Aqueous Phase Reforming of ethylene glycol - Role of intermediates in catalyst performance

Article abstract of DOI:10.1016/j.jcat.2012.05.019

Liquid product formation during the aqueous catalytic reforming of ethylene glycol (EG) was studied up to 450 °C and 250 bar pressure. Methanol, ethanol, and acetic acid were the main liquid by-products during EG reforming in the presence of alumina-supported Pt and Pt-Ni catalysts. The effect of these by-products on selectivity and catalyst stability was further investigated by studying reforming of these components. Reforming of these products was shown to be responsible for the formation of alkanes. The high dehydrogenation activity of Pt-Ni catalysts leads to high H₂ yields during EG reforming by (i) suppressing the formation of methane during methanol reforming (a major by-product of EG reforming) and (ii) suppressing the formation of acetic acid. In addition, the decrease in acetic acid formation showed a positive effect on catalyst lifetime. Acetic acid was found to be responsible for hydroxylation of the Al₂O₃ support, leading to migration and coverage of the metal particles by Al(OH)_x and resulting in deactivation of the Pt-based catalysts.

Full text of DOI:10.1016/j.jcat.2012.05.019

Journal of Catalysis 292 (2012) 239–245
Contents lists available at SciVerse ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Aqueous Phase Reforming of ethylene glycol – Role of intermediates
in catalyst performance
⇑
D.J.M. de Vlieger, B.L. Mojet, L. Lefferts, K. Seshan
Catalytic Processes and Materials (CPM), MESA+ Institute, Faculty of Science and Technology, University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands
a r t i c l e i n f o
a b s t r a c t
Article history:
Liquid product formation during the aqueous catalytic reforming of ethylene glycol (EG) was studied up
to 450 °C and 250 bar pressure. Methanol, ethanol, and acetic acid were the main liquid by-products dur-
ing EG reforming in the presence of alumina-supported Pt and Pt–Ni catalysts. The effect of these by-
products on selectivity and catalyst stability was further investigated by studying reforming of these
components. Reforming of these products was shown to be responsible for the formation of alkanes.
The high dehydrogenation activity of Pt–Ni catalysts leads to high H2 yields during EG reforming by (i)
suppressing the formation of methane during methanol reforming (a major by-product of EG reforming)
and (ii) suppressing the formation of acetic acid. In addition, the decrease in acetic acid formation showed
a positive effect on catalyst lifetime. Acetic acid was found to be responsible for hydroxylation of the
Al2O3 support, leading to migration and coverage of the metal particles by Al(OH)x and resulting in deac-
tivation of the Pt-based catalysts.
Received 22 February 2012
Revised 27 April 2012
Accepted 27 May 2012
Available online 29 June 2012
Keywords:
Boehmite
Raman
FTIR spectroscopy
Deactivation
Acetic acid
Sustainable
Ó 2012 Elsevier Inc. All rights reserved.
1. Introduction
Many studies on the catalytic reforming of model oxygenates
under APR conditions have been reported recently [9–13]; ethanol,
In the future, hydrogen from renewable resources is projected
to replace fossil fuels, at least partly, to help reduce anthropogenic
CO2 emissions [1]. A widely studied process for the production of
‘green’ hydrogen is reforming of oxygenates, for example alcohols,
acids or aldehydes [2,3]. Such oxygenates appear in waste streams
of food and other bio-based processes, for instance, the aqueous
fraction of biomass pyrolysis oil is one such example. Typical con-
centrations of oxygenates in these streams are up to 20 wt% [4].
One way to generate hydrogen from these components is con-
ventional steam reforming followed by water gas shift reaction.
These process steps are energy intensive because (i) they are usu-
ally carried out at high temperatures (600–700 °C) [5] and (ii) in
the case of aqueous waste streams, it requires evaporation of the
large amounts of the water present [6]. Dumesic and co-workers
[2] tackled this problem by developing the Aqueous Phase Reform-
ing (APR) process by keeping water in the liquid phase by exerting
increased pressures. This process is typically carried out in the
range of 175–265 °C and 34–56 bar. Thermodynamics favor hydro-
gen at these low temperatures, but the reaction rates are slow [7].
Increasing the temperature to obtain commercially viable reaction
rates has the disadvantage of increased formation of undesired al-
kanes [8], thus reducing the hydrogen yield.
ethylene glycol, glycerol are typical examples. Information on
mechanistic sequences that take place during APR is limited, and
in the case of ethylene glycol (EG), Dumesic et al. [13] suggested
that reforming is initiated via dehydrogenation and formation of
an adsorbed oxygenate (acetylinic dialcohol type) intermediate.
In a desired sequence, C–C bond cleavage and subsequent water
gas shift leads to CO2 and H2. However, rearrangement/desorption
of the oxygenate intermediate leads to acids and/or one C–O bond
scission to formation of alcohols. Sequential reforming of acids and
alcohols often cause formation of alkanes as by-products. In addi-
tion, direct hydrogenation of COx formed can also give CH4 or high-
er alkanes through the Fischer–Tropsch reaction [13]. These
parallel routes lower the hydrogen yield, which is a serious
drawback.
