Governing chemistry of cellulose hydrolysis in supercritical water

Article abstract of DOI:10.1002/cssc.201403385

At extremely low reaction times (0.02 s), cellulose was hydrolyzed in supercritical water (T=400°C and P=25 MPa) to obtain a sugar yield higher than 95 wt %, whereas the 5-hydroxymethylfurfural (5-HMF) yield was lower than 0.01 wt %. If the reaction time was increased to 1 s, the main product was glycolaldehyde (60 wt %). Independently of the reaction time, the yield of 5-HMF was always lower than 0.01 wt %. To evaluate the reaction mechanism of biomass hydrolysis in pressurized water, several parameters (temperature, pressure, reaction time, and reaction medium) were studied for different biomasses (cellulose, glucose, fructose, and wheat bran). It was found that the H⁺ and OH^- ion concentration in the reaction medium as a result of water dissociation is the determining factor in the selectivity. The reaction of glucose isomerization to fructose and the further dehydration to 5-HMF are highly dependent on the ion concentration. By an increase in the pOH/pH value, these reactions were minimized to allow control of 5-HMF production. Under these conditions, the retroaldol condensation pathway was enhanced, instead of the isomerization/dehydration pathway. Just add water: A reaction mechanism for cellulose hydrolysis that can explain the huge selectivity of biomass hydrolysis in supercritical water is presented. The model of the reaction mechanism has been validated by several experiments carried out in a continuous pilot plant capable at various conditions. It was found that the proton and hydroxide anion concentration in the medium due to water dissociation (represented by the ionic product of water, Kw) is the determining factor in the selectivity of the process.

Full text of DOI:10.1002/cssc.201403385

DOI: 10.1002/cssc.201403385
Full Papers
Governing Chemistry of Cellulose Hydrolysis in
Supercritical Water
[
a]
Danilo A. Cantero, M. Dolores Bermejo, and M. Josꢀ Cocero*
At extremely low reaction times (0.02 s), cellulose was hydro-
lyzed in supercritical water (T=4008C and P=25 MPa) to
obtain a sugar yield higher than 95 wt%, whereas the 5-hy-
droxymethylfurfural (5-HMF) yield was lower than 0.01 wt%. If
the reaction time was increased to 1 s, the main product was
glycolaldehyde (60 wt%). Independently of the reaction time,
the yield of 5-HMF was always lower than 0.01 wt%. To evalu-
ate the reaction mechanism of biomass hydrolysis in pressur-
ized water, several parameters (temperature, pressure, reaction
time, and reaction medium) were studied for different bio-
masses (cellulose, glucose, fructose, and wheat bran). It was
+
ꢀ
found that the H and OH ion concentration in the reaction
medium as a result of water dissociation is the determining
factor in the selectivity. The reaction of glucose isomerization
to fructose and the further dehydration to 5-HMF are highly
dependent on the ion concentration. By an increase in the
pOH/pH value, these reactions were minimized to allow con-
trol of 5-HMF production. Under these conditions, the retroal-
dol condensation pathway was enhanced, instead of the iso-
merization/dehydration pathway.
Introduction
Biomass exploitation as a raw material is growing as an alter-
native sustainable method of production of fuels and chemi-
of process represents an advantageous intensification that will
reduce the energetic and equipment requirements in the scale
up.
[
1]
cals. Cellulose is one of the main compounds of biomass and
[
2]
represents the most abundant biopolymer. An important
challenge in the processing of cellulosic biomass is to hydro-
lyze the b(1ꢀ4) glucose–glucose bond to produce a stream of
sugars with a low concentration of byproducts, by using an ef-
Cellulose depolymerization in hot pressurized water has
been done in different kinds of reactors (batch, semibatch, and
continuous) at different temperatures and pressures, with or
[10]
without catalysts. The yield of sugars after biomass hydroly-
sis is enhanced by using supercritical water reactors operated
[
3]
ficient process. These sugar streams could be further trans-
formed into valuable chemicals such as pyruvaldehyde, glyco-
in continuous mode at high temperature for short reaction
[
4]
[5]
[9b,11]
laldehyde, 5-hydroxymethylfurfural (5-HMF), organic acids,
times.
The combination of these two parameters is crucial
[
6]
or polyalcohols. Acid and enzymatic hydrolysis of cellulose
are two conventional methods that need long treatment times
for obtaining high yields of sugars. With long reaction times,
the sugars are derived; at low reaction temperatures, several
side reactions take place and produce many compounds. In
fact, it was observed that some reactions are avoided at super-
critical conditions. In particular, attention should be paid to the
formation of 5-HMF. The production of 5-HMF from cellulose in
pressurized water is highly dependent on the reaction temper-
ature. In Figure 1, several experimental results are shown for
cellulose hydrolysis in pressurized water from T=300–4008C at
different pressures and reaction times.
