Reaction products and transformations of intermediates in the aqueous-phase reforming of sorbitol

Article abstract of DOI:10.1002/cssc.200900254

Aqueous phase reforming of sorbitol over Pt supported on an alumina catalyst is investigated, in order to identify the intermediates involved in the transformation of the initial feed. Parameters such as the sorbitol feed rate and temperature are studied. To identify the intermediates, an approach based on analysis of the gas and liquid phases as well as the total carbon content was developed. According to analysis by gas chromatography combined with mass spectrometry of volatile substances collected with solid-phase microextraction, over 260 compounds are involved in the transformation of sorbitol. Of these, 50 of the major products are identified with high reliability. It is shown that a great variety of compounds, bearing different functionalities, form part of the reaction network. The formation of the majority of identified compounds is explained and a reaction network for the formation of sorbitol and intermediate molecules transformation is proposed.

Full text of DOI:10.1002/cssc.200900254

DOI: 10.1002/cssc.200900254
Reaction Products and Transformations of Intermediates in
the Aqueous-Phase Reforming of Sorbitol
Alexey V. Kirilin,[a, b] Anton V. Tokarev,[a] Elena V. Murzina,[a] Leonid M. Kustov,[b]
Jyri-Pekka Mikkola,*[a, c] and Dmitry Yu. Murzin*[a]
Aqueous phase reforming of sorbitol over Pt supported on an
alumina catalyst is investigated, in order to identify the inter-
mediates involved in the transformation of the initial feed. Pa-
rameters such as the sorbitol feed rate and temperature are
studied. To identify the intermediates, an approach based on
analysis of the gas and liquid phases as well as the total
carbon content was developed. According to analysis by gas
chromatography combined with mass spectrometry of volatile
substances collected with solid-phase microextraction, over
260 compounds are involved in the transformation of sorbitol.
Of these, 50 of the major products are identified with high reli-
ability. It is shown that a great variety of compounds, bearing
different functionalities, form part of the reaction network. The
formation of the majority of identified compounds is explained
and a reaction network for the formation of sorbitol and inter-
mediate molecules transformation is proposed.
Introduction
Results and Discussion
Catalyst characterization
Aqueous-phase reforming (APR) of biomass for the production
of liquid fuels is of high importance nowadays because it pro-
vides an alternative to the utilization of fossil fuels; the availa-
bility of which is constantly decreasing.[1] However, the removal
of oxygen atoms from the initial feedstock (typically, sugars
and sugar alcohols) is a challenge,[2] and a significant number
of papers dealing with the conversion of biomass to hydrogen,
fuels, or chemicals can be found in the literature. Pioneering
studies in this area, performed by Dumesic and co-workers,
have mostly been devoted to the transformation of sugar alco-
hols or polyols (glycols and sorbitol) to hydrogen and alka-
nes.[3–8] Sorbitol, being a product of glucose hydrogenation, is
of high interest in the APR process, not only for the production
of molecular hydrogen but also for the production of alkanes.
The effective development of APR requires knowledge of
the intermediates involved and their transformations during
the process. Until now, aqueous processing is performed in the
presence of molecular hydrogen at relatively high sorbitol-to-
hydrogen feed ratios (i.e., 13.5).[7] Therefore, the main pathway
of sorbitol transformation is considered to comprise sequential
dehydration–hydrogenation steps. This process is called aque-
ous phase dehydration–hydrogenation (APD/H). Typically, bi-
functional catalysts that bear an active metal component (Pd,
Pt, Rh, Ru) on a support with acidic properties are applied.[9]
The present study is mainly targeted at investigating the re-
action network in the sorbitol transformations during APR pro-
cess, in order to reveal the main intermediates, product distri-
butions, and study a plausible reaction scheme. A commercial-
ly available 5 wt% Pt/Al2O3 (Degussa, F214 XSP) catalyst was
used. Parameters such as weight hourly space velocity (WHSV),
feedstock concentrations, and temperature were varied during
the experiments.
The metal dispersion of the catalyst was investigated by CO-
impulse chemisorption. For Pt/Al2O3 the metal dispersion was
34%, which corresponds to an average platinum metal particle
size of 3 nm. The temperature-programmed reduction profile
of the catalyst showed three peaks, at 75, 180, and 3508C
(Figure 1). The first two peaks can be attributed to reduction
of the Pt particles to their metallic state, while the last peak
may be assigned to additional adsorption of hydrogen by the
Pt species.
Identification of reaction products and intermediates:
General approach
To investigate the reaction products and identify the numerous
intermediate compounds that emerge during APR of sorbitol,
we focused our analysis on the gaseous products, because the
[a] A. V. Kirilin, Dr. A. V. Tokarev, E. V. Murzina, Prof. J.-P. Mikkola,
Dr. D. Yu. Murzin
Laboratory of Industrial Chemistry and Reaction Engineering Process
Chemistry Centre, ꢀbo Akademi University
Biskopsgatan 8, 20500 ꢀbo-Turku (Finland)
Fax:(+35) 82-154-479
E-mail: dmurzin@abo.fi
jpmikkol@abo.fi
[b] A. V. Kirilin, Prof. L. M. Kustov
Zelinsky Institute of Organic Chemistry
Leninsky Prosp. 47, 199991, Moscow (Russia)
[c] Prof. J.-P. Mikkola
Department of Chemistry, Technical Chemistry
Chemical Biological Centre, Umeꢁ University
SE-901 87 Umeꢁ (Finland)
Supporting Information for this article is available on the WWW under
http://dx.doi.org/10.1002/cssc.200900254.
708
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2010, 3, 708 – 718

Reaction Products and Intermediates in the Aqueous-Phase Reforming of Sorbitol
HPLC and GC were selected and confirmed in the present
study. Consequently, over 80% of the compounds present in
the total intermediate network were positively identified (the
other 180 compounds accounted for less than 20% of the
peaks, and their identification is still a challenge).