Deactivation and catalyst instability are other issues during
APR. Activity loss during APR is often associated with loss of cata-
lytic sites (leaching, sintering) or blockage by carbonaceous spe-
cies, but specific details are not available in literature. However,
in the case of high temperature gas-phase steam reforming of oxy-
genates, reasons for catalyst deactivation were identified [14–20].
For example, coke deposition caused by oligomerization of olefins
formed was reported to be an issue during steam reforming of eth-
anol [14,15]. Formation of unsaturated CxHy species via dehydroge-
nation reactions is often proposed as a precursor to coke [16–18].
We have shown earlier that acetic acid is converted to acetone dur-
ing steam reforming, which further reacts to form coke precursors
⇑
Corresponding author.
E-mail address: k.seshan@utwente.nl (K. Seshan).
0021-9517/$ - see front matter Ó 2012 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.jcat.2012.05.019

240
D.J.M. de Vlieger et al. / Journal of Catalysis 292 (2012) 239–245
such as mesityl oxide or mesitylene [19]. The product CO can also
disproportionate exothermically to form carbon through the Bou-
douard reaction (2CO ? CO2 + C) [20].
(300 ꢁ 7.8 mm) column. An aqueous solution of 0.005 M H2SO4
was used as elutant phase and was flown at rate of
0.600 mL minꢀ1. The column was operated at 35 °C. Sample vol-
umes of 20 L were injected onto the column. Gas analysis was
carried out with a Varian CP-4900 Micro GC using MS5 and PPQ
columns. Relative errors of maximum 2% were introduced by the
GC measurements.
a
We reported recently that Pt/Al2O3 catalyst showed conversion
rates for APR of ethylene glycol that are very attractive for indus-
trial application [8]. However, this catalyst was found to deactivate
during the reaction. Pt/Al2O3 also gave lower H2 selectivity due to
the formation of alkanes. We showed further that the presence of
Ni in the Pt/Al2O3 catalyst suppressed the above problems such
as catalyst deactivation and high amounts of alkane formation
and provided the basis for the development of an efficient catalyst
[8].
In this study, the role of intermediates that are formed during
aqueous reforming of ethylene glycol on the formation of alkanes
is studied. Further, the reason for the loss of activity is investigated
in relation to intermediates formed during reactions. Correlations
thus drawn are discussed in relation to the promotion of Pt/Al2O3
with Ni. Improving catalyst stability and at the same time enhanc-
ing hydrogen selectivity at the expense of alkanes for the reform-
ing of ethylene glycol is essential in the design of an efficient
APR catalyst.
l
2.3. Definitions of conversion and selectivity
In this paper, carbon-to-gas conversion is defined as the per-
centage of carbon in the feed that is transformed to carbon in gas-
eous products (COx, hydrocarbons). Carbon removed from the
liquid phase can either go to gas phase products or coke. Determin-
ing the amount of carbon in gas could be done by combining the
GC results and gas production results. The ideal gas law can then
be applied to calculate the moles of carbon in the gas phase. This
method contains many parameters and assumptions that compro-
mise its accuracy. If coke formation does not occur, as was the case
for this study, then all carbon removed from the liquid phase gas-
ified. The carbon content in liquid streams can be determined very
accurately (relative error <1%) by the TOC analyzer, and therefore,
the following expression was used to calculate carbon to gas con-
2. Experimental
\moles C"Feedꢀ\moles C"Reactor
liquid effluent
version
ꢁ 100%.
\moles C"Feed
2.1. Catalyst preparation and characterization
Selectivity to carbon containing gas phase products (i = COx, CH4
\moles C"in species
\moles C"in gas
i
and C2+) was calculated according to selec:i% ¼
phase ꢁ
Experimental details regarding catalyst preparation and charac-
terization are discussed Table 1: Characterictics of the studied cat-
alystsin a previous publication [8]. Details of the catalysts used for
this study are summarized in Table 1. Metal loadings were deter-
100%. Side-reactions leading to carbon containing molecules in
the liquid phase were not taken into account in these selectivity
calculations.
mined using
a Philips X-ray fluorescence spectrometer (PW
3. Results and discussion
1480). Pt metal particle size was both determined by CO-chemi-
sorption and TEM imaging. FT-IR (Bruker Tensor 27, 4 cmꢀ1 resolu-
tion with 64 scans, range between 600 and 4000 cmꢀ1, 5 mm
aperture) and Raman (Bruker Senterra, laser wavelength
532 cmꢀ1 with 10 mW power, 3–5 cmꢀ1 resolution and range be-
tween 65 and 4000 cmꢀ1) spectra were recorded at room temper-
ature for fresh and used catalysts.
3.1. APR of ethylene glycol
Reforming of aqueous bio-organic streams is only interesting
for industrial applications when feeds with high oxygenate con-
centrations (>15 wt%) can be used. However, larger concentrations
of intermediates and products formed under these conditions can
lead to extensive secondary reactions. Thus, two sets of reforming
experiments were carried out: (1) 20 wt% ethylene glycol (EG)
solution was studied in supercritical water (450 °C and 250 bar)
and (2) a 1 wt% EG solution was studied at similar conversions ob-
tained by using milder reaction conditions (275 °C and 200 bar).
