly promoted at reaction
temperatures between 150 and 2008C and in concentrated sol-
utions of formic acid (>56 wt%; Figure 11a). The maximum LA
yields estimated in this study ranged between 37 and
40 mol%. However, these yields of LA were lower than those
obtained from the dehydration of d-glucose catalysed by
H2SO4, which gave LA yields between 64–68 mol% at a reaction
temperature of 1408C.[40] At reaction temperatures over 2008C,
LA yields obtained by using formic acid as the catalyst could
reach 40 mol%. Conversely, LA yields obtained at these tem-
peratures by using H2SO4 with
Table 5. Initial conditions of the experiments for simultaneous dehydra-
tion by formic acid of xylose, arabinose and glucose.
Sample
Concentration [mm]
Formic acid [wt%]
xylose arabinose glucose
furfural
A1
A2
50
50
6
5
15
10
20
20
23.0
37.0
well as the predicted data obtained from the kinetic models
are shown in Figure 12. Although the initial d-xylose concen-
tration used in these experiments (50 mm) was lower than that
used in the aforementioned kinetic study (200–270 mm), the
kinetic model predicts the dehydration reaction of d-xylose to
furfural satisfactorily. This also confirms that the kinetic models
developed in this study can be applied satisfactorily in a wide
range of reaction conditions.
A previous study performed by Danon et al.[41] has shown
that the presence of d-glucose can promote the degradation
an equivalent acidity were only
30 mol%.[40] Under these circum-
stances, that is, reaction temper-
atures above 2008C and formic
acid concentrations between 25
and 64 wt%, it would be pref-
erable to use formic acid over
sulfuric acid to catalyse the de-
hydration reaction of d-glucose
to LA.
Simultaneous dehydration of
d-xylose, l-arabinose and d-
glucose in aqueous mixtures of
formic acid
To evaluate the applicability of
the kinetic models presented in
this work, two solutions that
contained d-xylose, l-arabinose,
d-glucose and furfural were re-
acted at three different tempera-
tures (150, 160, 1708C) and two
formic acid concentrations (23
and 37 wt% HCOOH). The reac-
tion solutions were prepared to
resemble the expected composi-
tion of the liquid phase obtained
from the fractionation process of
lignocellulosic materials using
formic acid.[13] The initial compo-
sitions of the solutions and the
reaction conditions of the ex-
periments presented in this sec-
tion are listed in Table 5.
The experimental pentoses
~
&
*
Figure 12. Concentrations of d-xylose ( ), l-arabinose ( ) and furfural ( ) as a function of time for experiments at
and furfural concentrations as different temperatures. Solid lines correspond to the predictions by the kinetic models.
ChemSusChem 0000, 00, 0 – 0
www.chemsuschem.org
15
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
rate of furfural if equivalent
amounts of furfural and d-glu-
cose (50 mm furfural/50 mm d-
glucose) were reacted at tem-
peratures between 160 and
2008C using a mixture of 50 mm
HCl/500 mm NaCl as the catalyst.
Based on their results at 1708C,
the furfural loss after 1 h was
only 6% in the absence of d-glu-
cose, whereas the furfural loss
increased to 22% if an equimo-
lar mixture of furfural and d-glu-
cose was used as the reactant.
They attributed this effect to the
formation of LA, which would
enhance the degradation rate of
furfural. In this study, however,
the formation and degradation
rates of furfural were not affect-
ed by the presence of d-glucose,
which was present either in
sample A1 or A2 at a molar ratio
of 20 mm furfural/10–15 mm d-
glucose. This could be explained
by the differences in the type of
catalyst and also the lower con-
centrations of furfural and d-glu-
cose in the solutions, which
were 2–5 times lower than those
used in the study reported by
Danon et al.[41]
The concentrations of d-glu-
cose, HMF and LA obtained by
the reaction of the samples A1
and A2 at temperatures between
150 and 1708C and formic acid
concentrations between 23 and
37 wt% are shown in Figure 13.
