Hydrothermal upgrading of biomass to biofuel; studies on some monosaccharide model compounds

Article abstract of DOI:10.1016/j.carres.2004.04.018

During the hydrothermal upgrading of biomass, hydrolysis to glucose is an important step. To elucidate some of the reaction pathways that follow this initial hydrolysis, the hydrothermal treatment (340°C, 27.5MPa, 25-204s) of dilute (50mM) solutions of D-glucose and some other monosaccharides were studied. As a result of the increase of K_w under subcritical conditions, both acid and base catalysed reactions occur. The acid catalysed reactions are mainly dehydrations leading initially to 5-hydroxymethylfurfural. Important base catalysed reactions result in glycolaldehyde and glyceraldehyde. Further fragmentations and dehydrations lead to a variety of low molecular weight compounds such as formic acid, acetic acid, lactic acid, acrylic acid, 2-furaldehyde and 1,2,4-benzenetriol. Important pathways leading to a decrease of the O-content of the liquid reaction products start from the intermediate glyceraldehyde, which forms pyruvaldehyde, which in its turn is converted into formic acid and acetaldehyde. The latter compound can also be formed via isomerisation of glyceraldehyde into lactic acid followed by decarbonylation.

Full text of DOI:10.1016/j.carres.2004.04.018

Carbohydrate
RESEARCH
Carbohydrate Research 339 (2004) 1717–1726
Hydrothermal upgrading of biomass to biofuel; studies on some
monosaccharide model compounds
€
Zbigniew Srokol, Anne-Gaelle Bouche, Anton van Estrik, Rob C. J. Strik,
Thomas Maschmeyer and Joop A. Peters*
Laboratory for Applied Organic Chemistry and Catalysis, Delft University of Technology, Julianalaan 136,
2628 BL Delft, The Netherlands
Received 4 March 2004; received in revised form 27 April 2004; accepted 28 April 2004
Abstract—During the hydrothermal upgrading of biomass, hydrolysis to glucose is an important step. To elucidate some of the
reaction pathways that follow this initial hydrolysis, the hydrothermal treatment (340 ꢀC, 27.5 MPa, 25–204 s) of dilute (50 mM)
solutions of -glucose and some other monosaccharides were studied. As a result of the increase of Kw under subcritical conditions,
D
both acid and base catalysed reactions occur. The acid catalysed reactions are mainly dehydrations leading initially to 5-hydroxy-
methylfurfural. Important base catalysed reactions result in glycolaldehyde and glyceraldehyde. Further fragmentations and
dehydrations lead to a variety of low molecular weight compounds such as formic acid, acetic acid, lactic acid, acrylic acid, 2-
furaldehyde and 1,2,4-benzenetriol. Important pathways leading to a decrease of the O-content of the liquid reaction products start
from the intermediate glyceraldehyde, which forms pyruvaldehyde, which in its turn is converted into formic acid and acetaldehyde.
The latter compound can also be formed via isomerisation of glyceraldehyde into lactic acid followed by decarbonylation.
ꢁ 2004 Elsevier Ltd. All rights reserved.
Keywords: Sustainable energy; Renewables; Subcritical reactions; Glucose
1. Introduction
ing this process, the oxygen content of the organic
material is reduced from about 40% to between 10% and
15%. The removed oxygen ends up in CO2, H2O and
CO. The main components of biomass resources typi-
cally are lignin, cellulose, hemicellulose and minerals.
Sasaki et al. have studied the decomposition of cellulose
in water under subcritical and supercritical conditions.4
After 1.6 s at 320 ꢀC and 25 MPa, 47% conversion was
obtained yielding hydrolysis products (cellobiose, glu-
cose, etc., 44%) and decomposition products of glucose
(erythrose, 1,6-anhydroglucose, 5-hydroxymethylfur-
fural, 3%).4 Furthermore, it has been shown that cello-
biose decomposes via hydrolysis to glucose and via
pyrolysis to glycosylerythrose and glycosylglycolalde-
hyde, which are further hydrolysed into glucose, ery-
throse and glycolaldehyde.5
The anticipated depletion of fossil fuels and the concerns
about global warming ask for initiatives to search for
renewable energy sources. Particularly interesting in this
respect are forms of biomass including wood, grass,
agricultural waste and domestic waste. Several technol-
ogies have been developed to convert biomass into a
biofuel with a higher heating value, such as gasification,
fast pyrolysis1 and hydrothermal upgrading (HTU).2;3
In the latter process, the biomass is treated during 5–
20 min with water under subcritical conditions (300–
350 ꢀC, 10–18 MPa) to give a heavy organic liquid
(‘biocrude’) with a heating value of 30–35 MJ/kg. Dur-
* Corresponding author. Tel.: +31-15-2785892; fax: +31-15-2784286;
e-mail: j.a.peters@tnw.tudelft.nl
Therefore, glucose may be a good model compound
for studying the reaction paths of the HTU reaction of
cellulose in biomass. Various groups have studied
Current address: School of Chemistry, The University of Sydney,
NSW 2006, Australia.