The obtained results are shown in Table 2. Conversion to gas-
and liquid-phase products is mentioned separately in Table 2 to al-
low a good comparison of the gasification efficiency. Total EG con-
version to liquid and gas products was 90% for both reactions
with the remaining carbon being unconverted EG. Carbon balances
were complete within an error margin of 5%. It can be seen that at
the higher EG concentration and higher temperature, the amount
of carbon containing products in the liquid phase increased. Alkane
formation is also higher under these conditions. The alkane forma-
tion was much lower in the experiment with 1 wt% EG. In both
cases, componential HPLC analysis of the liquid reactor effluent
showed that (in addition to unconverted EG) three other main
components were present in the reactor effluent. These were acetic
acid, methanol, and ethanol.
2.2. Catalytic testing
The experimental setup used is described in detail in an earlier
publication [8]. A 2 mL minꢀ1 flow of 1 wt% oxygenate solution
(ethylene glycol (P99.9%, Sigma–Aldrich), acetic acid (P99.9%,
Merck), or methanol (P99.9%, Merck), or ethanol (P99.9%, Merck))
was first preconditioned to the operating conditions (275 °C and
200 bar) before it entered the reactor in which 1.0 g of catalyst
was placed. The reforming reactions were studied during 5-h
time-on-stream. The reactor effluent was separated into gas and li-
quid products. An online Shimadzu TOC-VCSH analyzer deter-
mined carbon concentrations in the feed solution and the liquid
effluent. TOC analyses of the liquid phases are subjected to a rela-
tive error of 1%. A Shimadzu HPLC was used to identify the liquid
products formed during the reforming reactions. The HPLC was
equipped with a RID-10A detector and an Aminex HPX-87H
Table 1
Characteristics of the studied catalysts [8].
As discussed previously, components such as acids and alcohols
are often suggested as precursors for the formation of alkanes. Fur-
ther, we also stated that the addition of Ni to Pt/Al2O3 suppressed
the formation of alkanes during reforming of EG [8]. In order to
probe the role for Ni, a detailed componential HPLC analysis of
the liquid reactor effluents of 1 wt% EG reforming (275 °C and
200 bar) in the presence of Pt or Pt–Ni catalysts was carried out.
The results are shown in Fig. 1. It is seen from Fig. 1 that the reactor
Catalyst
Metal
loading by
XRF (wt%)
Average metal
particle
size (nm)
BET surface
area (m2/g)
c
-Al2O3
–
1.45
–
200
198
Pt/c-Al2O3
61 (TEM)
1.6 (CO-chemisorption)
61 (TEM)
Pt–Ni/
c
-Al2O3
1.15 Pt–0.35 Ni
198

D.J.M. de Vlieger et al. / Journal of Catalysis 292 (2012) 239–245
241
Table 2
Conversions and alkane selectivities for ethylene glycol reforming under different conditions in the presence of 1.5 wt% Pt/Al2O3.
EG (wt%)
Reaction Temp. (°C)
Reaction Press. (bar)
WHSV (hꢀ1
)
Conversion to gas (%)
Conversion to liquid (%)
Alkanes in gas (%)
20
1
450
275
250
200
12
1.2
74
78
15.4
11.7
16.9
0.3
effluent in presence of the Ni-promoted catalyst showed the same
by-products (acetic acid, methanol, and ethanol). The concentra-
tions of methanol and ethanol in the reactor effluent were not
influenced significantly by the presence of Ni. However, the con-
centration of acetic acid was found to be three times lower for
the Ni-promoted catalyst.
genate intermediates lead to longer residence times on the metal
and result in increased dehydrogenation activity. It would be use-
ful to estimate the energetics for the two routes (dehydrogenation
vs. isomerization) on Pt and Pt–Ni bimetallic surfaces by theory,
but that is beyond the scope of the current study. In addition, the
dissociation of the O–H bond is suggested to be the rate-limiting
step in the conversion of ethylene glycol [21]. The higher dehydro-
genation activity of the Pt–Ni catalyst results in faster dehydroge-
nation of the O–H groups and hence a higher overall reforming
rate. This is in agreement with earlier studies [8,25] where Pt–
Ni/Al2O3 catalysts showed higher conversions for the reforming
of ethylene glycol than Pt/Al2O3.