Analysis of the gaseous phase
The influence of the WHSV on sorbitol APR was studied by
flow rate intervals, from 0.1 to 0.5 mLminꢀ1. Several reference
experiments were applied to monitor the catalyst deactivation.
It was found that the reaction products were affected in differ-
ent ways. The concentration of low-molecular-weight products
such as CO2 and H2 slightly decreased, moreover, the molar
ratio between H2 and CO2 changed from 0.713 to 0.710 after
116 h time-on-stream (TOS; 99.6% of initial molar ratio) and
can be considered as constant. Total catalyst deactivation with
respect to CO2 and H2 observed after 116 h was ca. 5%, and
can be neglected. Generally, the gas phase in APR of sorbitol
contains hydrocarbons such as ethane, butane, pentane, and
hexane, the concentrations of which decrease with TOS, while
propane and isobutane are formed in larger amounts at higher
TOS values (Figures 2, 3, and 4. The presence of hydrocarbons
was confirmed by GC-MS analysis. Besides, the presence of aro-
Figure 1. Temperature-programmed reduction profile of the 5 wt% Pt/Al2O3
catalyst.
composition of the liquid phase has very recently been de-
scribed in detail by Li and Huber under similar conditions with
the same type of catalyst (Pt/Al2O3-SiO2).[10] For the identifica-
tion of the reaction products and intermediates, an analytical
approach was developed (Scheme 1).
Scheme 1. Analytical approach for the identification of reaction products
and intermediates in APR of sorbitol.
Figure 2. Formation of the APR products CO2 and H2 with TOS. Reaction
conditions: 2258C, 29.3 bar, 10 wt% sorbitol solution, 30 mLminꢀ1 N2 flow
rate.
Knowledge about the structure and concentrations of gas-
eous and volatile products formed in the APR process was ob-
tained through analysis of gas samples by means of GC-MS
and micro-GC equipment. Solid-phase microextraction (SPME),
a powerful technique for the investigation of products present
in vapors above a liquid phase, was employed (see Experimen-
tal Section; no separate organic phase, in addition to the aque-
ous one, was observed during the experiments). After adsorp-
tion the samples were analyzed by GC-MS, showing the pres-
ence of more than 260 compounds (see Supporting Informa-
tion, Figure S1; over 50 major products were identified with a
probability higher than 80%. The quality of peak determina-
tion implied the degree of coincidence of the spectrum ob-
tained for the molecule with the library database data typically
used in GC-MS analysis. The quality of determination is esti-
mated according to the fragmentation spectrum presented in
the database. Because the product composition was very com-
plicated, compounds with higher peak areas identified by
Figure 3. Formation of C2–C4 linear hydrocarbons with TOS. Reaction condi-
tions: 2258C, 29.3 bar, 10 wt% sorbitol solution, 30 mLminꢀ1 N2 flow rate.
ChemSusChem 2010, 3, 708 – 718
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemsuschem.org
709

J.-P. Mikkola, D. Yu. Murzin et al.
Figure 4. Formation of C5–C6 linear hydrocarbons and iso-butane with TOS.
Reaction conditions: 2258C, 29.3 bar, 10 wt% sorbitol solution, 30 mLminꢀ1
N2 flow rate.
matic compounds such as benzene and its derivatives was also
detected in trace amounts. The influence of change in catalytic
activity on the formation of these hydrocarbons is shown in
Table 1. CO and methane were not formed in reference experi-
ments (WHSV of 0.1 hꢀ1 at different TOS). The ratio of carbon
content in CO2 (APR) towards CxHy (side reactions) is shown in
Table 2, and was found to be independent of TOS.
All experiments were performed in two series: from low to
high WHSV and then backwards at intermediate WHSV values,
causing apparent scattering due to minor deactivation. The in-
Figure 5. a) Rate of formation, and b) yields of CO2 and H2 as a function of
WHSV in APR of sorbitol. Conditions: 2258C, 29.3 bar, 10 wt% sorbitol solu-
tion, 30 mLminꢀ1 N2 flow rate.
sorbitol contacts the cata-
lyst at the same level of
conversion (100%). Thus,
Table 1. Sorbitol APR gas-phase reaction products, and influence of deactivation.
TOS [h]
Gas-phase products [mmolminꢀ1
H2 CO2 C2H6
]
C3H8
i-C4H10
C4H10
C5H12
C6H14
more sorbitol is converted
to produce more H2 and
CO2. However, at WHSV>
2 hꢀ1 a decrease in con-
version (H2 and CO2 for-
mation) is observed. The
7
123
1.22ꢁ10ꢀ1 1.71ꢁ10ꢀ1
1.15ꢁ10ꢀ1 1.62ꢁ10ꢀ1
9.26ꢁ10ꢀ3 9.42ꢁ10ꢀ3
7.90ꢁ10ꢀ3 1.03ꢁ10ꢀ2
1.15ꢁ10ꢀ5 4.41ꢁ10ꢀ3 6.97ꢁ10ꢀ4 1.48ꢁ10ꢀ4
1.41ꢁ10ꢀ4 3.52ꢁ10ꢀ3 5.19ꢁ10ꢀ7 9.24ꢁ10ꢀ5
Difference ꢀ5.7%
ꢀ5.3%
ꢀ14.6%
+9.4%
+1121.8%
ꢀ20.3%
ꢀ25.6%
ꢀ37.6%
yields of H2 and CO2 decrease linearly with increasing WHSV, as
shown in Figure 5b.