0008-6215/$ - see front matter ꢁ 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.carres.2004.04.018

1718
Z. Srokol et al. / Carbohydrate Research 339 (2004) 1717–1726
reactions of glucose in water under HTU conditions.4–10
Usually, these reactions were performed at short reac-
tion times (0.02–10 s). Remarkably, all studies indicate
that the amount of CO or CO2 that is formed is
negligible, under the conditions applied, whereas the
reaction products identified usually can be ascribed to
fragmentations and dehydrations. Recently, a study
of hydrothermal treatment under HTU conditions
(310–410 ꢀC, 30–50 MPa, 15 min) of baby food as
a model for phytomass (lignin free) was reported.11
Large amounts of CO2 were evolved under these
conditions. A gap exists between the studies on glucose
at very short residence times and the latter study,
which is closer to the conditions of the actual HTU
process.
Therefore, we report here the results of a systematic
study on the hydrothermal reactions of diluted solutions
(50 mM) of glucose, fructose, mannose and galactose for
longer reaction times (25–200 s). To obtain more insight
into the reaction paths, the C3–C5 sugars glyceralde-
hyde, erythrose and arabinose, glycolaldehyde and
the identified initial reaction products of these com-
pounds were treated under similar conditions. For
comparison, sorbitol was included as substrate. The
results are discussed in relation to the HTU process of
biomass.
2.1.
D
D
-Glucose (1),
-galactose (4) and
D
-mannose (2),
D-sorbitol
D-fructose (3),
Hydrothermal treatment at 340 ꢀC and 27.5 MPa of
50 mM aqueous solutions of these compounds resulted
in brown solutions with pH values between 2.5 and 2.8
and small amounts of tar. The amount of tar increased
with the residence time (s). No gases were observed
during these reactions. The liquid samples were complex
mixtures of low molecular weight compounds, most of
1
which could be identified by HPLC and H NMR. The
14 identified compounds are compiled in Table 1. These
compounds have also been detected in studies using
shorter residence times, reported in the literature.4–9 The
major initial compounds are glycolaldehyde (10) and
5-hydroxymethylfurfural (HMF, 7). These compounds
Table 1. Compounds that have been identified after HTU reactions of
monosaccharides
Name
Formula
Concentration
(mmol/L)a
5-Hydroxymethylfurfural (7)
2-Furaldehyde (9)
Glycolaldehyde (10)
Dihydroxyacetone (12)
Glyceraldehyde (13)
1,2,4-Benzenetriol (18)
Pyruvaldehyde (20)
Lactic acid (21)
C6H6O3
C5H4O2
C2H4O2
C3H6O3
C3H6O3
C6H6O3
C3H4O2
C3H6O3
C3H4O2
C2H4O
CH2O2
8.0
4.6
15.3
0.5
<0.1
2.5
0.6
1.6
2. Results and discussion
Acrylic acid (22)
<0.1
1.1
Acetaldehyde (23)
Formic acid (24)
Acetic acid (26)
The hydrothermal reactions of the model compounds
were performed with 50 mM solutions in water using a
continuous flow-type reactor (see Fig. 1). This allowed
short residence times (25–204 s) and rapid heating and
cooling of the reaction mixture. The reaction tempera-
tures were measured by a thermocouple located inside
the reactor.
4.3
1.2
C2H4O2
C2H4O3
C3H6O
Glycolic acid (28)
Acetone (30)
3.1
0.5
Total amount of identified carbon
Total organic carbon
141.2 mmol C/L
234 mmol C/L
aIn reaction mixture of glucose (50 mM, 340 ꢀC, 27.5 MPa, 120 s).
Figure 1. Experimental set-up for hydrothermal reactions.

Z. Srokol et al. / Carbohydrate Research 339 (2004) 1717–1726
1719
Figure 2. Formation of 5-hydroxymethylfurfural (m) and glycolaldehyde (d) during the hydrothermal reactions of some monosaccharides (50 mM,
340 ꢀC, 27.5 MPa).
are clearly intermediates, as their concentrations initially
increase until they reach a maximum at s ¼ 30–100 s,
after which they decrease (see Fig. 2). The concentration
of all other compounds initially steadily increased to an
almost constant value at a s-value of about 100 s.