[7]
(
tꢁ3 h) to obtain a product with poor selectivity (<60 wt%).
The use of ionic liquids as both solvent and reaction medium
has been intensively studied owing to the possibility of dis-
solving cellulose and making it more “accessible” for the hy-
[
8]
drolysis reaction. However, this kind of process takes at least
h of hydrolysis to obtain a selectivity near to 30 wt% of re-
3
[
8a]
ducing sugars. These processing methods require long reac-
tion times (hours), which will demand big reactors during the
scale up. The use of pressurized water as the reaction medium
for the processing of cellulosic biomass in a one-step fast pro-
cess is an alternative. Total hydrolysis of cellulose can be ach-
ieved in a reaction time of 0.02 s in a supercritical water
medium and produces a stream of water-soluble sugars with
It can be observed that 5-HMF production was faster, but
the yield was lower, when the reaction temperature was in-
creased from 300 to 3508C. The reaction time was reduced
from 40 to 10 s by increasing the reaction temperature. This
behavior was expected and it follows the Arrhenius law. How-
ever, an expected behavior was detected if the reaction tem-
perature was increased over the critical point of water: the
production of 5-HMF was highly avoided. Although this behav-
[
9]
a low concentration of derived products (<2 wt%). This kind
[
a] Dr. D. A. Cantero, Dr. M. D. Bermejo, Prof. Dr. M. J. Cocero
Department of Chemical Engineering and Environmental Technology
University of Valladolid
[4a,b,9a,12]
ior was previously described in the literature,
a clear
and quantitative explanation has not been developed yet. The
different discovered behaviors can be classified into three
main groups. 1) The maximum amount of 5-HMF from cellu-
lose in pressurized water without catalyst is produced at tem-
Dr. Mergelina S/N, 47005, Valladolid (Spain)
E-mail: mjcocero@iq.uva.es
Supporting Information for this article is available on the WWW under
http://dx.doi.org/10.1002/cssc.201403385.
ChemSusChem 0000, 00, 0 – 0
1
ꢁ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
actions that water participates in as a reagent or in the forma-
[13]
tion of intermediate states. Another important property of
2
ꢀ2
water as a reaction medium is the ion product (mol kg ),
which represents how dissociated the water molecules are (the
ꢀ
ion concentration). If the molal concentration of OH ions
(
square root of the ionic product) is multiplied by the density,
the molar concentration of protons or hydroxide anions in the
medium is obtained. This concentration parameter includes
both the variations in water volume and water dissociation.
ꢀ
+
The concentration of OH ions (which is the same for H ions)
in the proximity of the critical point of water is plotted in
Figure 2.
Figure 1. 5-HMF yield from cellulose hydrolysis in pressurized water against
reaction time (t ). Experimental temperatures: red: 4008C; yellow: 3508C;
and blue: 3008C. Experimental pressures: ^: 27 MPa; &: 25/23 MPa; and ~:
3/18 MPa.
r
2
[
9a,12b,h,j,l]
peratures lower than 3008C.
An increase in tempera-
ture benefits the retroaldol condensation reactions of fruct-
[
12l]
ose.
2) The production of 5-HMF is enhanced by increasing
+
the availability of protons (H ) in the reaction medium by ad-
[
12d–g,i]
dition of acids.
3) The production of 5-HMF is enhanced
in a pressurized water medium if the pressure is increased at
[
4a,b]
a constant temperature.
The aim of this work was to study the reactions of cellulose
hydrolysis, with a focus on 5-HMF production from sugars. The
yields were analyzed from a chemical point of view of the reac-
tion pathway. Several reactions were performed to obtain ac-
curate data. Cellulose hydrolysis was tested at T=300, 325,
ꢀ
1
Figure 2. Hydroxy ion concentration (molL ) against temperature and pres-
ꢀ
[14]
sure. pOH=ꢀlog(jOH j)=pH. Water density was calculated according to
the International Association for the Properties of Water and Steam industri-
[
14a]
al formulation,
whereas the molal ionic product of water was calculated
by following the “International Formulation of Ionic Product of Water Sub-
3
50, 375, and 4008C at 25 MPa of pressure. Also, the pressure
effect was tested at T=300, 350, and 4008C between P=18–
7 MPa. Studies were also conducted to analyze glucose and
fructose hydrolysis in pressurized water between T=300–
008C at 25 MPa. Finally, the results were contrasted with the
[
14b]
stance”.
2
4
Important changes in the identity of the medium can be ob-
tained if the temperature and pressure are changed at the
same time. For example, the density of water at T=3008C and
products obtained from wheat-bran hydrolysis in supercritical
water. A reaction pathway was developed, and a novel kinetic
model was tested for understanding the behavior of glucose
and fructose reactions in supercritical water.