At this point, it is useful to recall the mechanistic sequences
proposed in the literature for the aqueous phase reforming of EG
by Dumesic et al. [13]. It is suggested, without experimental evi-
dence, that dehydrogenation of EG over Pt is the first step in the
reforming pathway. During this step, the hydrogen atoms from
the hydroxyl groups are removed by O–H cleavage on the Pt sur-
ꢀ
face [21]. In an optimal situation, the resulting adsorbed O–CH2–
CH2–Oꢀ species is further dehydrogenated and undergoes C–C
cleavage to form adsorbed CO [21]. The adsorbed intermediate spe-
cies ꢀO–CH2–CH2–Oꢀ can also undergo rearrangement over the
3.2. APR of model components
Aqueous phase reforming experiments with liquid products
formed during the APR of EG was carried out using individual com-
ponents (Fig. 2). Reforming experiments (with acetic acid, metha-
nol, ethanol) were conducted at 275 °C and 200 bar in the
presence of Pt/Al2O3 or Pt–Ni/Al2O3. Feed concentrations of 1 wt%
were used so as to work in the same concentration range as the li-
quid by-products formed during EG-reforming experiments. For
example, a total conversion of 15% to methanol, ethanol and acetic
acid was observed during the reforming of 20 wt% EG (Table 2),
resulting in a total concentration of 3 wt% liquid by-products. As
the three were in the same range, concentration of each was fixed
at 1 wt% for further experiments.
acidic
c-Al2O3 support to form acetic acid [13]. However, in an-
other study, we also observed the formation of acetic acid during
APR of EG using a carbon-supported Pt catalyst [22]. The chemical
inertness of this type of support material excludes the rearrange-
ꢀ
ment of O–CH2–CH2–Oꢀ over the support to form acetic acid. The
ionic nature of hot compressed water is reported to catalyze organ-
ic reactions in the aqueous phase, such as the Cannizzaro reaction
[23] and isomerization reactions [24]. As discussed later in this
manuscript, the Cannizzaro-derived reaction is not a dominant
contributor to acetic acid formation. The isomerization reaction
(either catalyzed by the support or by the ions from the liquid
ꢀ
water) of adsorbed O–CH2–CH2–Oꢀ is responsible for the majority
Blank reforming experiments (275 °C and 200 bar) with metha-
nol, ethanol, and acetic acid without catalysts showed conversions
below 5%. In the presence of Pt/Al2O3 and Pt–Ni/Al2O3 catalysts,
conversions were in the range of 50–80% (Fig. 2). For methanol,
the conversion was around 60% (Fig. 2A) and for ethanol higher,
80% (Fig. 2B). The higher reforming activity of ethanol on both
catalysts may be related to the reforming sequences that take
place. The initial step in the reforming of aliphatic alcohols in-
volves the dissociation of the O–H bond resulting in adsorbed
methoxy or ethoxy species on the Pt surface [26,27]. Similar disso-
ciation energies are reported [28] for breaking the O–H bond in
these alcohols and can therefore not be responsible for the differ-
ence in reforming activity. During reforming of alcohols, hydrogen
formation requires the formation of adsorbed carbon monoxide
which can undergo WGS to form H2 and CO2. In the case of meth-
anol to form CO, it is necessary for the adsorbed methoxy species
to undergo three C–H bond cleavages. For ethanol, the adsorbed
ethoxy species must undergo two C–H and one C–C bond cleavage.
Further it was suggested that the cleavage of the C–C bond in eth-
anol leads to an adsorbed methyl species [27]. We speculate that
the different routes to decompose the alcohols may be the reason
for the observed difference in reactivity. For the reforming of acetic
acid (Fig. 2C), initial conversion levels of 50% were obtained. How-
ever, the conversion decreased rapidly with time for both the Pt
and Pt–Ni catalysts, this will be discussed later on in the
manuscript.
of acetic acid formed.
Competition between dehydrogenation and isomerization of
adsorbed ꢀO–CH2–CH2–Oꢀ dictates the relative formation of hydro-
gen and acetic acid. Dehydrogenation will be catalyzed by metals
(Pt, Ni), and isomerization would be assisted by an acidic oxide
(in this case c-Al2O3).
The role of Ni in the suppression of acetic acid formation (Fig. 1)
can be inferred from DFT calculations that predicted ‘‘increased’’
binding energies for adsorbed species derived from ethylene glycol
on the surface monolayer Ni–Pt (111) as compared to Pt (111)
[21]. We suggest that the increased binding energies of the oxy-
3.3. Alkane formation during APR
The formation of alkanes during the catalytic reforming of ace-
tic acid, methanol and ethanol are shown in Fig. 3. The selectivity
toward alkanes (mainly methane) was found to be the highest
Fig. 1. Liquid by products during catalytic reforming of 1 wt% ethylene glycol at
275 °C and 200 bar (WHSV = 1.2 hꢀ1). Results for Pt and Pt–Ni catalysts showed
conversions to gas phase of 78% and 88%, respectively.

242
D.J.M. de Vlieger et al. / Journal of Catalysis 292 (2012) 239–245
Fig. 3. Formation of alkane during the catalytic reforming of 1 wt% acetic acid,
methanol, and ethanol at 275 °C and 200 bar. Conversions to gas phase (conv.) are
mentioned in the Figure.
activity for C–O bond breaking [13] during the reforming of meth-
anol can explain the observed differences for methane formation.