Qualitatively, no differences existed in the products
formed from the sugars investigated. However, the
amounts of the various compounds and their rates of
formation depend strongly on the nature of the starting
sugar.
and fructose have been observed in the reaction mixtures
obtained upon hydrothermal treatment of glucose
using very short residence times.9 Possibly, successive
OH
OH
O
HO
HO
OH
D-mannose
2
The reaction mixtures obtained from these sugars
contain compounds that are characteristic for acid deg-
radation of sugars next to others that are characteristic
for basic degradation. For example, HMF (7) is a prod-
uct that typically is formed upon acid degradation of
sugars,12 whereas glycolaldehyde (10) has been observed
as an important product in alkaline degradation.13–17 The
occurrence of both acidic and base catalysed reactions
can be ascribed to the increase of the value of the ion
product of water, Kw, near the critical point.18;19 Under
those conditions Kw is about three orders of magnitude
higher than under ambient conditions.
HC OH
OH
OH
HOH2C
C
OH
O
OH
O
HO CH
HO
HO
CH2OH
HC OH
HC OH
CH2OH
5
OH
OH
HO
D-glucose
D-fructose
3
1
H2C OH
OH
OH
C
C
OH
O
HO
The similarity of the hydrothermal reaction products
D-glucose (1), D-mannose (2), D-fructose (3) and
HO
HC OH
HC OH
CH2OH
OH
of
OH
D-galactose
D-galactose (4) may be explained by the base cataly-
sed Lobry de Bruyn–Alberda van Ekenstein rearrange-
ment, which results in interconversion of glucose
(1), mannose (2) and fructose (3) via the 1,2-enediol an-
ion (5, see Scheme 1).20 It should be noted that mannose
4
6
Scheme 1. Rearrangements of monosaccharides via enediol during the
hydrothermal reaction.

1720
Z. Srokol et al. / Carbohydrate Research 339 (2004) 1717–1726
rearrangements via other enediol anions (e.g., 6) result in
interconversion of all C6-monosaccharides.
Although the reaction products of the four mono-
saccharides 1–4 are qualitatively similar, a remarkable
difference exists between the product distribution from
fructose (3) and those of the other sugars. For fructose
(3), HMF (7) is the major reaction product and glycol-
aldehyde (10) is much less abundant than with the other
sugars. Of the four sugars studied, fructose (3) has the
energetically most favourable furanose form. Under
ambient conditions, 28% of this anomeric form occurs,
whereas glucose (1), mannose (2) and galactose (4) occur
almost exclusively in pyranose forms. Furthermore, the
epimerisation rate of glucose to fructose is much faster
than the reverse reaction.8 This supports the formation
of HMF (7) via successive acid catalysed dehydrations
of fructofuranose (3) as proposed by Antal et al. (see
Scheme 2).12
Glycolaldehyde (10) can be formed from glucose in
the open form via a retro-aldol reaction (see below,
Scheme 4). Formation of large amounts of this com-
pound is remarkable, since Kabyemela et al. have
observed glyceraldehyde (13) as an important reaction
product after extremely short reaction times (s ¼ 0:02–
2 s). Under the conditions that we applied, glyceralde-
hyde was not observed (13); only minor amounts of its
isomerisation product dihydroxyacetone (12, 6 1 mM)
were detected. Both the formation of glyceraldehyde
(13) and its consecutive reactions are very fast,8;21 and,
therefore, possibly only the secondary reaction products
are observed at the relative long residence times that we
applied (see also below).
Figure 3. The total amount of nonexchangeable protons of the organic
substrate during the hydrothermal reaction of glucose.
given as Cn(H2O)m (see Table 1). Only some minor
compounds had a different H/O ratio (acetaldehyde (23),
formic acid (24), glycolic acid (28) and acetone (30)).
This supports the suggestion that decarboxylation and
decarbonylation are relatively unimportant pathways
under the conditions applied.
The exchange of the OH and COOH protons and the
water protons is rapid on the 1H NMR time scale.
Consequently only an averaged signal can be observed.
For the reaction of glucose, the total integral of the
other (nonexchangeable) protons indicated that the
amount of this type of protons decreased from 350 to
about 267 milliatom/L during the first 120 s of the
reaction (see Fig. 3). Similar behaviour was observed for
the other sugars. This indicates that about 24 mol %
dehydration occurs under these conditions, which cor-
responds with a decrease of the amount of oxygen with
the same percentage.
During the hydrothermal reactions of the monosac-
charides only minor amounts of gases were formed; the
amount was negligible for
D
-glucose (1). It should be
The total organic carbon number (TOC) of the sam-
ples at the longest residence times showed that in the
liquid samples between 60% and 96% of the total
amount of carbon was recovered. The losses can be
ascribed to tar deposited on the wall of the reactor.
Upon increase of the concentration of the substrate, the
amount of tar increased, which frequently resulted in
blocks of the reactor.
noted that the solubility of CO2 in water is about
33 mmol/L. However, a 13C NMR spectrum of a sample
taken at s ¼ 400 s showed that the amount of bicar-
bonate in the sample is negligible. The molecular
formulae of most of the identified compounds can be
OH
OH
OH
Surprisingly,
D-sorbitol remained unreacted when
O
O
HO
it was subjected to similar hydrothermal condi-
tions (50 mM solution, 240 s at 340 ꢀC and 27.5 MPa).