ꢀ
3
P=27 MPa is around 750 kgm ; this value can be decreased
ꢀ
3
to 130 kgm if the conditions are modified to T=4008C and
+
ꢀ
P=23 MPa. The H /OH concentration varies by six orders of
magnitude in the region of the critical point, which allows the
possibility of working with markedly different reaction medi-
Results and Discussion
+
ꢀ
ums. The H /OH concentration at T=3008C and P=23 MPa
is around 2ꢂ10 molL , which means that the medium has
Reaction medium and reaction pathway
ꢀ
6
ꢀ1
+
ꢀ
Supercritical water is water at temperature and pressure values
a high concentration of ions ([H ] and [OH ]) and ionic reac-
[
4a,15]
+
ꢀ
above the critical points (T =3748C and P =22.1 MPa). In the
tions are favored.
The H /OH concentration will take
c
c
ꢀ
12
ꢀ1
proximity of the critical points, the properties of water can be
highly influenced by changing pressure and temperature, so
the identity of the medium can be modified without changing
the solvent. The medium density represents the quantity of
a value of 5.5ꢂ10 molL if the temperature and pressure
are changed to 4008C and 23 MPa; this reaction medium
[16]
would favor radical reactions.
The reactions were assumed to follow the reaction pathway
shown in Scheme 1. This reaction pathway was built by follow-
ing the schemes developed in Reference [12b]. The reaction of
ꢀ
3
water per volume unit (kgm ); this is a measurement of water
concentration, an important factor to take into account in re-
&
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
To evaluate the medium
effect in the selectivity, the
model was solved in three dif-
ferent ways: 1) by considering
only the concentration of cellu-
lose and its derived products,
2) by considering also the water
concentration, and 3) by consid-
ering the concentration of cellu-
lose, its derived products, and
the proton or hydroxide anion
concentrations in the reactions
of glucose isomerization and
fructose dehydration. Therefore,
the concentration of fructose
(
nf) was calculated according to
Equation (2) or Equation (3) in
the resolution of models 2 or 3,
respectively, for which nw is the
Scheme 1. Main reaction pathways of cellulose hydrolysis in pressurized water.
ꢀ
1
glucose isomerization occurs through ring opening and keto–
enol tautomerism. These reactions take place to form transition
water concentration (molL ), n
is the concentration of
OH-H
ꢀ
+
ꢀ1
OH or H ions in the medium (molL ), n is the concentra-
g
ꢀ
+
states with OH or H ions. Also, fructose dehydration takes
tion of glucose, and k , k , and k are the kinetic constants of
gf
fh
fg
+
place to form transition states incorporating H ions (one per
the reactions indicated in Scheme 1.
[
17]
H O molecule lost). To identify these reactions in Scheme 1,
2
ꢀ
+
dnf
the symbols OH /H have been added above the reaction
arrow. The production of glycolaldehyde was enhanced at su-
percritical conditions because the hydroxide/proton concentra-
tion is highly decreased (pH=pOH=13) and so is the concen-
tration of fructose and its derived products. Although the reac-
tion of glucose isomerization is avoided at low concentrations
of hydroxide anions, a fructose yield of near to 10 wt% was
obtained at supercritical conditions.
¼
¼
k n n ꢀ k n n ꢀ k n
ð2Þ
ð3Þ
gf
g
w
fh
f
w
fg
f
dt
dnf
dt
k n n
ꢀ k n n
ꢀ kfgnf
gf
g
OHꢀH
fh
f
OHꢀH
Sugar production from cellulose hydrolysis
As shown in Scheme 1, fructose can follow two main reac-
tion pathways: fructose dehydration or retroaldol condensa-
tion. The second reaction was more beneficial than the first
one because, in this way, glyceraldehyde was obtained as the
main product from fructose. The maximum quantity of 5-HMF
was obtained by working at T=3508C and P=23 MPa (pH=
pOH=6) for a reaction time of 10 s. Under those conditions,
the yield was around 15 wt%.
The evolution of the sugar yield along with reaction time can
be seen in Figure 3 (see also Tables S.1–S.10 in the Supporting
Information). The best conditions to obtain soluble sugars (up
A reaction model was built by considering cellulose as the
starting material. The concentration of each compound shown
in Scheme 1 was calculated against the reaction time. The dif-
ference between the calculated concentrations and the experi-
mental ones was minimized to obtain the kinetic constant of
[
9a]
the reactions. Equation (1) describes the evolution of com-
ꢀ
1
pound i along with reaction time, for which n and n (molL )
i
j
are the concentrations of compounds i and j, respectively, t (s)
ꢀ
1
is the time, k (s ) is the kinetic constant of the reaction in
ji
which compound j reacts to produce compound i, and kio is
the kinetic constant of the reaction in which compound
i reacts to produce a derived product.
n
n
Figure 3. Sugar yields from cellulose hydrolysis in pressurized water against
reaction time. Experimental temperatures: red: 4008C; yellow: 3508C; and
blue: 3008C. Experimental pressures: ^: 27 MPa; &: 25/23 MPa; and ~: 23/
X
X
dni
dt
¼
k n ꢀ
k ni
io
ð1Þ
ji
j
1
1
18 MPa.
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
to six glucose units) from the hydrolysis of cellulose were ach-
ieved by working at T=4008C with extremely short reaction
times (t=0.015 s). If the reaction time was increased, the
sugars were hydrolyzed and the yield decreased, as shown in
Figure 4. At T=4008C, the yield of soluble sugars was higher
sis would have a sugar yield of between 80–98 wt% if the re-
action time was between 0.015–0.2 s.