~
&
*
Figure 13. Concentrations of d-glucose ( ), HMF ( ) and LA ( ) as a function of time for experiments at different
temperatures. Solid lines correspond to the predictions by the kinetic models.
The predicted data calculated from the kinetic model fit well
with the experimental data. These results support the fact that
the presence of furfural in the reaction mixture had no signifi-
cant effects on the conversion rate of d-glucose as well as on
the selectivity to HMF and LA.
Table 6. Estimated production of furfural, HMF and LA from the hydro-
lysed carbohydrates in different biomass feedstocks.
Biomass
Miscanthus[37] Wheat straw[43] Norway spruce[44,45]
Carbohydrate composition [wt% dry basis]
If we take into account the previous observations, the pro-
duction of furfural, HMF and LA can be estimated in aqueous
streams obtained from different types of biomass in the pres-
ence of formic acid. An estimation of the formation of these
products in aqueous streams obtained from Miscanthus, wheat
straw and Norway spruce is presented in Table 6. In general,
high hydrolysis yields (>80%) can be attained during the ex-
traction of hemicellulose sugars at high temperatures with
weak acids.[42] Molar yields of furfural between 50 and
60 mol% are expected from the conversion at 1708C of feed-
stocks with high arabinoxylan content, such as Miscanthus,
wheat straw and hardwoods. However, softwoods, such as
spruce, would lead to relatively higher furfural yields
Arabinan
Xylan
Glucan
2.2
19.6
40.7
1.0
2.1
21.5
34.6
0.7
2.0
6.2–9.0
50–65
15–25
Galactomannan
Lignin
Estimated yields [kgtonbiomassꢀ1] (molar yield%)[a]
23.8
16.1
~30
Furfural
HMF[b]
LA[b]
94 (59.3)
5 (9.4)
9 (18.6)
99 (57.8)
4 (9.5)
8 (18.3)
43–56 (72.2–69.6)
15–24 (8.7–9.0)
34–50 (20.8–20.1)
[a] Only 15% of the glucan content was considered to be hydrolysed
before conversion to HMF and LA. Reaction time to attain the maximum
production of furfural with 60 wt% HCOOH and 100 gLꢀ1 of biomass at
1708C. [b] The kinetic parameters for the conversion of glucose were
used for the estimation of product formation from galactose/mannose.
&
ChemSusChem 0000, 00, 0 – 0
www.chemsuschem.org
16
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
the oil bath and immersed in a cold water bath to quench the re-
action. The effective concentration of acid as hydronium ion [H+]
was calculated from the dissociation equilibrium constant and the
activity coefficients evaluated at the reaction temperature using
the Debye–Hꢃckel equation and the temperature-dependent func-
tion of the dissociation constant of formic acid as proposed by Kim
et al.[46]
(>70 mol%) because of the lower amount of pentose carbohy-
drates present in these materials. Similarly, as a result of the
high content of hexose sugars, that is, glucose, galactose and
mannose, greater amounts of HMF and LA can be formed
from hemicellulose hydrolysates obtained from softwoods (up
to 24 kg HMF and 50 kg levulinic acid tonꢀ1 of Norway spruce).