This indicates that the hemi-acetal function of the
sugars is essential for their conversion. Most likely,
the enediol (and/or its anion) is an important intermediate
for the basic catalysed pathways, whereas cyclic struc-
tures are important in the acid catalysed reactions.
OH
CH2OH
CH2OH
OH
OH
3
- 3 H2O
O
HOH2C
CHO
2.2.
D-Arabinose (8)
7
When the C5 monosaccharide
subjected to the hydrothermal treatment, mainly glycol-
aldehyde (10) and 2-furaldehyde (9) were obtained.
D
-arabinose (8) was
Scheme 2. Formation of furan derivatives during the hydrothermal
D-fructose is shown
reactions of monosaccharides. Here the reaction of
as an example.

Z. Srokol et al. / Carbohydrate Research 339 (2004) 1717–1726
1721
H
O
O
HO OH
HO
H
O
O
O
O
H2O
HO
OH
7
H
O
OH
OH
H
O
HO
HO
OH
OH
-H2O
HO
HO
OH
17
18
Scheme 3. Pathway proposed for the rearrangement of HMF (7) to
1,2,4-benzenetriol (18).22
be formed via several pathways. For example, ketose 14
(see Scheme 4), which can fragment via a retro-aldol
reaction into formaldehyde (15) and a C5-monosacchar-
ide (16).
Alternatively, the initially formed 1,2-enediol may be
converted into the corresponding 1,2-diketone by an a-
elimination. Subsequent a-dicarbonyl cleavage results in
Figure 4. Distribution of the most important species during the
hydrothermal treatment of
-arabinose (50 mmol/L, 340 ꢀC,
27.5 MPa). m 5-hydroxymethylfurfural, j 1,2,4-benzenetriol, d gly-
colaldehyde, . acetone, ꢀ glycolic acid, + formic acid, · 2-furalde-
hyde, ꢀ acetic acid, - lactic acid.
D
Smaller amounts of glycolic acid (28), lactic acid (21),
acetic acid (26) and acetaldehyde (23) were identified in
the reaction mixture (see Fig. 4). Apparently, similar
reaction paths occur as with the C6 monosaccharides.
Once again, initially dehydration and retro-aldol con-
densation seem to be important.
H
O
C
H
O
C
H
O
HC OH
HO CH
H
O
(ii)
(ii)
C
HC OH
HC OH
C
2
+
CH2OH
CH2OH
HC OH
HC OH
10
10
CH2OH
2.3. Reaction pathways via 5-hydroxymethylfurfural (7)
11
CH2OH
1
To untangle the reaction pathways of the reaction of
-glucose (1), we also performed hydrothermal reac-
D
(i)
tions on compounds that were identified as the initial
reaction products of this sugar. HMF (7) showed some
decomposition; at s ¼ 400 s, the concentration of HMF
(7) in the reaction mixture was decreased from 55 to
28 mM. The 1H NMR spectrum showed the presence of
several minor reaction products of which 1,2,4-benze-
netriol (18) was by far the most abundant (1.3 mM at
s ¼ 400 s). This compound was identified before by
Luijkx et al.7 It has been suggested that this compound
is formed from HMF (7) by hydrolysis of the furan ring
followed by a rearrangement to a hexatriene (17), elec-
trolytic rearrangement and dehydration (see Scheme
3).22 1,2,4-Benzenetriol (18) is stable under HTU con-
ditions, so it is one of the end products of this reaction.
However, aqueous samples of this compound are not
stable in air. They slowly convert into a dimer via an
oxidative coupling.
CH2OH
H
O
C
O
C
CH2OH
(ii)
HO CH
HC OH
+
C
O
HC OH
HC OH
CH2OH
2
CH2OH
CH2OH
13
12
(i)
CHO
CHOH
(i)
HC OH
HC OH
CH2OH
CH2OH
(i)
C
O
(ii)
(ii)
CH2OH
(ii)
H
H2C
O
+
HC OH
HC OH
HC OH
15
CH2OH
C
O
O
CH2OH
16
C
O
C
+
HC OH
HC OH
CH2OH
CH2OH
CH2OH
H
O
CH2OH
HC OH
13
10
C
+
No 2-furaldehyde (9) could be detected in the reaction
mixtures from the hydrothermal reaction of HMF. This
implies that the 2-furaldehyde (9) that has been observed
in the reaction mixture from the C6 monosaccharides
(see Table 1) has been formed via another pathway,
most likely through C5 ketoses as already proposed
previously by Kallury et al.10 The latter compounds may
CH2OH
14
C
O
10
CH2OH
Scheme 4. Fragmentation of monosaccharides via (i) Lobry de Bruyn–
Alberda van Ekenstein rearrangements and (ii) retro-aldol condensa-
tions. For convenience the saccharides are represented in their linear
forms.