The best combination of conditions to obtain a high yield of
glycolaldehyde can be achieved by working at T=4008C and
P=23 MPa with a reaction time of 1 s. In those conditions, the
yield of glycolaldehyde was around 60 wt%. Sometimes, the
production of 5-HMF is undesired, especially if a microorganism
[19]
postmodification process is needed. At T=4008C, the 5-HMF
production was highly avoided for all studied pressures. Inde-
pendently of the residence time (0.015 s for sugars or 1 s for
glycolaldehyde), the concentration of 5-HMF was lower than
0.1 wt%.
Reaction model evaluation: Cellulose hydrolysis at
P=25 MPa for different temperatures
The reaction pathway shown in Scheme 1 was kinetically eval-
uated by following Equations (1), (2), or (3). The results for the
three tested models are shown in Figure 4. The behavior of
the reaction models will be compared by means of the results
for the kinetic constants of fructose dehydration to 5-HMF. The
obtained kinetic constants for the reaction rates k , k , and
og
gg
kfg (as indicated in Scheme 1) obtained by the resolution of
model 1 follow the Arrhenius law. However, the kinetics of glu-
cose isomerization and fructose dehydration showed a break
point near the critical point of water. As can be observed in
Figure 4A, the model linearly predicts the kinetic constants of
fructose dehydration at subcritical temperatures. However,
near the critical point of water a break point in the kinetic was
observed, which represents a deviation from the Arrhenius law.
This phenomenon could be predicted because of the low con-
centration of 5-HMF in the products. Moreover, the concentra-
tion profiles for 5-HMF obtained at T=4008C were lower than
those found at T=350 or 3008C (these concentration profiles
can be seen in the Supporting Information). This deviation
from the Arrhenius behavior suggests that another chemical
effect is not taken into account in the kinetic evaluation. To
solve this problem, the water or ion concentrations were con-
sidered as well as the reagent concentration.
The reactions of glucose and fructose in near critical water
are usually analyzed together with the medium density to ex-
Figure 4. Kinetic constants for fructose dehydration to 5-HMF at P=25 MPa
and T=300–4008C. A) Kinetic evaluation with consideration of the biomass
concentration as the reagent concentration. B) Kinetic evaluation with con-
sideration of the biomass and water concentrations as the reagent concen-
tration. C) Kinetic evaluation with consideration of the biomass and ion con-
centrations as the reagent concentration. Error bars represent the experi-
mental and fitting errors.
[3b,4a,b,9a]
plain the selectivity of the process.
Qualitatively, the
theory of water transition states works well: if the medium
density is low, the concentration of 5-HMF will be low. To
quantify the participation of water in the reaction, the water
concentration was added as a reagent concentration in the re-
actions of glucose isomerization and fructose dehydration
(model 2). Unfortunately, the obtained kinetic constants with
consideration of the water concentration as a reagent concen-
tration did not follow the Arrhenius law at sub- or supercritical
water (Figure 4B). Thus, from a quantitative point of view, the
theory of water transition states would not explain the low
concentration of 5-HMF at supercritical conditions. Finally, the
kinetic constants of the whole reaction system followed the Ar-
rhenius law if model 3 was resolved (Figure 4C). After this eval-
uation, it can be concluded that the reactions of glucose iso-
merization and fructose dehydration are highly dependent on
than 95 wt%. Similar methods of cellulose hydrolysis in pres-
surized water were developed in the literature with sugar se-
[
11,12j,18]
lectivities of lower than 77 wt%.
The combination of the
supercritical water medium and the effective method of reac-
tion-time control presented in this work allow cellulose hydrol-
ysis a high yield of sugars. This is because, under those condi-
tions, the cellulose hydrolysis kinetics are fast enough and the
glucose hydrolysis kinetics are slow enough to allow our reac-
tor to stop the reactions after total hydrolysis and before glu-
[
9b]
cose degradation. It was observed that the cellulose hydroly-
&
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
the ion concentration in the reaction medium. This phenomen-
on was also observed for the reaction of methyl tert-butyl
[
20]
ether hydrolysis in sub- and supercritical water.
The kinetic constants for cellobiose hydrolysis (k ), oligosac-
h
charides hydrolysis (k ), glucose isomerization to fructose (k ),
og
gf
and glucose retroaldol condensation (k ) are plotted in Fig-
gg
ure 5A–D, respectively. It can be observed that the resolution
of model 3 produces a series of kinetic constants that follows
the Arrhenius law for the whole system. The dotted lines in
Figure 5 represent the uncertainty of the kinetic constants with
consideration of the experimental errors and the fitting errors.