For methanol reforming with the Pt–Ni catalyst, almost no
methane was formed (Fig. 3). Methanol was found to be one of
the major products during the reforming of EG (Fig. 1) that ex-
plains why Pt–Ni/Al2O3 catalyst shows almost no methane forma-
tion in agreement with earlier studies [8,25]. The lower methane
production observed during methanol reforming in the presence
of Pt–Ni can be explained by the enhanced dehydrogenation activ-
ity of Pt–Ni catalyst [21], favoring C–H cleavage instead of C–O
cleavage in methanol. Promoting the Pt catalyst with Ni did not af-
fect the alkane (methane) selectivity for acetic acid or ethanol
reforming as can be expected from the reforming routes. Reform-
ing of these compounds inherently lead to CH3 species on the cat-
alyst surface, which is a precursor for methane formation. In case
of methanol reforming, the intermediate CH3O must undergo C–
O cleavage to form CH3 (and hence methane). Methane formation
during methanol reforming can be prevented by avoiding C–O
cleavage.
3.4. Catalyst deactivation
Reforming experiments with 1 wt% methanol (Fig. 2A) and eth-
anol (Fig. 2B) solutions showed stable activity levels for both Pt and
Pt–Ni catalysts. Catalysts used for acetic acid reforming deacti-
vated during the reaction (Fig. 2C). Both catalysts showed similar
deactivation rates. The catalysts lost all activity after 3 h on stream
with final conversion levels (5%) being similar to the reforming
experiment without a catalyst.
Deactivation of the catalyst with acetic acid is commonly ob-
served in reforming reactions [19]. The formation of methane
resulting from acetic acid decomposition is reported in literature
to be a possible cause for the deactivation of catalysts [30]. It is
suggested that C–H bond breaking in methane further results in
CHx species (1 6 x 6 2) on the catalytic surface. These fragments
can further oligomerize to coke species [19]. However, catalysts
used for ethanol reforming were found to be stable while also pro-
ducing high amounts of methane (20%). It is also known that Pt and
Ni catalysts are stable for the reforming of methane. Therefore, it is
concluded that methane-induced deactivation is unlikely, and the
dominant deactivation pathway is due to other causes.
Fig. 2. Conversion to gas phase for reforming (275 °C and 200 bar) of 1 wt%
methanol (A), ethanol (B), or acetic acid (C) using 1.5 wt% Pt/Al2O3 (s) and 1.15–
0.35 wt% Pt–Ni/Al2O3 (j) catalysts.
( 47%) during acetic acid reforming. For the reforming of methanol,
a selectivity of 8% toward CH4 was observed for the Pt catalyst. In
the case of ethanol reforming, the Pt catalyst showed a CH4 selec-
tivity of 20%. The higher amounts of methane observed during
ethanol reforming is attributed probably to the CHx fragment
formed by C–C bond breaking of ethanol [26] on the Pt surface.
CHx fragments can recombine with adsorbed H atoms on the Pt
surface to form methane. A similar sequence is also suggested for
acetic acid on Pt-based catalysts [19,29]. In the case of methanol,
C–O bond breaking is necessary to generate these adsorbed CHx
species that lead to the production of methane. High C–C bond–
breaking activities for Pt during the reforming of ethanol and low
Further, liquid products formed during the APR of acetic acid,
methanol, and ethanol were determined to find any role for them
in catalyst deactivation. During acetic acid reforming, a conversion
of 7% to liquid by-products was observed for the Pt catalyst. By-

D.J.M. de Vlieger et al. / Journal of Catalysis 292 (2012) 239–245
243
products were identified as formaldehyde and iso-propanol. For
ethanol, a conversion of 6% to liquid by-products was found for
the Pt catalyst. The main product was acetic acid, and the remain-
ing was acetaldehyde (<0.5%). The Cannizzaro reaction could be
responsible for some of the acetic acid formed during ethanol
reforming. The Pt and Pt–Ni catalysts studied did not influence
the amount of acetic acid, and acetaldehyde formed during ethanol
reforming, while during EG reforming, the amount of acetic acid
was strongly dependent on the type of catalyst. This shows that
the Cannizzaro-derived reaction is not a dominant contributor to
acetic acid formation during EG reforming.
No liquid by-products were observed during methanol reform-
ing. Interestingly, the presence of Ni in the Pt–Ni catalyst did not
show any significant differences on the type or amount of second-
ary liquid by-products during the reforming of acetic acid, metha-
nol, or ethanol compared to the mono-metallic Pt catalyst. No
apparent conclusions can thus be derived from these experiments.