Table 2. Carbon distribution between CO2 and hydrocarbons
The distribution of carbon between the liquid and gas
phases is shown in Table 3. All mass balances were monitored
independently by means of micro-GC and total organic carbon
(TOC), and were close to 100%.
TOS [h]
CO2 [mol]
ꢀ(X·CxHy) [mol]
ꢀCxHy/CO2
7
123
1.71ꢁ10ꢀ4
1.62ꢁ10ꢀ4
6.89ꢁ10ꢀ5
6.45ꢁ10ꢀ5
0.4
0.4
At a WHSV of 0.6 hꢀ1, ca. 73% of carbon was present in the
gas phase as CO2 and hydrocarbons. With an increase of the
WHSV to 0.9 hꢀ1, the carbon content decreased to 59.7% as
expected. Owing to the shorter contact time between the feed
solution and the catalyst, less fragmentation of sorbitol took
place and, hence, more carbon remained in the solution. As
shown in Figure 8, the carbon content in the liquid phase in-
creased with increasing WHSV, and 96% of carbon-containing
products were present in the liquid phase at a WHSV of
3.0 hꢀ1. For all C2–C6 hydrocarbons concentration maxima were
observed as a function of WHSV (Figure 6a and 7a), although
the yields decreased with increasing WHSV (Figure 6b and 7b).
fluence of deactivation for each substance can be taken into
account based on Table 1. Trends in concentration and yield
changes are shown by lines in Figures 5, 6, and 7.
Based on these trends, one can distinguish two types of be-
havior: The concentration of most gas products goes through
a maximum, observed in the WHSV interval from 1 to 2 hꢀ1
(Figure 5a, 6a, and 7a). The formation of H2 and CO2 is linearly
proportional to the WHSV in the interval 0.6–1.5 hꢀ1, and ex-
hibits maxima at 1.8 hꢀ1 (Figure 5a). At lower WHSV values H2
and CO2 are formed at 1.1ꢁ10ꢀ4 and 1.6ꢁ10ꢀ4 molminꢀ1, re-
spectively; however, these values increase with increasing
WHSV (lower contact time) due to the fact that actually more
In the case of propane this maximum corresponded to 1.2 hꢀ1
,
710
www.chemsuschem.org
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2010, 3, 708 – 718

Reaction Products and Intermediates in the Aqueous-Phase Reforming of Sorbitol
Figure 8. Liquid- and gas-phase carbon as a function of WHSV in APR of sor-
bitol. Reaction conditions: 2258C, 29.3 bar, 10 wt% sorbitol solution,
30 mLminꢀ1 N2 flow rate.
and the alkane flow rate to ca. 1.25ꢁ10ꢀ5 molminꢀ1 (Fig-
ure 6a). The propane yield decreased from 2.8% to 0.4% when
the WHSV was changed from 0.6 to 3 hꢀ1. The formation of
ethane and n-butane exhibited maxima at 1.5 hꢀ1, correspond-
ing to flow rates of 6ꢁ10ꢀ6 and 10ꢀ5 molminꢀ1, respectively.
The yields of linear C2–C4 hydrocarbons exhibited a similar ten-
dency and decreased with increasing WHSV (Figure 6b). As
shown in Figure 7a, concentration maxima versus WHSV were
also found for hydrocarbons with higher number of atoms as
well as for nonlinear ones (isobutane). Although the yields of
these did not exceed 0.25%, a tendency of decreasing yield
with increasing WHSV was observed (Figure 7b). In general,
CO2 was the main component in the gas phase, varying in con-
tent from 71% at 0.6 hꢀ1 to 88% at 3 hꢀ1 (60%–80% of the
gas phase was reported to be CO2 in APD/H of sorbitol at high
WHSV of 2.9–11.6 hꢀ1[10]).
Figure 6. a) Rate of formation, and b) yields of C2–C4 linear hydrocarbons
and CO as a function of WHSV in APR of sorbitol. Conditions: 2258C,
29.3 bar, 10 wt% sorbitol solution, 30 mLminꢀ1 N2 flow rate.
The CO concentration during experiments was very low, and
close to 0 at lower WHSV values (0.6–1.2 hꢀ1; Table 3), but an
increase in concentration was observed between 1.5 hꢀ1 and
2.4 hꢀ1 with a maximum flow rate of 1.15ꢁ10ꢀ5 (Figure 6a; se-
lectivity 2.5%). These results agree well with earlier litera-
ture,[10] where under APD/H conditions the concentration of
CO was below the detection limit. Thus, low concentrations of
CO and, on the other hand, a high gas-phase selectivity to CO2
(71%–88%) reveal that CO is almost totally converted into CO2
and H2 via the water–gas shift (WGS) reaction which, as known
from literature,[12] is near equilibrium under these conditions.
Thus, at lower space velocities (0.6–1.2 hꢀ1) all CO formed is
converted to CO2 via the WGS reaction. With an increase of
WHSV, the activity of the catalyst towards WGS decreases
slightly and, as a result, the CO/CO2 ratio rises (to 0.03 at 1.5
and 1.8 hꢀ1). The decrease of the CO/CO2 ratio to 0.01 with fur-
ther increase of the WHSV to 2.1 and 2.4 hꢀ1 and the absence
of CO at higher WHSV of 3 hꢀ1 shows that the formation of CO
at these conditions became less favorable (Table 4).
The selectivity to CxHy hydrocarbons increased with decreas-
ing WHSV, as shown in Table 4 (from 2.1 to 0.6 hꢀ1). This cor-
roborates earlier findings,[10] where selectivity to C2–C6 alkanes
increased with decreasing WHSV. However, in the present work
the selectivity to alkanes in the gas phase exhibited much
Figure 7. a) Rate of formation, and b) yields of iso-butane and C5–C6 hydro-
carbons as a function of WHSV in APR of sorbitol. Reaction conditions:
2258C, 29.3 bar, 10 wt% sorbitol solution, 30 mLminꢀ1 N2 flow rate.