It is important to take into account the dimension of the kinet-
ic constant. For model 3 resolution, the k , k , and k con-
h
og
gg
ꢀ
1
stants have the same units (s ). However, the k and khmf con-
gf
ꢀ
1
ꢀ1
stants have second-order units (Lmol s ). This is because
these two reactions were considered to be dependent on the
concentrations of two reagents. In addition, the obtained
values of the kgf and khmf constants were higher than the
values of the k , k , and k constants.
h
og
gg
To describe the kinetic behavior against temperature for the
different reactions, the activation energy (E ) and the pre-expo-
a
nential factor (ln k ) were calculated. These parameters are
o
shown in Table 1 for all of the kinetics fitted. It can be also ob-
Table 1. Activation energy and pre-exponential factor for glucose reac-
tions in pressurized water medium.
Kinetic constant
E
a
Error
[kJmol ]
ln k
o
Error
ꢀ
1
ꢀ1
[
kJmol
]
ꢀ
1
k
k
k
k
k
h
[s
]
90.4
4.5
6.1
34.1
7.4
17.4
28.6
127.3
28.7
0.9
1.2
6.6
1.4
5.0
ꢀ
1
og [s
gf [Lmol
]
145.5
566.3
147.4
348.1
ꢀ
1
1
ꢀ1
ꢀ1
s
]
ꢀ
1
gg [s
]
ꢀ
fh [Lmol
s
]
26.0
84.8
served in this table that the error in the Arrhenius parameters
is around 5% with consideration of the experimental and fit-
ting uncertainties. Also, as previously analyzed for the kinetic
constants of kgf and khmf, the values of the Arrhenius parame-
ters for these constants were higher than for the others.
Reaction model evaluation: Cellulose hydrolysis at different
pressures and temperatures
The evaluation of the reaction model was also tested for the
reactions of cellulose hydrolysis at different pressures and tem-
peratures. The cellulose hydrolysis was tested at T=3008C (P=
Figure 5. Kinetic constants for cellulose hydrolysis at P=25 MPa and
T=300–4008C. A) Kinetic constants of cellobiose hydrolysis. B) Kinetic con-
stants of oligosaccharide hydrolysis. C) Kinetic constants of glucose isomeri-
zation to fructose. D) Kinetic constants of glucose retroaldol condensation.
The dotted line represents the uncertainty of the kinetics behavior.
18, 23, and 27 MPa), T=3508C (P=18, 23, and 27 MPa), and
T=4008C (P=23, 25, and 27 MPa). A comparison of the three
model results can be seen in Figure 6A. Once again, the out-
comes of models 1 and 2 did not follow the Arrhenius law. The
fructose dehydration kinetics followed the Arrhenius law if
model 3 was used to solve the kinetics system. Although the
results shown in Figure 6A correspond to the P=27 MPa
series, the results of model 3 follow the Arrhenius law for all
experimental pressures, as shown in Figure 6B. It should be
taken into account that the experiments were always done in
pressurized liquid or the supercritical phase. This is because
the series at P=18 MPa was only tested at T=300 and 3508C.
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
Figure 6. A) Kinetic constants of fructose dehydration with consideration of
Figure 7. Glucose hydrolysis kinetic constants at T=350, 385, and 4008C.
P=25 MPa. A) Kinetic constants of glucose isomerization to fructose (~)
and fructose dehydration (&) with consideration of the OH concentration
ꢀ
OH concentration (~), with consideration of water concentration (Cw) (&),
ꢀ
ꢀ
and without consideration of the OH or water concentrations (*).
P=27 MPa. B) Kinetic constant of fructose dehydration with consideration of
OH concentration at P=27 (*), 25 (&), 23 (~), and 18 MPa (^). Error bars
represent the experimental and fitting errors.
as a reagent concentration. B) Kinetic constants of glucose retroaldol con-
densation (&), fructose retroaldol condensation (^), and the formation of
pyruvaldehyde from glyceraldehyde (*). Error bars represent the experi-
mental and fitting errors.
ꢀ
Reaction model evaluation: Glucose hydrolysis at P=25
MPa for different temperatures
merization and fructose dehydration took similar orders of
magnitude for the different analyzed situations.
The reaction pathway and the kinetic model developed for cel-
lulose hydrolysis were also tested by analyzing glucose reac-
tions in pressurized water. Glucose hydrolysis reactions were
tested at P=25 MPa of pressure and at temperatures around
the critical point of water (T=350, 385, and 4008C), at which
point the change in the ionic product of water is the highest.