1722
Z. Srokol et al. / Carbohydrate Research 339 (2004) 1717–1726
formic acid (24) and the C5-monosaccharide. The C5-
monosaccharides can then dehydrate to 2-furaldehyde
(9) as has been demonstrated in the hydrothermal
treatment of
D-arabinose (8, see above). Most likely,
first a Lobry de Bruyn–Alberda van Ekenstein rear-
rangement to ribulose takes place. Then, following
the conclusions of the work of Antal et al.,12 three
consecutive dehydration steps may lead to 2-furaldehyde
(9).
Quantitatively, the compounds identified with NMR
and HPLC cannot account for the decrease in the con-
centration of HMF (7) observed during the reaction. We
assume that tar deposited on the wall of the reactor tube
is responsible for this.
Figure 5. Expanded 1H NMR spectrum of the reaction product of the
hydrothermal treatment of glycolaldehyde (13), showing the presence
of a complex mixture of polyhydroxy compounds.
2.4. Reaction pathways via glycolaldehyde (10) and
-erythrose (11)
D
An important pathway of the hydrothermal reaction of
-glucose and the other C6 sugars starts with a retro-
aldol condensation to form glycolaldehyde (10, see
s ¼ 400 s during the hydrothermal treatment of glucose.
This compound can be formed from glyceraldehyde (13)
via another Lobry de Bruyn–Alberda van Ekenstein
rearrangement.
When glyceraldehyde (13) was applied as the feed-
stock for the hydrothermal reaction, a rapid conversion
into a mixture of compounds was observed that con-
tained most minor compounds observed in the reaction
D
Scheme 4). The second product should be a C4 sugar,
such as D-erythrose (11), but these sugars could not be
1
identified by H NMR because of the presence of many
overlapping signals in the region of the spectrum where
the resonances of these compounds should be expected.
Also in the HPLC chromatogram no C4 sugars could be
identified. To get some insight in possible reactions
starting from this type of sugars, we performed an
of
D-glucose (1): dihydroxyacetone (12), pyruvaldehyde
(20), lactic acid (21), acetaldehyde (23), formic acid (24),
acrylic acid (22), acetic acid (26), glycolic acid (28) and
acetone (30). After 200 s, lactic acid (21) was the major
component (25 mM, see Fig. 6). The concentrations of
pyruvaldehyde (20) and acetone (30) have maxima after
about 50 s, suggesting that these compounds are inter-
mediates. Pyruvaldehyde (20) can be formed from
analogous reaction with
D-erythrose (11). The main
reaction product was glycolaldehyde, which can be
formed via a retro-aldol reaction.
It should be noted that glycolaldehyde (10) cannot be
formed directly from a retro-aldol condensation from D-
fructose (3). This compound should first isomerise to
glucose to enable this reaction. This is reflected in the
low initial rate of formation of glycolaldehyde from
fructose.
D-
Glycolaldehyde (10), when subjected to the same
reaction conditions, converted with small yield into a
complex mixture of compounds with NMR resonances
at chemical shifts (3–5 ppm) that are characteristic for
polyhydroxy compounds (see Fig. 5). Most likely these
products are the result of condensation reactions.
2.5. Reaction pathways via glyceraldehyde (13)
Another initial product of the hydrothermal reactions of
monosaccharides is, once again as the result of a retro-
aldol condensation reaction, glyceraldehyde (13, see
Scheme 4). This reaction can only occur in 2-ketoses,
such as fructose (3). Therefore, starting from
D-glucose
(1) a Lobry de Bruyn–Alberda van Ekenstein rear-
rangement should precede this route. Only minor
amounts of glyceraldehyde (13) were detected, but up to
1 mM of dihydroxyacetone (12) was observed at
Figure 6. Distribution of the major reaction products during the
hydrothermal treatment of glyceraldehyde (13). d glycolaldehyde, .
acetone, ꢀ glycolic acid, - lactic acid, s pyruvaldehyde.

Z. Srokol et al. / Carbohydrate Research 339 (2004) 1717–1726
1723
H
O
H
O
H
O
(10) are relatively important under basic conditions.
Major products were glycolic acid (28) and acetone (30),
which were observed in all other reactions only in minor
amounts. Many possible pathways can be envisaged for
the formation of glycolic acid (28) and acetone (30), for
example, starting from the 2,3-enediol 6. An a-elimina-
tion followed by enol-keto rearrangement gives diketone
27, which may give glycolic acid (28) and compound 29
after an a-dicarbonyl cleavage (see Scheme 6). The latter
could be converted into acetone (30) and formic
acid (24) via dehydration, enol-keto rearrangement
followed by a base catalysed fragmentation (see Scheme
6).