Conclusions
Analytical measurements
This report presents a comprehensive study on the conver-
sions of the monosaccharides found in the hemicellulose frac-
tion of biomass to high-value chemicals, such as furfural, 5-hy-
droxymethylfurfural and levulinic acid. We started with a sys-
tematic investigation of the kinetics of major components in
the hemicelluloses fraction of biomass, that is, d-xylose, l-ara-
binose and d-glucose. The kinetics experiments for each sugar
were performed at temperatures between 130 and 1708C in
various formic acid solutions (10–64 wt%) that act as the reac-
tion solvent as well as the catalyst. In this study, three of the
reaction mechanisms reported most commonly for the conver-
sion of pentoses and hexose were examined thoroughly and
fitted to the experimental data. Based on the statistical analy-
ses and the kinetic model validation using additional experi-
mental data, the kinetic models for the dehydration reactions
of d-xylose, l-arabinose and d-glucose were selected. The im-
plications of these kinetic models on the selectivity of each
sugar to the desired products were discussed extensively. Fi-
nally, the kinetic models developed for each sugar were com-
bined and used to model the reaction kinetics of solutions
that resemble the liquid phase obtained from the fractionation
process using formic acid. The kinetic models developed in
this study provide detailed insights on the conversion of the
hemicellulose fraction of biomass to valuable chemicals, and
further optimisation studies can utilise these kinetic models to
design and develop efficient processes for the fractionation
and valorisation of lignocellulosic biomass.
The concentrations of d-xylose, l-arabinose, d-glucose, HMF, LA
and furfural in the samples were quantified using HPLC. The mea-
surement of sugars was conducted by using an ICS-3000 system
(Dionex Corp., Sunnyvale, CA) equipped with an AS50 autosampler
(10 mL sample-loop injection), dual pump, column oven, and inte-
grated electrochemical detector for the quantification of the carbo-
hydrates. An ion exchange column (Hi-Plex H 0.77ꢂ30.0 cm, 8 mm)
was used for the combined elution of the monosaccharides. The
column was operated at 658C by isocratic elution with deionised
water (0.65 mLminꢀ1) and a post-column flow of sodium hydroxide
(0.3m, 0.1 mLminꢀ1) for detection in the electrochemical detector.
In addition, a diode array detector (DAD-3000RS) set at 278 nm
was used for the measurement of furfural and HMF. LA was sepa-
rated in a Dionex Acclaim Organic Acid (5 mm, 0.4ꢂ25.0 cm)
column, which was operated at 308C and was eluted using
100 mm Na2SO4 at pH of 2.65 at a flow rate of 0.8 mLminꢀ1
.
Acknowledgements
The authors acknowledge financial support provided by the Irish
Competence Centre for Biorefining & Biofuels (TCBB) through an
Enterprise Ireland research grant (CC/2009/1305A).
Keywords: biomass · carbohydrates · hydrolysis · kinetics ·
reaction mechanisms
[1] A. Gandini, Green Polymerization Methods, Wiley-VCH, 2011, pp. 29–56.
[2] J.-P. Lange, E. Van der Heide, J. Van Buijtenen, R. Price, ChemSusChem
2012, 5, 150–166.
[3] R. M. West, Z. Y. Liu, M. Peter, C. A. Gꢁrtner, J. A. Dumesic, J. Mol. Catal.
A 2008, 296, 18–27.
[4] C. J. Barrett, J. N. Chheda, G. W. Huber, J. A. Dumesic, Appl. Catal. B
2006, 66, 111–118.
Experimental Section
Materials and chemicals
[5] E. Christensen, A. Williams, S. Paul, S. Burton, R. L. McCormick, Energy
Fuels 2011, 25, 5422–5428.
[6] J. Q. Bond, A. A. Upadhye, H. Olcay, G. A. Tompsett, J. Jae, R. Xing, D. M.
Alonso, D. Wang, T. Zhang, R. Kumar, A. Foster, S. M. Sen, C. T. Marave-
lias, R. Malina, S. R. H. Barrett, R. Lobo, C. E. Wyman, J. A. Dumesic, G. W.
Huber, Energy Environ. Sci. 2014, 7, 1500–1523.
d-Xylose, d-glucose, l-arabinose, formic acid (>98.0%), LA (>98),
HMF (>99.0%) and furfural (>99.0%) were of analytical grade and
purchased from Sigma–Aldrich. All reaction solutions and dilutions
for HPLC analysis were prepared with deionised water
(18.2 MWcm).
[7] M. J. Antal Jr., T. Leesomboon, W. S. Mok, G. N. Richards, Carbohydr. Res.