The fitted kinetic constants of kgf and khmf are plotted against
temperature in Figure 7A. The fitted kinetic constants of kgg,
k , and k are plotted against temperature in Figure 7B. As ex-
Testing the concept with modified reaction mediums and
natural biomass
Finally, to test the developments in the kinetics mechanism,
fructose was hydrolyzed in modified reaction mediums (with
2,2,6,6-tetramethylpiperidin-1-oxyl (TEMPO) and oxalic acid). In
addition, wheat bran was hydrolyzed in supercritical water for
testing the production of 5-HMF from a natural biomass in su-
percritical water.
gp
fg
pected, the obtained kinetic constants follow the Arrhenius
law if the concentration of ions is considered as a reagent con-
centration (model 3).
The yields of the main products obtained after fructose hy-
drolysis are shown in Figure 8. The experiments were carried
out in the experimental setup explained above. However, the
reaction medium was modified by pumping TEMPO or oxalic
acid to the reactor. TEMPO is a free radical kidnapper that is
usually employed to control radical reactions in organic syn-
The reaction mechanism proposed in this work was tested
in three different ways: cellulose hydrolysis with constant pres-
sure and changing temperature, cellulose hydrolysis with
changing pressure and temperature, and glucose hydrolysis
with constant pressure and changing temperature near the
critical point of water. For the three situations, the kinetic con-
stants of the glucose hydrolysis reactions follow the Arrhenius
parameters if the ion concentration of the medium is taken
into account. In addition, the kinetic constants of glucose iso-
[21]
thesis and polymerization. Oxalic acid was used to increase
the concentration of ions in the reaction medium. As expected,
the production of 5-HMF from fructose at supercritical condi-
tions (T=4008C and P=23 MPa) was negligible, with pyruval-
dehyde as the main product after 0.9 s of reaction time. The
addition of oxalic acid to the reaction medium increased the
&
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
soluble sugars or 60 wt% of glycolaldehyde. From the view-
point of chemistry, the selectivity of the process is governed
by the ion concentration in the reaction medium. A reaction
mechanism model was built and intensively tested to demon-
strate its reliability. The reactions of glucose and the fructose
retroaldol condensation are not very demanding of ions. In
fact, these reactions are highly improved if the water mole-
cules (reaction medium and reagent) are highly associated. On
the other hand, the isomerization reactions of glucose–fructose
and the dehydration reactions of these sugars are extremely
diminished if the water molecules are associated. For first time,
the reasons why the production of 5-HMF is highly avoided at
supercritical water conditions were quantitatively explained
and demonstrated. It was achieved by adding the concentra-
tion of protons or hydroxide ions as a result of water dissocia-
tion as a reagent concentration into the kinetic modeling of
the reactions.
Figure 8. Yields of fructose hydrolysis at T=4008C, P=23 MPa, and t=0.9 s.
The reaction medium was modified with TEMPO or oxalic acid.
availability of protons in the medium, which would promote
the fructose dehydration reaction. In fact, if the reaction
medium was acidified, the production of 5-HMF in supercritical
water was enhanced to 15 wt%. The same behavior was ob-
served with TEMPO as the reaction-medium modifier. This free
radical kidnapper has an acid role as a result of the dissociation
of the OH group bonded to the nitrogen atom. Once again, an
acid medium promoted the production of 5-HMF, which corro-
borated the fact that 5-HMF production is highly dependent
on the proton availability in the medium.
The extraordinary changes in the chemical and physical
properties of supercritical water allow biomass hydrolysis with
the choice of desired products by simply selecting the correct
reaction temperature and pressure.
Experimental Section
The experiments were carried out in a continuous pilot plant that
was able to work at temperatures up to 4258C and pressures up
to 30 MPa. A schematic diagram of the process is shown in
Figure 10. Two streams continuously fed a microreactor: a cellulose
Wheat bran was also hydrolyzed in the aforementioned ex-
perimental setup. The reaction temperature was set at 4008C
with a reactor pressure of 25 MPa. The reaction time was
varied from 0.19 s to 0.69 s. Fortunately, as can be seen in
Figure 9, the yields of 5-HMF were lower than 0.05 wt%. A de-
tailed description of wheat-bran hydrolysis for sugar and lignin
[
22]
production can be found in a previous work.
Figure 10. Schematic diagram of the supercritical water hydrolysis facility.
stream and a supercritical water stream. Strict control of the reac-
tion times was achieved by a combination of three factors: 1) rapid
heating by supercritical water injection of the cellulose suspension
stream, 2) rapid cooling by sudden depressurization down to at-
mospheric pressure and Tꢂ1008C by using a micrometering valve
that was able to stand temperatures up to T=4258C, and 3) selec-
tion of a series of tubular reactors of different accurately deter-
mined volumes. The volume of the used reactors varied from 0.12–
ꢀ
1
Figure 9. Yields of wheat-bran hydrolysis at T=4008C and P=25 MPa for re-
action times of between t=0.19 and 0.69 s.