C
C
C
C
C
-H2O
HC OH
OH
O
CH2OH
CH2
CH3
13
19
20
OH-
OH-
OH-
H
O
O-
O
OH
O
O-
H
OH
O-
H
OH
O-
C
C
C
C
C
C
C
C
OH
C
O
HC OH
CH3
O
O-
CH3
CH3
CH3
CH3
21
2H+
OH
C
H
O
CH2O
H+, -H2O
25
C
OH
-H2O,-CO
In the presence of added HCl (6 mM), initially HMF
(7) is formed (see Fig. 7C). However, it rapidly decom-
poses to unidentified products. Most likely tars are being
formed. After the reaction a large amount of deposited
tar was observed in the reactor. The base catalysed
reactions are not fully suppressed, glycolaldehyde (10) is
still present in considerable amounts. The most abun-
dant products of the reaction in the presence of HCl are
acetaldehyde (23) and formic acid (24), which may be
explained by reaction pathways starting from initially
formed glyceraldehyde (13, see Scheme 5).
CH3
23
+
HC OH
CH3
+
COO-
CH
CH3COO-
HCOO-
26
CH2
22
24
Scheme 5. Possible pathways of the reactions during the hydrothermal
treatment of glyceraldehyde (13).
glyceraldehyde (13) by dehydration to compound 19 (see
Scheme 5), followed by a keto-enol rearrangement.
Pyruvaldehyde (20) in turn can be converted into the
isomeric lactic acid (21) by means of a benzilic rear-
rangement. An a-dicarbonyl cleavage can explain the
formation of formic acid (24) and acetaldehyde (23) and
of acetic acid (26). The latter compound should be
accompanied by formaldehyde (25), which due to its
high reactivity probably has escaped observation. It may
be concluded that the pathway glyceraldehyde
(13) fi pyruvaldehyde (20) fi acetaldehyde (23) + formic
acid (24), formaldehyde (25) + acetic acid (26) is an
important route to compounds that have a decreased O-
content with respect to glucose.
Experimentally, starting from lactic acid as feedstock
(21), next to acrylic acid (22), once again acetaldehyde
(23) was observed. Both reactions are acid catalysed.
Dehydration results in acrylic acid (22), whereas an acid
catalysed decarbonylation may afford acetaldehyde (see
Scheme 5).
3. Conclusions
During the initial stages of the hydrothermal treatment
(340 ꢀC, 27.5 MPa) of a low concentration of C6
monosaccharides (50 mM) both acid and base catalysed
reactions play a role. The acid catalysed reactions are
mainly dehydrations with HMF as a major intermedi-
ate. Several base catalysed reactions lead to fragmenta-
tions via, for example, retro-aldol condensation and
beta elimination. Decarbonylation and decarboxylation
reactions play a minor role at the high dilution of the
feedstocks used in this research. Some highly reactive
compounds, including glycolaldehyde (10) and glycer-
aldehyde (13) are key intermediates in the pathways that
were observed in this investigation. It may be expected
that, at higher concentrations, condensation reactions
may occur similar to those observed for the alkaline
degradation of sugars at lower temperature.12–15 The
occurrence of condensation reactions was already ob-
served during the reaction of glycolaldehyde (10), which
resulted in significant amounts of condensation prod-
ucts. Therefore, at the higher concentrations that are
applied for the HTU reaction of biomass it may be ex-
pected that fragments formed subsequently enter in
condensation reactions with other fragments to afford a
complex mixture of compounds. In this dynamic system
irreversible reactions similar to those observed in this
study may occur, including the formation of 1,2,4-benz-
enetriol (18). Reactions similar to that of the pathway
2.6. Effects of added acid and base
Since both acid and base catalysed reactions were ob-
served during the hydrothermal reactions of D-glucose,
we decided to investigate the effect of added acid and
base. In the presence of 6 mM NaOH, the acid catalysed
reaction paths via HMF (7) and furfural (9) appeared to
be completely suppressed (see Fig. 7B), neither HMF
(7), 2-furaldehyde (9) nor 1,2,4-benzenetriol (18) were
observed. The amounts of lactic acid (21), acetic acid
(26) and acetaldehyde (23) were considerably higher,
indicating that the reaction paths via glyceraldehyde

1724
Z. Srokol et al. / Carbohydrate Research 339 (2004) 1717–1726
Figure 7. Distribution of the most important products during the hydrothermal reaction of
D
-glucose (50 mM, 340 ꢀC, 27.5 MPa), (A) without
addition of acid or base, (B) in the presence of 6 mM NaOH, (C) in the presence of 6 mM HCL. m 5-hydroxymethylfurfural, ꢁ 1,2,4-benzenetriol, d
glycolaldehyde, . acetone, ꢀ glycolic acid, + formic acid, · 2-furaldehyde, ꢀ acetic acid, - lactic acid.