ChemSusChem 2010, 3, 708 – 718
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemsuschem.org
711

J.-P. Mikkola, D. Yu. Murzin et al.
ics, and their derivatives. Data
collected from analysis of the
liquid phase in APR of sorbitol at
various WHSV values is present-
ed in the Table 1. In the present
study it was possible to identify
66% to 91% of the compounds
presented in the liquid phase,
which is comparable to literature
data (70 to 100% of compounds
were reported to be identified in
under APD/H conditions for a
5 wt% solution of sorbitol).[10]
Table 3. Carbon distribution between the gas and liquid phase, as well as total carbon balance, in sorbitol
APR. Reaction conditions: 2258C, 29.3 bar, 10 w.% sorbitol solution, 30 mLminꢀ1 N2 flow rate.
Flow rate WHSV CxHy
CO
[molminꢀ1] [molminꢀ1] [molminꢀ1] CO2 inlet
ratio [molminꢀ1
CO2
CO/ Carbon at Gas-phase Liquid-
carbon [%] phase
carbon [%]
Carbon
balance [%]
[mLminꢀ1
]
[hꢀ1
]
]
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.6
0.9
1.2
1.5
1.8
2.1
2.4
3
6.89ꢁ10ꢀ5
7.75ꢁ10ꢀ5
9.57ꢁ10ꢀ5
0
0
0
1.71ꢁ10ꢀ4
2.21ꢁ10ꢀ4
2.84ꢁ10ꢀ4
0
0
0
3.3ꢁ10ꢀ4 72.7
5.0ꢁ10ꢀ5 59.7
6.7ꢁ10ꢀ4 57.0
24.4
35.9
39.3
52.2
55.6
84.5
78.4
96.0
97.1
95.6
96.3
95.0
101.8
101.7
99.8
5.92ꢁ10ꢀ5 9.29ꢁ10ꢀ6 2.88ꢁ10ꢀ4 0.03 8.3ꢁ10ꢀ4 42.8
9.15ꢁ10ꢀ5 1.15ꢁ10ꢀ5 3.59ꢁ10ꢀ4 0.03 1.0ꢁ10ꢀ3 46.2
2.21ꢁ10ꢀ5 1.94ꢁ10ꢀ6 1.77ꢁ10ꢀ4 0.01 1.2ꢁ10ꢀ3 17.2
1.06ꢁ10ꢀ4 2.09ꢁ10ꢀ6 1.76ꢁ10ꢀ4 0.01 1.3ꢁ10ꢀ3 21.4
0.5
3.22ꢁ10ꢀ5
0
n.d.
0
1.7ꢁ10ꢀ3
>2.0
>98.0
At a space velocity of 0.6 hꢀ1
the liquid phase consisted mainly of C4 compounds (selectivity
38.4%), C6 (19.5%), C3 (8.9%), and C5 (8%), whilst the presence
of C2 species was found to be minor (2.3%). With an increase
of WHSV from 0.6 to 0.9 hꢀ1 the selectivity towards C6 com-
pounds increased to 21.6%. This can be rationalized by consid-
ering that when the contact time between the catalyst and the
feed decreases, less fragmentation of initial sorbitol takes
place. This tendency was observed for the whole range of
WHSV values studied. At a higher WHSV value the selectivity
towards C6 molecules was around 31%. With an increase of
the WHSV from 0.6 to 0.9 hꢀ1 the selectivity towards C2 and C5
molecules increased from 2.3% to 7.4% and from 8.0% to
12%, respectively, whereas the presence of C3 molecules de-
creased by 1% only, from 8.9% to 7.9% (Table 5). The selectivi-
Table 4. Selectivity towards main sorbitol APR gas-phase components.
Reaction conditions: 2258C, 29.3 bar, 10 wt% sorbitol solution,
30 mLminꢀ1 N2 flow rate.
WHSV
Gas-phase
carbon [%]
Selectivity to
CxHy [%]
Selectivity to
CO [%]
Selectivity to
CO2 [%]
[hꢀ1
]
0.6
0.9
1.2
1.5
1.8
2.1
2.4
3
72.7
59.7
57.0
42.8
46.2
17.2
21.4
2–4
29
26
25
16.6
19.5
11
37.3
>50
0
0
0
2.4
2.5
1
0.7
0
71
74
75
81
78
88
62
n.d.
higher values (>50%) at higher
WHSV (see Table 4). These results
can be explained by a much
shorter contact time between
the feed and the catalyst, lead-
ing to a decrease of the APR re-
action but promoting alkane for-
mation. Still, at these WHSV
values (2.4 and 3 hꢀ1) the flow
rates of hydrocarbons were low
(1.06ꢁ10ꢀ4
10ꢀ5 molminꢀ1
and
3.22ꢁ
, respectively).
Moreover, most of the carbon
was present in the liquid phase
(78.4% and 96% for 2.4 and
3 hꢀ1), hence, the yield of hydro-
carbons can be considered low.
Figure 9. Compounds detected in APR of sorbitol with quality of determination >80% by MS.
Analysis of the liquid phase
ty to C4 compounds decreased with increasing WHSV (decrease
of the contact time).
Analysis of the liquid phase was performed mainly by means
of SPME followed by GC-MS. The compounds observed in the
liquid phase when applying various feed rates and tempera-
tures were classified according to their functional groups.
Compounds detected during the analysis of the liquid phase
with of peak determination quality over 80% are presented in
Figure 9. The compounds identified by HPLC were mainly ke-
tones, alcohols, furanes, cyclic ethers, carboxylic acids, aromat-
The carbon distribution for molecules comprising various
numbers of carbon atoms is shown in Figure 10. The selectivity
to C6 decreases and selectivity to C4 and C5 increases with de-
creasing WHSV, as discussed above. However, the selectivity to-
wards C3 goes through a maximum at approximately 1.8 hꢀ1.