1991, 217, 71–85.
[8] M. R. Nimlos, X. Qian, M. Davis, M. E. Himmel, D. K. Johnson, J. Phys.
Chem. A 2006, 110, 11824–11838.
Dehydration experiments
The dehydration reactions were performed in ACE glass pressure
tubes (2.54 cm O.D.ꢂ10.2 cm) that comprised polytetrafluoroethy-
lene (PTFE) plugs and fluoroelastomer with tetrafluorolethylene ad-
ditives (FETFE) O-rings. The tube reactors can hold an operational
pressure up to 1.0 MPa. An aqueous solution of the model com-
pound (5 mL, 2–250 mm) was mixed with the corresponding
amount of formic acid and was poured into the pressure tubes.
The tubes were sealed, and each tube was placed for a defined
period of time in an oil bath set at the particular reaction tempera-
ture (130–1708C). On completion, each tube was removed from
[9] Q. Xiang, Y. Lee, R. Torget, Appl. Biochem. Biotechnol. 2004, 115, 1127–
1138.
[10] W. Hongsiri, B. Danon, W. de Jong, Ind. Eng. Chem. Res. 2014, 53, 5455–
5463.
[11] D. J. M. Hayes, Analysis of Lignocellulosic Feedstocks for Biorefineries
with a Focus on The Development of Near Infrared Spectroscopy as
a Primary Analytical Tool, PhD thesis, University of Limerick, 2011.
[12] S. Willfçr, A. Sundberg, J. Hemming, B. Holmbom, Wood Sci. Technol.
2005, 39, 245–257.
[13] K. Dussan, B. Girisuta, D. Haverty, J. J. Leahy, M. H. B. Hayes, Carbohydr.
Polym. 2014, 111, 374–384.
ChemSusChem 0000, 00, 0 – 0
www.chemsuschem.org
17
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
[14] M. J. Antal, Jr., W. S. L. Mok, G. N. Richards, Carbohydr. Res. 1990, 199,
91–109.
[32] J. Zheng, X. Du, W. Yan, Y. Li, J. Zhang, Adv. Mater. Res. 2012, 450–451,
710–714.
[33] E. R. Garrett, B. H. Dvorchik, J. Pharm. Sci. 1969, 58, 813–820.
[34] D. Laroque, C. Inisan, C. Berger, ꢇ. Vouland, L. Dufossꢄ, F. Guꢄrard, Food
Chem. 2008, 111, 1032–1042.
[35] A. H. Conner, L. Anderson, Carbohydr. Res. 1972, 25, 107–116.
[36] R. K. Schmidt, M. Karplus, J. W. Brady, J. Am. Chem. Soc. 1996, 118, 541–
546.
[15] R.-J. van Putten, J. C. van der Waal, E. de Jong, C. B. Rasrendra, H. J.
Heeres, J. G. de Vries, Chem. Rev. 2013, 113, 1499–1597.
[16] B. Girisuta, L. P. B. M. Janssen, H. J. Heeres, Green Chem. 2006, 8, 701–
709.
[17] B. P. Lavarack, G. J. Griffin, D. Rodman, Biomass Bioenergy 2002, 23,
367–380.
[37] B. Girisuta, K. Dussan, D. Haverty, J. J. Leahy, M. H. B. Hayes, Chem. Eng.
J. 2013, 217, 61–70.
[18] A. Esteghlalian, A. G. Hashimoto, J. J. Fenske, M. H. Penner, Bioresour.
Technol. 1997, 59, 129–136.
[38] B. Girisuta, L. P. B. M. Janssen, H. J. Heeres, Ind. Eng. Chem. Res. 2007, 46,
1696–1708.
[39] R. Weingarten, J. Cho, R. Xing, W. C. Conner, G. W. Huber, ChemSusChem
2012, 5, 1280–1290.