CH2OH
CH2OH
CH2OH
CH2OH
CH2OH
acid (26) may ultimately lead to the desired reduction of
the O-content of the liquid reaction products. These
pathways seem to be favoured under more acidic con-
ditions, whereas under more basic conditions pathways
similar to those observed for basic degradation of sugars
seem to be dominant. Then, routes via diketones to
glycolic acid (28) and acetone (30) seem to be of
special interest with regard to the reduction of the
O-content during the HTU reaction. Furthermore, it
may be expected that condensation reactions of the ini-
tially formed products may afford new compounds and
maybe also new pathways may become operative.
This might include pathways that result in formation of
CO2.
Another important conclusion of this research is that
the HTU reaction mixture of biomass may contain con-
siderable amounts of highly reactive aldehydes, which
will have a bad influence on the stability of the
biocrude. Further upgrading, for example by hydro-
deoxygenation may be required to obtain a more stable
fuel.
C
C
OH
OH
C
C
O-
C
O
C
C
O
O
COO-
28
OH-
OH-
OH
C
OH
+
CH
HC OH
HC OH
HC OH
HC OH
CH2OH
CH2
H
O
C
HC OH
CH2OH
HC OH
CH2
CH2OH
CH2OH
6
27
HC OH
CH2
29
OH
H
O-
H
C
O
H
O
H
O
HCOOH
25
+
CH3
C
C
C
H+
OH-
CH2 H+
CH2
CH2
CH2
, -H2O
C
O
C
OH
HC OH
C
O
C
O
CH3
CH2
CH2
29
CH3
CH3
30
Scheme 6. Possible pathways during the hydrothermal treatment of
glucose in the presence of base.
glucose (1) fi glyceraldehyde (13) fi pyruvaldehyde (20)
fiacetaldehyde (23)+CO/formaldehyde (25)+acetic

Z. Srokol et al. / Carbohydrate Research 339 (2004) 1717–1726
1725
1
4. Experimental
HPLC and H NMR. Assignments of peaks were made
by comparison of chromatograms and spectra with
those of authentic samples. The quantitative results of
the HPLC and 1H NMR analyses agreed with each
other within the experimental error (ca. 5%). Some of
the samples were also analysed for total organic carbon
(TOC).
The HPLC analyses were carried out with a Milli-
pore-Waters 590 pump and Phenomenex Rezex organic
acid column at 60 ꢀC. The eluent was 0.01 M aq triflu-
oroacetic acid and the flow rate 0.5 mL/min. A refractive
index (RI) (Shodex model SE-51) detector and an
ultraviolet (UV) (Shimadzu spectrometric detector
model SPD-6A) at 215 nm were used. The peak areas
were determined using an integrator (Kipp & Zonen).
HPLC quantitation was achieved using THF as an
internal standard.
1H NMR spectra were measured on a Varian Unity
INOVA-300 spectrometer at 300 MHz). A weighed
amount of tert-butanol in D2O was added to the sam-
ples to lock and as an internal standard. The chemical
shifts are reported with respect to the CH3 signal of tert-
butanol, which was set at 1.20 ppm. The water reso-
nance was suppressed using presaturation with the
transmitter. The peak areas were determinded by
deconvolution of the spectrum with lorentzian peaks.
Total organic carbon analyses were performed using a
Shimadzu T.O.C. 5050A instrument.
4.1. Chemicals
All chemicals used were obtained from Aldrich. A 50 mM
solution of glycolaldehyde in water was heated for 10 min
at 60 ꢀC to hydrolyse any oligomers present just prior to
the hydrothermal reaction. All other substrates were
used without further purification. The water used was
demineralised and had a conductivity of 18.2 mX cm.
4.2. Apparatus and method
The home-built apparatus used for the HTU reactions is
schematically depicted in Figure 1. A continuous flow-
type reactor (made of stainless steel 316, length 35 mm,
internal diameter 4 mm) allowed the short residence
times required to study the initial products. A solid
cylinder of aluminium with an outer diameter of 25 mm
surrounded the reactor. The reactor and the aluminium
housing were heated in an oven. The temperature was
measured by a chromel–alumel thermocouple located
inside the reactor. The inlet and outlet of the reactor
were cooled by a cooling water jackets, which decreased
the temperature to less than 30 ꢀC. The feedstock had a
concentration of 50 mmol/L of the component under
study, unless stated otherwise, and was fed to the reactor
using an HPLC pump at a rate of 1.0–9.9 mL/min
depending on the residence time required. The flow rate
was frequently tested by measuring the volume of the
water pumped through the reactor as a function of
the setting of the flow rate of the HPLC pump. From the
flow rates of the feed solutions, the residence time of the
solutions in the reactor were calculated. The pressure in
the reactor was controlled by a Tradinco dome-loaded
backpressure regulator with a Teflon membrane to
which a hydraulic reference oil pressure was applied
with a Barnet dead-weight tester (maximum pressure
70 MPa). The pressure regulator is designed to allow
smooth pressure release for liquids as well as liquid/gas
mixtures. No pressure fluctuations could be observed as
long as the membrane was clean. All experiments were
carried out with a reactor temperature of 340 ꢀC and a
pressure of 27.5 MPa.