In the case of the Pt/Al2O3 catalyst that deactivated during ace-
tic acid reforming, CO-chemisorption experiments showed a Pt dis-
persion of only 0.03% (fresh catalyst 67%), indicating that almost all
accessible Pt surface area was lost. Loss of catalytic metal by leach-
ing, metal particle growth by sintering and coke formation are
common causes for the loss of catalytic surface and hence the deac-
tivation of catalysts. XRF elemental analysis of the deactivated Pt/
Al2O3 catalyst used for acetic acid reforming showed that no Pt was
lost by leaching from the catalyst. Sintering is a thermo-physical
effect and therefore dictated by the operating conditions (temper-
ate and pressure). Catalyst stability was not affected during the
reforming of methanol and ethanol when the reaction was per-
formed at the similar operating conditions as for acetic acid
reforming, thus excluding sintering as cause of catalyst deactiva-
tion. The formation of coke on the surface of Pt based catalysts dur-
ing steam reforming of acetic acid is often reported [19]. Carbon
deposits on the deactivated Pt/Al2O3 catalyst that was used for
aqueous reforming of acetic acid was studied by TGA-MS up to a
temperature of 800 °C in oxygen (1 v/v% O2 in N2). No weight loss
attributed to the removal of carbonaceous deposits was observed
during the experiment, indicating that carbonaceous deposits were
not present on the catalyst. IR and Raman experiments (discussed
below) also did not show presence of any carbonaceous species. In
addition, the regeneration of the deactivated Pt catalyst used for
acetic acid reforming was attempted by oxygen treatment (up to
500 °C), but the treated catalyst did not regain the activity for
the reforming of EG. Based on these results, carbonaceous deposits
as cause for catalyst deactivation can also be ruled out. Under con-
ventional gas-phase-reforming conditions, it is well accepted that
steam reforming of acetic acid can lead to coke formation via inter-
mediates involving acetone, diacetone alcohol, and mesitylene
[19,31,32]. Hot-compressed water has a high tendency to dissolve
organic species, including carbonaceous deposits [3,33]. Carrying
out reforming in hot-compressed water therefore plays an impor-
tant role in keeping the catalytic surface clean of any coke deposits
that prevent accessibility of oxygenates to catalytic sites.
Fig. 4. Raman spectra of spent 1.5 wt% Pt/Al2O3 catalysts used for reforming (275 °C
and 200 bar) of acetic acid, methanol, and water only.
Fig. 5A. FT-IR spectra of spent 1.5 wt% Pt/Al2O3 catalysts used for the reforming
(275 °C and 200 bar) of acetic acid, methanol, and water only.
vibrations of O–H bonds [34]. A weak band at 3565 cmꢀ1 is visible
in the spectra of the acetic acid reforming catalyst, and this band
can be attributed to hydroxyl groups with a higher acidity. Inter-
estingly, the ratio between the normalized bands at 362 cmꢀ1
and the OH bands at 3076 and 3212 cmꢀ1 show clearly that more
hydroxyl species are present in case of the deactivated Pt catalyst
compared with the active catalysts.
The differences in hydroxyl groups and the cause for deactiva-
tion of the Pt catalyst were further investigated by Fourier-Trans-
formed Infrared spectroscopy (FT-IR). FT-IR spectra of spent Pt
catalysts were taken at room temperature in air. Spectra were nor-
malized for sample weight and are shown in the frequency range of
1000–2400 cmꢀ1 in Fig. 5A. No information could be derived from
the IR spectrum above 2400 cmꢀ1 due to the broad and intensive
molecular water band that covered the whole range up to the max-
imum of 4000 cmꢀ1. For all catalysts, a broad band in the frequency
region 1000–1200 cmꢀ1 was observed which can be assigned to
the Al–OH bending vibrations in the Boehmite crystal structure
[35]. Multiple weak bands are observed in the region 1300–
1900 cmꢀ1, and these are attributed to different kinds of adsorbed
carbon pollutants (i.e., CO2 adsorption from air leading to surface
carbonates) and water adsorption. Two strong bands appeared at
1975 and 2112 cmꢀ1, and these have been suggested to originate
from complex Al–O–H zigzag structures in Boehmite [35].
Raman spectroscopy was used to study spent Pt catalysts used
for reforming of methanol, acetic acid, and water only (blank
experiment). Raman spectra were taken at room temperature in
air. The obtained spectra are shown in Fig. 4. The spectra were nor-
malized based on the most pronounced Raman band at 362 cmꢀ1
for easy comparison. In all three cases, Raman spectra of the used
catalysts showed five sharp adsorption bands in the range of 65–
4000 cmꢀ1. These bands are attributed to Boehmite [34,35], which
is formed as a result of the phase change of the c-Al2O3 support in
our catalyst when subjected to hot-compressed water [8,36]. The
three bands at 362, 494, and 672 cmꢀ1 are attributed to vibrational
modes of the Boehmite Al–O bond [34]. The two other adsorption
bands, located at 3212 and 3076 cmꢀ1, are attributed to stretching

244
D.J.M. de Vlieger et al. / Journal of Catalysis 292 (2012) 239–245
responsible for the dissolution of Al2O3 [38]. The increase in hydro-
xyl groups on the surface of the catalysts used for acetic acid
reforming indicates that the surface structure of Al2O3 is affected
by acetic acid. Dissociation constants for acids are much higher
in hot-compressed water compared with normal water [33], and
therefore, we propose that acetic acid behaves as a stronger acid
in hot-compressed water resulting in protonation and sequential
leaching of the alumina surface. In hot-compressed water, where
solubility of inorganic oxide materials is low, it can be expected
that hydroxylated alumina that leached off can be re-deposited
on the catalyst. Fig. 6 shows the TEM image of the Pt-deactivated
catalyst. The Pt particle and the lattice distances of the Al2O3 are
clearly visible. The Al2O3 lattice distance is measured in the picture
to be 0.7 nm. The Pt particle is covered by a layer that shows the
same lattice distances of 0.7 nm as alumina, indicating that the
Pt is covered by alumina. From these results, we conclude that
the cause of deactivation of the Pt catalyst is coverage of catalytic
Pt sites by hydroxylated alumina layer.