[40] K. Dussan, B. Girisuta, D. Haverty, J. J. Leahy, M. H. B. Hayes, Bioresour.
Technol. 2013, 149, 216–224.
[41] B. Danon, L. van der Aa, W. de Jong, Carbohydr. Res. 2013, 375, 145–
152.
[42] T. Zhang, R. Kumar, C. E. Wyman, Carbohydr. Polym. 2013, 92, 334–344.
[43] W. J. J. Huijgen, J. H. Reith, H. den Uil, Ind. Eng. Chem. Res. 2010, 49,
10132–10140.
[44] A. Vꢈrnai, M. Siika-aho, L. Viikari, Enzyme Microb. Technol. 2010, 46, 185–
193.
[19] Q. Jing, X. Lu, Chin. J. Chem. Eng. 2007, 15, 666–669.
[20] R. Weingarten, J. Cho, J. W. C. Conner, G. W. Huber, Green Chem. 2010,
12, 1423–1429.
[21] K. Lamminpꢁꢁ, J. Ahola, J. Tanskanen, Ind. Eng. Chem. Res. 2012, 51,
6297–6303.
[22] L. Kupiainen, J. Ahola, J. Tanskanen, Chem. Eng. Res. Des. 2011, 89,
2706–2713.
[23] J. J. Villaverde, P. Ligero, A. de Vega, J. Cleaner Prod. 2010, 18, 395–401.
[24] D. Haverty, K. Dussan, A. V. Piterina, J. J. Leahy, M. H. B. Hayes, Bioresour.
Technol. 2012, 109, 173–177.
[25] R. Xing, W. Qi, G. W. Huber, Energy Environ. Sci. 2011, 4, 2193–2205.
[26] J. E. Morinelly, J. R. Jensen, M. Browne, T. B. Co, D. R. Shonnard, Ind. Eng.
Chem. Res. 2009, 48, 9877–9884.
[45] M. Wiman, D. Dienes, M. A. T. Hansen, T. van der Meulen, G. Zacchi, G.
Lidꢄn, Bioresour. Technol. 2012, 126, 208–215.
[46] M. H. Kim, C. S. Kim, H. W. Lee, K. Kim, J. Chem. Soc. Faraday Trans.
1996, 92, 4951–4956.
[27] E. R. Ziegel, J. W. Gorman, Technometrics 1980, 22, 139–151.
[28] D. Nabarlatz, X. Farriol, D. Montanꢄ, Ind. Eng. Chem. Res. 2004, 43,
4124–4131.
[29] B. Danon, W. Hongsiri, L. van der Aa, W. de Jong, Biomass Bioenergy
2014, 66, 364–370.
[30] F. Carvalheiro, G. Garrote, J. C. Parajꢅ, H. Pereira, F. M. Gꢆrio, Biotechnol.
Prog. 2005, 21, 233–243.
Received: November 28, 2014
Revised: January 7, 2015
[31] K. Gairola, I. Smirnova, Bioresour. Technol. 2012, 123, 592–598.
Published online on && &&, 0000
&
ChemSusChem 0000, 00, 0 – 0
www.chemsuschem.org
18
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

FULL PAPERS
Hemicellulose’s next top model: The
pretreatment of lignocellulosic biomass
produces liquid streams with hemicellu-
lose sugars, which can be further con-
verted to high-value chemicals. This
study investigates the reaction kinetics
of the conversion of major hemicellu-
lose sugars of biomass using formic acid
as catalyst. Kinetic models are used to
model the conversion of sugars in solu-
tions obtained from the fractionation of
biomass.
K. Dussan, B. Girisuta,* M. Lopes,
J. J. Leahy, M. H. B. Hayes
&& – &&
Conversion of Hemicellulose Sugars
Catalyzed by Formic Acid: Kinetics of
the Dehydration of d-Xylose, l-
Arabinose, and d-Glucose
ChemSusChem 0000, 00, 0 – 0
www.chemsuschem.org
19
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