6
4.5 mL, which in combination with flows of between 1–2 gs
and taking into account the density of water under the experimen-
tal conditions, gives reaction times of 0.004–40 s. A detailed de-
scription of the experimental setup is given in the Supporting In-
formation. The reactor is fed from two streams (biomass and
water); however, no extra water is needed in the process once the
steady state is achieved. As can be seen in Figure 10, after the re-
actor, a flash chamber separator produces two streams: vapor
Conclusions
The process presented in this work shows an efficient alterna-
tive to hydrolyze cellulose selectively. The control of the reac-
tion time is the key to obtain yields higher than 95 wt% of
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
(
water) and liquid (dissolved sugars). The vapor is almost pure
[10] D. A. Cantero, M. D. Bermejo, M. J. Cocero, J. Supercrit. Fluids 2015, DOI:
1
0.1016/j.supflu.2014.1007.1003.
water that can be recycled into the system.
[
[
11] M. Sasaki, T. Adschiri, K. Arai, AIChE J. 2004, 50, 192–202.
12] a) M. Sasaki, B. Kabyemela, R. Malaluan, S. Hirose, N. Takeda, T. Adschiri,
K. Arai, J. Supercrit. Fluids 1998, 13, 261–268; b) M. Sasaki, K. Goto, K.
Tajima, T. Adschiri, K. Arai, Green Chem. 2002, 4, 285–287; c) M. Sasaki,
Z. Fang, Y. Fukushima, T. Adschiri, K. Arai, Ind. Eng. Chem. Res. 2000, 39,
Acknowledgements
The authors thank the Spanish Ministry of Economy and Compet-
itiveness for the Projects CTQ2013-44143-R, CTQ2011-27347, and
ENE2012-33613. The authors thank Repsol for technical support.
M.D.B. thanks the Spanish Ministry of Economy and Competitive-
ness for the Ramꢀn y Cajal research fellowship RYC-2013-13976
2
883–2890; d) M. J. Antal, Jr., W. S. L. Mok, G. N. Richards, Carbohydr.
Res. 1990, 199, 91–109; e) F. S. Asghari, H. Yoshida, Ind. Eng. Chem. Res.
007, 46, 7703–7710; f) F. S. Asghari, H. Yoshida, Carbohydr. Res. 2010,
45, 124–131; g) M. Bicker, D. Kaiser, L. Ott, H. Vogel, J. Supercrit. Fluids
005, 36, 118–126; h) D. Brçll, C. Kaul, A. Krꢄmer, P. Krammer, T. Richter,
M. Jung, H. Vogel, P. Zehner, Angew. Chem. Int. Ed. 1999, 38, 2998–
014; Angew. Chem. 1999, 111, 3180–3196; i) V. Choudhary, S. H. Mush-
2
3
2
3
rif, C. Ho, A. Anderko, V. Nikolakis, N. S. Marinkovic, A. I. Frenkel, S. I.
Sandler, D. G. Vlachos, J. Am. Chem. Soc. 2013, 135, 3997–4006; j) K.
Ehara, S. Saka, J. Wood Sci. 2005, 51, 148–153; k) F. Jin, Z. Zhou, T.
Moriya, H. Kishida, H. Higashijima, H. Enomoto, Environ. Sci. Technol.
2005, 39, 1893–1902; l) D. Klingler, H. Vogel, J. Supercrit. Fluids 2010,
55, 259–270; m) X. Lꢃ, S. Saka, J. Supercrit. Fluids 2012, 61, 146–156;
n) T. Sakaki, M. Shibata, T. Miki, H. Hirosue, N. Hayashi, Energy Fuels
1996, 10, 684–688.
Keywords: biomass · carbohydrates · hydrolysis · kinetics ·
water chemistry
[
1] A. J. Ragauskas, C. K. Williams, B. H. Davison, G. Britovsek, J. Cairney,
C. A. Eckert, W. J. Frederick, J. P. Hallett, D. J. Leak, C. L. Liotta, J. R. Mie-
lenz, R. Murphy, R. Templer, T. Tschaplinski, Science 2006, 311, 484–489.
2] D. Klemm, B. Heublein, H. P. Fink, A. Bohn, Angew. Chem. Int. Ed. 2005,
[
[
44, 3358–3393; Angew. Chem. 2005, 117, 3422–3458.
[13] M. Akizuki, T. Fujii, R. Hayashi, Y. Oshima, J. Biosci. Bioeng. 2014, 117,
10–18.
3] a) J. Tollefson, Nature 2008, 451, 880–883; b) K. Arai, R. L. Smith, Jr.,
T. M. Aida, J. Supercrit. Fluids 2009, 47, 628–636; c) K. Barta, P. C. Ford,
Acc. Chem. Res. 2014, 47, 1503–1512.
[14] a) W. Wagner, R. J. Cooper, A. Dittmann, J. Kijima, J. H. Kretzschmar, A.