Fig. 5B. Difference FT-IR spectra of spent Pt catalysts.
Although the exact mechanism responsible for coverage of the
Pt particle by alumina is not yet known, we show here that this
mechanism likely involves hydroxylation of the
and is propagated by a strong acidic environment, resulting from
c-Al2O3 support
FT-IR spectra of the spent catalysts were subtracted from each
other as shown in Fig. 5B to study the differences in chemical
groups on the surfaces of the used Pt catalysts. Subtracting the
spectrum of the catalyst that was exposed to water only from that
of aqueous methanol solution resulted in a more or less flat line,
indicating that the surface composition of the active catalysts
(used for methanol and water only) are similar. Subtracting the
spectrum of the catalyst subjected to only water from the spent
deactivated acetic acid catalyst resulted in 3 strong bands; one lat-
tice vibration in the region 1000–1200 cmꢀ1 and two sharp OH re-
lated bands located at 1975 and 2112 cmꢀ1. Also in the Raman
spectra, the acetic acid treated sample showed the highest O–H
intensity (Fig. 4). These results indicate the presence of a highly
hydroxylated type of Boehmite in the acetic acid deactivated cata-
lyst compared with the active catalyst used in methanol APR.
A cause for the deactivation of the catalyst can be inferred from
the increased number of hydroxyl groups. It is known that the sta-
bility of the Al2O3 surface strongly depends on the pH of the med-
ium [37]. Acids are reported to cause dissolution of Al2O3 [38], and
this is an important process that can affect catalyst properties. The
dissolution rate of metal oxide supports increases at lower pH val-
ues and involves protonation of Al–O–Al bonds to form Al–OH. Fur-
ther protonation and/or rearrangement of oxide ligands are
the high proton dissociation of acids in hot-compressed water.
3.5. Reaction pathways
The reaction pathways involved in the reforming of ethylene
glycol proposed by Dumesic et al. [13] have been adapted with
the new insights obtained from this study. The resulting-extended
reaction sequences are proposed in Fig. 7. Modifications include
formation of methanol from ethanol and deactivation by acetic
acid. We have identified methanol as an important liquid by-prod-
uct. Methanol can be formed during the reforming of ethanol as a
result of C–C bond cleavage. Reforming of methanol can either lead
to methane or to hydrogen, and the preferential route is controlled
by the dehydrogenation activity of the catalyst. The formation of
acetic acid, ethanol, and methanol during the reforming of EG is
considered undesired because consecutive reactions involving
these compounds are responsible for the formation of alkanes
(mainly methane). In addition, the presence of acetic acid can
cause the deactivation of the catalyst. The formation of liquid by-
products and alkanes mainly result from C–O bond cleavage, dehy-
dration, and hydrogenation reactions. The pathway to high hydro-
Fig. 6. TEM image of an alumina covered Pt particle.

D.J.M. de Vlieger et al. / Journal of Catalysis 292 (2012) 239–245
245
Fig. 7. Schematic representation of modified reaction pathways during aqueous reforming of ethylene glycol.
[6] E.C. Vagia, A.A. Lemonidou, Int. J. Hydrogen Energy 32 (2007) 212.
[7] R.R. Davda, J.A. Dumesic, Angew. Chem. Int. Ed. 42 (2003) 4068.
[8] D.J.M. de Vlieger, A.G. Chakinala, L. Lefferts, S.R.A. Kersten, K. Seshan, D.W.F.
Brilman, Appl. Catal. B 111–112 (2012) 536.
[9] J.W. Shabaker, G.W. Huber, J.A. Dumesic, J. Catal. 222 (2004) 180.
[10] A.O. Menezes, M.T. Rodrigues, A. Zimmaro, L.E.P. Borges, M.A. Fraga, Renew.
Energy 36 (2011) 595.
[11] B. Roy, K. Loganathan, H.N. Pham, A.K. Datye, C.A. Leclerc, Int. J. Hydrogen
Energy 35 (2010) 11700.
[12] A. Wawrzetz, B. Peng, A. Hrabar, A. Jentys, A.A. Lemonidou, J.A. Lercher, J. Catal.
269 (2010) 411.
[13] R.R. Davda, J.W. Shabaker, G.W. Huber, R.D. Cortright, J.A. Dumesic, Appl. Catal.
B 56 (2005) 171.
gen yields involves dehydrogenation of ethylene glycol. Competi-
tion between dehydrogenation and isomerization of the adsorbed
intermediate to acetic acid determines the selectivity toward al-
kanes (and hence H2 yields). Addition of Ni to alumina-supported
Pt catalysts increased the dehydrogenation activity. As a conse-
quence, a decrease in the formation of alkanes and improved sta-
bility was accomplished by suppressing acetic acid formation.
4. Conclusions
[14] P.D. Vaidya, A.E. Rodrigues, Chem. Eng. J. 117 (2006) 39.
[15] A.L. Alberton, M.M.V.M. Souza, M. Schmal, Catal. Today 123 (2007) 257.