Acknowledgements
This research was carried out under a grant from the
Netherlands Organisation for Scientific research (NWO)
with financial contributions of Shell Global Solutions,
the Dutch Ministry of Economics Affairs (Senter) and
Novem. Thanks are due to E. Wurtz and R. Sloter for
assistance with the construction of the experimental set-
up and to Dr. F. Goudriaan, Dr. L. Petrus and Dr. G.
C. A. Luijkx for helpful discussions. We are also
indebted to K. Djanashvili and J. Knoll for performing
the NMR and T.O.C. analysis.
Reactor shut-down was performed by flushing with
water during cooling-down, and if tar had been depos-
ited in the reactor, it was dissolved by pumping acetone
through the reactor after cooling down to room tem-
perature. Tar is defined here as any black material that is
insoluble in water but soluble in acetone.
References
1. Bridgwater, A. V. Chem. Eng. J. 2003, 91, 87.
2. Goudriaan, F.; Peferoen, D. G. R. Chem. Eng. Sci. 1990,
45, 2729–2734.
3. Goudriaan, F.; van de Beld, B.; Boerefijn, F. R.; Bos,
G. M.; Naber, J. E.; van de Wal, S.; Zeevalkink, J. A.
In Proceedings Progress in Thermochemical Biomass Con-
version; Bridgwater, A. V., Ed.; Blackwell Science Ltd:
Oxford, UK, 2001.
4.3. Analyses
4. Sasaki, M.; Fang, Z.; Fukushima, Y.; Adschiri, T.; Arai,
K. Ind. Eng. Chem. Res. 2000, 39, 2883–2890.
5. Kabyemela, B. M.; Takigawa, M.; Adschiri, T.; Malaluan,
R. M.; Arai, K. Ind. Eng. Chem. Res. 1988, 37, 357–361.
At each residence time, three samples of 10 mL were
taken. Any solid particles formed were centrifuged or
filtered off and the clear solutions were analysed by

1726
Z. Srokol et al. / Carbohydrate Research 339 (2004) 1717–1726
6. Bobleter, O.; Bonn, G. Carbohydr. Res. 1983, 124, 185–193.
15. de Bruijn, J. M.; Touwslager, F.; Kieboom, A. P. G.; van
Bekkum, H. Starch 1987, 39, 49–52.
16. de Bruijn, J. M.; Kieboom, A. P. G.; van Bekkum, H.; van
der Poel, P. W.; de Visser, N. H. M.; de Schutter, M. A.
M. Int. Sugar J. 1987, 89, 208–210.
17. Ellis, A. V.; Wilson, M. A. J. Org. Chem. 2002, 67, 8469–
8474.
18. Akiya, N. A.; Savage, P. E. Chem. Rev. 2002, 102, 2725–
2750.
19. Marshall, W. L.; Franck, E. U. J. Phys. Chem. Ref. Data
1981, 10, 295–304.
20. Lobry de Bruyn, C. A.; Alberda van Ekenstein,
W. Recl. Trav. Chim. Pays-Bas 1895, 14, 203–
216.
7. Luijkx, G. C. A.; van Rantwijk, F.; van Bekkum, H.;
Antal, M. J. Carbohydr. Res. 1995, 272, 191–202.
8. Kabyemela, B. M.; Adschiri, T.; Malaluan, R. M.; Arai,
K. Ind. Eng. Chem. Res. 1997, 36, 1552–1558.
9. Kabyemela, B. M.; Adschiri, T.; Malaluan, R. M.; Arai,
K. Ind. Eng. Chem. Res. 1999, 38, 2888–2895.
10. Kallury, R. K. M. R.; Ambidge, C.; Tidwell, T. T.;
Boocock, D. G. B.; Agblevor, F. A.; Stewart, D. J.
Carbohydr. Res. 1986, 158, 253–261.
11. Kruse, A.; Gawlik, A. Ind. Eng. Chem. Res. 2003, 42, 267–
279.
12. Antal, M. J., Jr.; Mok, W. S. L.; Richards, G. N.
Carbohydr. Res. 1990, 199, 91–109.
13. de Bruijn, J. M.; Kieboom, A. P. G.; van Bekkum, H.;
van der Poel, P. W. Sugar Technol. Rev. 1986, 13, 21–52.
14. de Bruijn, J. M.; Kieboom, A. P. G.; van Bekkum, H.
Starch 1987, 39, 23–28.
21. Kabyemela, B. M.; Adschiri, T.; Malaluan, R.; Arai, K.
Ind. Eng. Chem. Res. 1997, 36, 2025–2030.
22. Luijkx, G. C. A.; van Rantwijk, F.; van Bekkum, H.
Carbohydr. Res. 1993, 242, 131–139.