As discussed above, the formation of CO also displays a maxi-
mum at 1.8 hꢀ1. This suggests that the main formation of CO
712
www.chemsuschem.org
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2010, 3, 708 – 718

Reaction Products and Intermediates in the Aqueous-Phase Reforming of Sorbitol
structure was also observed by
Huber and Li.[10] These mole-
cules originate from several con-
secutive sorbitol dehydration–
hydrogenation steps, leading to
the formation of compounds
with less hydroxyl groups then
in the starting material. Howev-
er, the presence of low amounts
of triols (glycerol and hexane-
triol-1,2,6; selectivity 8.4%) indi-
cated that dehydration–hydro-
genation reactions occur easily.
Table 5. Carbon distribution in the liquid-phase APR of sorbitol.
WHSV
Liquid-phase
carbon [%]
Selectivity
[%]
Total products
detected [%]
[hꢀ1
]
C2
C3
C4
C5
C6
Unknown
0.6
0.9
1.2
1.5
1.8
2.1
3
24.4
35.9
43.0
52.2
53.8
82.8
96
2.3
7.4
4.9
4.8
4.5
6.7
8
8.9
7.9
38.4
41.7
23.3
28.4
16.9
9.3
8.0
12
6.8
–
3.9
3.6
3.6
19.5
21.6
31.7
25.6
15.8
35.5
26.9
22.9
9.3
77.1
90.6
78.7
83.7
66.1
68.5
70.4
12.0
24.9
25.0
13.4
16.9
21.3
16.3
33.8
31.5
29.6
15
Ketones (acetone, 2-butanone, 2-hexanone, and 3-hexanone)
and carboxylic acids (acetic, butyric and hexanoic acids) were
found in amounts of 5% and 4.9%, respectively, while the se-
lectivity for other compounds (hydroxyacetone, isosorbide, and
2,5-dimethyltetrahydrofurane) was 18.7%.
Overall scheme
Based on the experimental evidence obtained from various an-
alytical techniques, several possible pathways of sorbitol trans-
formation in the APR process can be envisaged (Schemes 2, 3,
and 4).
Figure 10. Carbon selectivity as a function of WHSV. Conditions: 2258C,
2.93 MPa, 50 mLminꢀ1 N2 flow, 1 g Pt/Al2O3 catalyst.
Initially, sorbitol undergoes aqueous reforming to produce
hydrogen. Hydrogen released from the APR and WGS reactions
is then used in hydrogenation and deoxygenation steps. Sorbi-
tol is further converted into less oxygenated derivatives, con-
taining five or less oxygen atoms, through sequential dehydra-
tion–hydrogenation reactions, as shown in Scheme 3. Oxygen-
ates comprising less than six carbon atoms (C4–C5) can origi-
nate from decarbonylation reactions, and the CO released is
converted into CO2 via WGS. For example, the formation of
linear mono- and diketones containing five and six carbon
atoms is apparently the result of several dehydration and hy-
drogenation reactions, omitting scission of CꢀC bonds as well
as possible decarbonylation processes. The cleavage of CꢀC
bonds leads to products with less carbon atoms, even resulting
in C1–C2 products: methanol, ethanol, formic acid, acetic acid,
and methane. These products were identified by SPME fol-
lowed by GC-MS as well as micro-GC. However, methane for-
mation was low during the experiments. This implies that
under the chosen experimental conditions C2 molecules (e.g.,
ethanol) are formed in very low amounts, because the pres-
ence of these compounds leads to elevated methane forma-
tion via various side reactions. On the other hand, the breaking
of the initial sorbitol skeleton can also result from retro-aldol
condensation reactions. This in turn leads to the formation of
polyols with three or four carbon atoms, that can be further
converted into linear ketones and alcohols, via dehydration–
hydrogenation, and then to light alkanes.
and, hence, H2 via APR occurs from C3 molecules. When the
CO formation decreases the concentration of C3 compounds
decreases also. The selectivity to C2 compound increased
slightly with an increase of WHSV.
The composition of the liquid products was similar at all
WHSV values studied, thus we report here only the detailed
analysis for WHSV=0.9 hꢀ1, where over 90% of the liquid
products could be reliably identified by HPLC. The distribution
for various classes of organic compounds is illustrated in the
Table 6. As expected, the main oxygenated species formed are
Table 6. Compound distribution according to functionality in liquid-phase
APR of sorbitol.
WHSV Liquid-
Product distribution[a]
[%]
Total products
detected [%]
[hꢀ1
]
phase
carbon [%]
alcohols diols triols ketones acids other
0.9
24.4
17.3
49.9 4.2 5.0
4.9 18.7 90.6
[a] (amount of compounds)/(total amount of detected compounds)ꢁ100%.
molecules that bear one or more hydroxyl groups: alcohols
(selectivity 17.3%) and diols (selectivity 49.9%). The alcohols
included mainly propanols (1-propanol and 2-propanol), 2-bu-
tanol, and tetrahydropyrane-2-methanol, while the diols com-
prised ethyleneglycol, propanediols, butanediols, 1,2-pentane-
diol, and hexanediols. The presence of compounds of similar
The formation of substituted and unsubstituted cyclic ethers
(saturated furane and pyrane derivatives, detected by SPME) is
supposed to result from the dehydration of intermediate oxy-
genates, that is, a ring-closure reaction. These species can then
undergo, for example, dehydrogenation, resulting in the for-
ChemSusChem 2010, 3, 708 – 718
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemsuschem.org
713

J.-P. Mikkola, D. Yu. Murzin et al.