[16] D.L. Trimm, Catal. Today 37 (1997) 233.
[17] F. Frusteri, S. Freni, V. Chiodo, L. Sparado, G. Bonura, S. Cavallaro, J. Power Sour.
132 (2004) 139.
We identified methanol, ethanol, and acetic acid as the main li-
quid by-products during the reforming of ethylene glycol in hot-
compressed water in the presence of Al2O3-supported Pt and Pt–
Ni catalysts. Side-reactions involving these liquid by-products lead
to the formation of high amounts of methane. Acetic acid was
shown to be responsible for the deactivation of Pt and Pt–Ni cata-
lysts by hydroxylation of the Al2O3 surface. Re-deposition of the
dissolved alumina on the catalyst leads to the blocking of catalytic
Pt sites and hence the deactivation of the catalyst. The increased
dehydrogenation activity of the Pt–Ni catalyst was found to sup-
press the formation of acetic acid during ethylene glycol reforming
and thereby increasing the H2 selectivity and catalyst lifetime.
[18] V. Klouz, V. Fierro, P. Denton, H. Katz, J.P. Lisse, S. Bouvot-Mauduit, C.
Mirodatos, J. Power Sour. 105 (2002) 26.
[19] K. Takanabe, K. Aika, K. Seshan, L. Lefferts, Chem. Eng. J. 120 (2006) 133.
[20] K.O. Christensen, D. Chen, R. Lodeng, A. Holmen, Appl. Catal. A 314 (2006) 9.
[21] O. Skoplyak, M.A. Barteau, J.G. Chen, Surf. Sci. 602 (2008) 3578.
[22] D.J.M. de Vlieger, D.B. Thakur, L. Lefferts, K. Seshan, Carbon nanotubes: A
promising catalyst support material for supercritical water gasification of
biomass waste, ChemCatChem, Accepted for publication, doi: 10.1002/cctc.
201200318.
[23] M. Osada, M. Watanabe, K. Sue, T. Adschiri, K. Arai, J. Supercrit. Fluids 28
(2004) 219.
[23] M. Osada, M. Watanabe, K. Sue, T. Adschiri, K. Arai, J. Supercrit. Fluids 28
(2004) 219.
Acknowledgments
[24] J. An, L. Bagnell, T. Cablewski, C.R. Strauss, R.W. Trainor, J. Org. Chem. 62 (1997)
2505.
[25] G.W. Huber, J.W. Shabaker, S.T. Evans, J.A. Dumesic, Appl. Catal. B 62 (2006)
226.
[26] A. Erdohelyi, J. Rasko, T. Kecskes, M. Toth, M. Dömök, K. Baan, Catal. Today 116
(2006) 367.
[27] S. Sa, H. Silva, L. Brandao, J.M. Sousa, A. Mendes, Appl. Catal. B 99 (2010) 43.
[28] S.J. Blanksby, G.B. Ellison, Acc. Chem. Res. 36 (2003) 255.
[29] B. Matas Guell, I. Babich, K. Seshan, L. Lefferts, J. Catal. 257 (2008) 229.
[30] M.M.V.M. Souza, M. Schmal, Appl. Catal. A 281 (2005) 19.
[31] X. Hu, G. Lu, Appl. Catal. B 99 (2010) 289.
[32] A.C. Basagiannis, X.E. Verykios, Appl. Catal. B 82 (2008) 77.
[33] P.E. Savage, J. Supercrit. Fluid. 47 (2009) 407.
[34] H.D. Ruan, R.L. Frost, J.T. Kloprogge, J. Raman Spectrosc. 32 (2001) 745.
[35] A.B. Kiss, G. Keresztury, L. Farkas, Spectrochim. Acta A 36 (1980) 653.
[36] R.M. Ravenelle, J.R. Copeland, W.-G. Kim, J.C. Crittenden, C. Sievers, ACS Catal. 1
(2011) 552.
This project was supported by ACTS (Project Number
053.61.023). The authors greatly acknowledge Ing. Louise Vrielink
for BET and XRF analysis, Dr. Rico Keim for TEM-EDX imaging, Kar-
in Altena for dispersion measurements, Ing. Benno Knaken and Ing.
Bert Geerdink for technical support.
References
[1] A. Tanksale, J.N. Beltramini, G.M. Lu, Renew. Sustain. Energy Rev. 14 (2010)
166.
[2] R.D. Cortright, R.R. Davda, J.A. Dumesic, Nature 418 (2002) 964.
[3] Y. Guo, S.Z. Wang, D.H. Xu, Y.M. Gong, H.H. Ma, X.Y. Tang, Renew. Sustain.
Energy Rev. 14 (2010) 334.
[4] L. Garcia, R. French, S. Czernik, E. Chornet, Appl. Catal. A 201 (2000) 225.
[5] S. Adhikari, S. Fernando, F. To, R. Bricka, P. Steele, A. Haryanto, Energy Fuel 22
(2008) 1220.
[37] J.A. Yopps, D.W. Feurstenau, J. Colloid Sci. 19 (1964) 61.
[38] B. Kasprzyk-Hordern, Adv. Colloid. Interfac. 110 (2004) 19.