Kruse, R. Mares, K. Oguchi, H. Sato, I. Stocker, O. Sifner, Y. Takaishi, I. Ta-
nishita, J. Trubenbach, T. Willkommen, J. Eng. Gas Turbines Power 2000,
122, 150–182; b) W. L. Marshall, E. U. Franck, J. Phys. Chem. Ref. Data
1981, 10, 295–304.
[15] a) N. Akiya, P. E. Savage, Chem. Rev. 2002, 102, 2725–2750; b) A. Kruse,
A. Gawlik, Ind. Eng. Chem. Res. 2003, 42, 267–279; c) H. Weingꢄrtner,
E. U. Franck, Angew. Chem. Int. Ed. 2005, 44, 2672–2692; Angew. Chem.
2005, 117, 2730–2752.
[
4] a) T. M. Aida, Y. Sato, M. Watanabe, K. Tajima, T. Nonaka, H. Hattori, K.
Arai, J. Supercrit. Fluids 2007, 40, 381–388; b) T. M. Aida, K. Tajima, M.
Watanabe, Y. Saito, K. Kuroda, T. Nonaka, H. Hattori, R. L. Smith, Jr., K.
Arai, J. Supercrit. Fluids 2007, 42, 110–119; c) B. M. Kabyemela, T. Ad-
schiri, R. M. Malaluan, K. Arai, Ind. Eng. Chem. Res. 1999, 38, 2888–2895;
d) B. M. Kabyemela, T. Adschiri, R. M. Malaluan, K. Arai, H. Ohzeki, Ind.
Eng. Chem. Res. 1997, 36, 5063–5067.
[
[
5] a) A. Corma, S. Iborra, A. Velty, Chem. Rev. 2007, 107, 2411–2502; b) 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] C. Promdej, Y. Matsumura, Ind. Eng. Chem. Res. 2011, 50, 8492–8497.
[17] R. S. Assary, T. Kim, J. J. Low, J. Greeley, L. A. Curtiss, Phys. Chem. Chem.
Phys. 2012, 14, 16603–16611.
[18] a) Y. Zhao, W.-J. Lu, H.-T. Wang, Chem. Eng. J. 2009, 150, 411–417; b) S.
Kumar, R. B. Gupta, Ind. Eng. Chem. Res. 2008, 47, 9321–9329.
[19] C. Schacht, C. Zetzl, G. Brunner, J. Supercrit. Fluids 2008, 46, 299–321.
[20] J. D. Taylor, J. I. Steinfeld, J. W. Tester, Ind. Eng. Chem. Res. 2001, 40, 67–
74.
6] a) A. M. Ruppert, K. Weinberg, R. Palkovits, Angew. Chem. Int. Ed. 2012,
5
1, 2564–2601; Angew. Chem. 2012, 124, 2614–2654; b) C. Luo, S.
Wang, H. Liu, Angew. Chem. Int. Ed. 2007, 46, 7636–7639; Angew. Chem.
007, 119, 7780–7783.
2
[
7] a) L. Shuai, X. Pan, Energy Environ. Sci. 2012, 5, 6889–6894; b) C. E.
Wyman, B. E. Dale, R. T. Elander, M. Holtzapple, M. R. Ladisch, Y. Y. Lee,
Bioresour. Technol. 2005, 96, 2026–2032.
[21] C. J. Hawker, Acc. Chem. Res. 1997, 30, 373–382.
[22] D. A. Cantero, C. Martinez, M. D. Bermejo, M. J. Cocero, Green Chem.
2015, 17, 610–618.
[
8] a) R. Rinaldi, R. Palkovits, F. Schꢃth, Angew. Chem. Int. Ed. 2008, 47,
8
047–8050; Angew. Chem. 2008, 120, 8167–8170; b) S. Morales-delaRo-
sa, J. M. Campos-Martin, J. L. G. Fierro, Chem. Eng. J. 2012, 181–182,
38–541.
9] a) D. A. Cantero, M. D. Bermejo, M. J. Cocero, J. Supercrit. Fluids 2013,
5, 48–57; b) D. A. Cantero, M. D. Bermejo, M. J. Cocero, Bioresour. Tech-
nol. 2013, 135, 697–703.
5
[
7
Received: December 9, 2014
Published online on && &&, 0000
&
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
Just add water: A reaction mechanism
for cellulose hydrolysis that can explain
the huge selectivity of biomass hydroly-
sis in supercritical water is presented.
The model of the reaction mechanism
has been validated by several experi-
ments carried out in a continuous pilot
plant capable at various conditions. It
was found that the proton and hydrox-
ide anion concentration in the medium
due to water dissociation (represented
by the ionic product of water, Kw) is the
determining factor in the selectivity of
the process.
D. A. Cantero, M. D. Bermejo,
M. J. Cocero*
&& – &&
Governing Chemistry of Cellulose
Hydrolysis in Supercritical Water
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