Scheme 2. Overall scheme of sorbitol transformation during APR.
mation of unsaturated derivatives (e.g., substituted furanes).
Another pathway for their transformation is hydrogenation of
CꢀO bonds, leading to alcohols and then alkanes. The forma-
tion of molecules that have more than six carbon atoms is con-
sidered to result from aldol-like condensation processes or var-
ious cyclization processes that involve OH and carbonyl groups
(Scheme 2).
undergo aldol condensation reactions (Schemes 2 and 3).
Moreover, another plausible pathway for the formation of
acids is a consecutive desorption–rearrangement step of ad-
sorbed polyol species on the metal surface, as proposed earlier
by Dumesic et al. (Scheme 4).[9] The formation of cyclic ketones
and furane derivatives typically occurs via ring-closing aldol-
condensation reactions or via dehydration of intermediate
products during sorbitol conversion.
Nonetheless, having observed some of the condensation
products with ketones, no aldehydes were detected (except
for low amounts of butanal and vinylaldehyde). Ketones them-
selves can undergo aldol condensation reactions without the
presence of aldehydes to form branched condensation prod-
ucts. The major reason for this could be their disproportiona-
tion over the metal surface, resulting in the formation of car-
boxylic acids and alcohols. Both acids and aliphatic alcohols
containing equal numbers of carbon atoms were observed,
suggesting dispropotionation of aldehydes as the origin of
their formation (Scheme 3). Nevertheless, this observation of
both acids and aliphatic alcohols containing equal number of
carbon atoms does not necessarily justify the assumption that
aldehydes undergo only disproportionation reactions. It is pos-
sible that aldehydes were not detected because they rapidly
An interesting observation that has no evident explanation
at the moment is the presence of aromatic compounds in the
final product mixture. Consecutive dehydration–dehydrogena-
tion of cyclic intermediates can lead to the formation of ben-
zene and its derivatives as well as aromatization of C6 alkanes.
Influence of reaction parameters on conversion and product
formation
Four experiments were carried out at a 0.1 mLminꢀ1 feed rate,
at various temperatures. The conversion of sorbitol as a func-
tion of temperature is displayed in Figure 11a and was close to
100%. Conversion of sorbitol as a function of feed rate is pre-
sented in Figure 11b. At low sWHSV the conversion was close
714
www.chemsuschem.org
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2010, 3, 708 – 718

Reaction Products and Intermediates in the Aqueous-Phase Reforming of Sorbitol
to 100%, whereas it decreased to 52% at 2.4 hꢀ1 and to 18%
at 3 hꢀ1, respectively. These results are in a good accordance
with those recently reported by Li and Huber.[10] At all temper-
atures compounds with various functional groups were ob-
served by SPME: alcohols, ketones, acids and products of their
aldol condensation, or compounds being the result of, for ex-
ample, retro-aldol reactions[13]
.
Thus, as discussed above, at 2258C and at a space velocity
of 0.6 hꢀ1 the final mixture of liquid-phase products was
mainly composed of C4 and C6 compounds bearing a single
carbonyl, carboxylic, or hydroxyl group. As a result of sorbitol
fragmentaion, molecules containing three and four carbon
atoms were also detected. These comprised to a major extent
carboxylic acids (butanoic acid) as well as linear ketones (ace-
tone and butanones). Apparently the formation of linear un-
branched ketones takes place via consecutive dehydration–hy-
drogenation reactions of initial sorbitol, occasionally followed
by decarbonylation (releasing CO), which in turn leads to the
formation of C5 products, and so forth.
The relative content of C6 products increased with increasing
feed rates. This can be explained by the shorter contact time,
which in turn leads to less CꢀC cleavage and fragmentation of
initial sorbitol.
Scheme 3. Plausible transformation of intermediates over metal and acid
sites of the APR catalyst.
Scheme 4. Formation of compounds bearing various functionalities, and their transformation during APR of sorbitol over Pt/Al2O3.
ChemSusChem 2010, 3, 708 – 718
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemsuschem.org
715

J.-P. Mikkola, D. Yu. Murzin et al.
Figure 11. Conversion of sorbitol a) as a function of temperature, and b) as a
function of WHSV at 2258C. Conditions: 1 g 5 wt% Pt/Al2O3 catalyst, 10 wt%
Figure 12. Total carbon content during APR of sorbitol a) in solution as a
function of initial substrate concentration at 2258C, and b) in the gas phase
as a function of temperature at 10 wt% sorbitol. Conditions: 5 wt% Pt/Al2O3,
sorbitol solution, feed rate 0.1 mLminꢀ1, N2 flow 30 mLminꢀ1
.
sorbitol feed rate 0.1 mLminꢀ1, N2 flow 30 mLminꢀ1
.
Influence of temperature, sorbitol feed rate, and feed
concentration on the carbon distribution
Conclusions
In the present study, the reaction intermediates and products
from APR of sorbitol over a Pt/Al2O3 catalyst have been investi-
gated under various experimental conditions (WHSV, tempera-
ture, sorbitol concentration). An approach for analyses of the
liquid- and gas-phase products (by GC-MS, micro-GC, HPLC,
and SPME) as well as TOC was developed. Over 260 com-
pounds were detected and over 50 of them (major compo-
nents) were reliably identified by using GC-MS. It was shown
that various molecules with different functional groups are
formed during APR process. They were classified according to
the presence of these groups and the number of carbon
atoms. It was demonstrated that the main reactions that occur
during APR are dehydration to olefins and their consecutive
hydrogenation, decarbonylation, and fragmentation of the ini-
tial substrate via CꢀC or CꢀO cleavage. CꢀO scission apparent-
ly occurs much more easily; however, the presence of adjacent
CꢀO bonds facilitates cleavage of CꢀC bonds.
The concentration of carbon in the liquid phase increased with
the feed rate in a linear manner (Figure 8). Figure 12a shows
that the total carbon content in the liquid phase depends line-
arly on the sorbitol concentration. Therefore, more concentrat-
ed solutions are better feedstocks for the production of liquid
fuels, because less fragmentation occurs, leading to less vola-
tile products. Consequently, when utilizing a 10 wt% substrate
solution, 31% of carbon was found in the liquid phase while in
the case of 20 and 30 wt% solutions 46% and 65% of carbon
was found in the liquid phase, respectively. Since with increas-
ing feed rate and, subsequently, shorter contact times the con-
version of sorbitol is lower, this leads to fewer consecutive re-
actions that would result in gas-phase products.
To investigate the influence of the temperature, APR was
conducted at three different temperatures at a constant feed
rate of 0.1 mLminꢀ1 over 1 g of Pt/Al2O3 catalyst. As demon-
strated in Figure 12b, the carbon content in the gas phase in-
creases with temperature. Thus, at 2158C 57% of carbon is
present in the gas phase, whilst at 217 and 2258C the amounts
of carbon are 60% and 75%, respectively. It is evident that
higher temperatures favor the fragmentation of the initial
feedstock through CꢀC and CꢀO scission, forming undesirable
volatile products.
The influence of the sorbitol feed rate, temperature, and so-
lution concentration on the total carbon content in the liquid
and gas phases was investigated. The total carbon content in
the gas phase increases with increasing temperature. Likewise,
a linear increase of total carbon in the liquid phase was ob-
served depending on the initial sorbitol concentration and
feed rate. This was explained by an increase of the residence
time or, in other words, a decrease of the conversion.
716
www.chemsuschem.org
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2010, 3, 708 – 718

Reaction Products and Intermediates in the Aqueous-Phase Reforming of Sorbitol
Based on the results of the experiments, a feasible reaction
scheme involving transformations of the substrate and inter-
mediates was proposed.
APR of sorbitol was studied at four different temperatures
(210, 217, 221, and 2258C). For each temperature, the internal
pressure in the reactor was maintained ca. 4 bar above the crit-
ical pressure of water vapor,[8] and the nitrogen flow rate was
varied from 20 to 50 mLminꢀ1. Because the generation of
gases during the reaction results in dilution of the inlet gas, N2
with 1% He was used to enable quantitative calculation of the
gaseous products. An aqueous solution of 10, 20, or 30 wt%
sorbitol (99.5%, Fluka) was fed continuously, at different flow
rates (0.1–0.5 mLminꢀ1). Prior to an experiment the reactor
was pressurized, gas and liquid flows were set to the desired
values, and the heating was turned on (58Cminꢀ1). As soon as
the process temperature was reached, the on-line GC analysis
was turned on.
Experimental
Catalyst
A commercially supplied 5 wt% Pt/Al2O3 (Degussa, F214 XSP) cata-
lyst was tested. The catalyst was pressed into pellets, crushed, and
then sieved to 125–250 mm fractions. The metal dispersion was de-
termined by CO pulse chemisorption in an apparatus manufac-
tured by Micrometrics (Autochem 2900). The catalyst was reduced
prior to measurement with the following program: 25–508C at
108Cminꢀ1 in He, dwell for 30 min, gas-switch to H2, 58Cminꢀ1 to
2508C, dwell for 2 h, followed by flushing for 60 min in He at
2508C. Thereafter the catalyst was cooled to ambient temperature
and CO pulses were introduced utilizing 10 vol% CO in He. A Pt/
CO stoichiometry of 1:1 was assumed. TPR of the catalyst was per-
formed in an AutoChem instrument by Micrometrics, using the fol-
lowing program: 58Cminꢀ1–4008C.
The on-line GC analysis of reaction products was performed
with a HP5890 Series II gas chromatograph. The product gas
stream was passed through an additional mass flow meter at a
flow rate of 10 mLminꢀ1. The sampling is an important aspect
of data collection, because there is the potential danger of a
pressure decrease to below the critical value. Therefore, the
sampling unit was constructed with two consecutive valves
with a sampling loop (2 mL) in between. A special liquid trap
was introduced to protect the flow meter from water droplets
and condensation. The instrument was equipped with both
a TCD and FID detector. The current response from detec-
tors was converted by a 35900D analog/digital converter and
then calculated by HP Chemstation Software (version A.05.04).
The GC instrument was equipped with a HP-MOLSIV column
Modification of the reactor
Special attention was paid to the reactor development, in par-
ticular to the precise temperature control. Thermocouples
were embedded along the catalyst bed at intervals of 1 cm
(total length: ca. 12 cm; see Supporting Information, Fig-
ure S2). Experiments with only one heater demonstrated that
the temperature inside the reactor was lower than the oven
temperature by at least 68C and varied by a range of 358C.
Most probably the main heat losses occurred along the metal-
lic tubing, and therefore the heating system was modified to
include three different heaters for separate heating to allow
more precise control of temperature in reactor inlet, outlet,
and the catalyst bed. A comparison between the two ap-
proaches for reactor heating is provided in Figure S2. After this
modification the temperature along the reactor was main-
tained within ꢁ18C throughout (it has been reported that the
difference in temperature between the middle and the top of
the reactor can reach 158C[10]). Although short fluctuations of
the temperature can occur due to the periodic character of
heating cycles and a hysteresis in the oven and temperature
inside the reactor, their impact was diminished by improving
the heat isolation of the reactor and parts in the vicinity of the
tubing (Figure S3). Nevertheless, the temperature profile was
significantly influenced by the liquid flow rate, which could be
countered by adjusting the temperature of the heating zones.
The influence of the liquid flow rate on the temperature distri-
bution within the reactor revealed that for each experimental
condition the heating zones should be configured in advance
by carrying out a blank experiment. The influence of the gas
flow was similar but less pronounced.
(30 m) at 408C, Tinj =1208C, Vloop =1 mL, TFID =2708C, TTCD
2108C.
=
Volatile compounds were analyzed by GC-MS as well as
headspace SPME (reported earlier to be successful for the de-
termination of furfural and 5-methylfurfural in liquid sam-
ples[11]). An aliquot of the solution (2 mL) containing the reac-
tion products was transferred into
a small (4 mL) flask
equipped with a rubber cap, and the sample flask was heated
to 320 K. A needle-containing fiber was penetrated through
the cap into the bottle and was exposed to the headspace of
the sample for 30 min. The fiber used for the extraction was
coated with 2 cm–75 mm of carboxen/polymethylsiloxane
(CAR/PDMS) and purchased from Supelco (Bellefonte, PA, USA).
The holder used for the manual injection was also supplied by
Supelco. The components were enriched on the surface and as
equilibrium was reached between the headspace and the fiber,
the syringe was injected into the GC-MS for analysis. The inlet
chamber was set to 543 K, at which temperature the absorbed
and adsorbed analytes were thermally desorbed into the hot
injector of the gas chromatograph. The desorption time was
10 min to ensure total desorption. Moreover, no memory effect
was observed as the same fiber was inserted for the second
time. The GC-MS was equipped with a capillary column (DB-
Petro 50 mꢁ0.2 mmꢁ0.5 mm). The following temperature pro-
gram was used: dwelling for 10 min at 313 K, heating
0.9 Kminꢀ1 to 348 K, followed by heating 1.1 Kminꢀ1 to 393,
heating 10 Kminꢀ1 to 473 K, and dwelling at 473 K for 20 min.
Liquid samples were periodically withdrawn from the reactor
and analyzed by HPLC (injection volume 2 mL, Aminex HPX-
During a typical experiment, the stainless steel reactor (Fig-
ure S3; d=4.2 mm) was loaded with 1 g of catalyst. Catalyst re-
duction was done in situ with molecular hydrogen (AGA Oy,
99.999%) at 2508C for 2 h.
ChemSusChem 2010, 3, 708 – 718
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemsuschem.org
717

J.-P. Mikkola, D. Yu. Murzin et al.
87H, eluent 5 mmolLꢀ1 H2SO4, flow rate 0.6 mLminꢀ1, 458C,
70 min) using a refractive index (RI) detector.
Keywords: biofuels · heterogeneous catalysis · platinum ·
reaction mechanisms · renewable resources
An Agilent Micro GC 3000 A was used for analysis of the
gas-phase products, enabling us to analyze all possible gas
products starting from H2 up to C7 hydrocarbons. The device
was equipped with four parallel GC systems with TCD detec-
tors. Four (Plot U, OV-1, Alumina, and Molsieve) columns were
used for product separation. Three additional short columns
served as precolumns. Undesirable admixtures in the main col-
umns were avoided by using the back-flash approach. Sorbitol
conversion was estimated from the decrease of the sorbitol
peaks areas:
[1] G. W. Huber, S. Iborra, A. Corma, Chem. Rev. 2006, 106, 4044.
[2] E. L. Kunkes, D. A. Simonetti, R. M. West, J. C. Serrano-Ruiz, C. A. Gaert-
ner, J. A. Dumesic, Science 2008, 322, 417.
[3] J. W. Shabaker, R. R. Davda, G. W. Huber, R. D. Cortright, J. A. Dumesic, J.
Catal. 2003, 215, 344.
[4] J. W. Shabaker, G. W. Huber, J. A. Dumesic, J. Catal. 2004, 222, 180.
[5] G. W. Huber, J. W. Shabaker, J. A. Dumesic, Science 2003, 300, 2075.
[6] J. N. Chheda, G. W. Huber, J. A. Dumesic, Angew. Chem. 2007, 119, 7298;
Angew. Chem. Int. Ed. 2007, 46, 7164.
[7] R. M. West, M. H. Tucker, J. A. Dumesic, Catal. Commun. 2009, 10, 1743.
[8] D. A. Simonetti, J. A. Dumesic, Catal. Rev. 2009, 51, 441,and references
cited therein.
ꢀ
ꢁ
Peak area ðsorbitolÞ
Conversion ¼ 1 ꢀ
ꢂ 100%
ð1Þ
Initial peak area ðsorbitolÞ
[9] R. R. Davda , J. W. Shabaker, G. W. Huber, R. D. Cortnight, J. A. Dumesic,
Appl. Catal. B: Environm. 2005, 56, 171.
[10] N. Li, G. W. Huber, J. Catal. 2010, 270, 48.
[11] L. Giordano, R. Calabrese, E. Davoli, D. Rotilio, J. Chromat. 2003, 1017,
141.
[12] R. R. Davda, J. A. Dumesic, Angew. Chem. 2003, 115, 4202; Angew. Chem.
Int. Ed. 2003, 42, 4068.
Acknowledgements
This work is part of the activities at the ꢀbo Akademi Process
Chemistry Centre (ꢀA-PCC) within the Finnish Centre of Excellence
Programme (2000–2011) appointed by the Academy of Finland.
Financial support from the Fortum Foundation is gratefully ac-
knowledged. The authors are grateful to Leif ꢂsterholm for TOC
analysis. The National Research Agency for Innovation and Tech-
nology (TEKES) is acknowledged for financial support.
[13] E. P. Maris, R. J. Davis, J. Catal. 2007, 249, 328.
Received: November 3, 2009
Revised: February 15, 2010
Published online on May 28, 2010
718
www.chemsuschem.org
ꢀ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemSusChem 2010, 3, 708 – 718