Aqueous-phase reforming of ethylene glycol to hydrogen on Pd/Fe 3O4 catalyst prepared by co-precipitation: Metal-support interaction and excellent intrinsic activity

Article abstract of DOI:10.1016/j.jcat.2010.07.014

A high-performance Pd/Fe₃O₄ catalyst for aqueous-phase reforming (APR) of ethylene glycol (EG) to H₂ was prepared facilely by the co-precipitation method. After proper activation, the Pd was present as highly dispersed metallic nanoparticles with dimension of <3 nm, and the Fe was present as magnetite. When compared to Pd catalyst supported on Fe₂O₃, NiO, Cr₂O₃, Al₂O₃, or ZrO₂ prepared by incipient wetness impregnation, the Pd/Fe₃O₄ catalyst displayed superior catalytic performance in terms of activity, selectivity, and stability. The intrinsic activity of the Pd/Fe₃O₄ catalyst was about three times of that of the second most active Pd/Fe₂O₃ catalyst under the same reaction conditions. In addition, the Pd/Fe ₃O₄ catalyst retained ～80% of its initial activity after reaching the steady-state. Notably, the Pd/Fe₃O₄ catalyst possessed the highest turnover frequency of H₂ (109 min ^-1) reported so far, showing its promise as a new practical catalyst for APR of biomass-derived oxygenates to H₂. The excellent catalytic performance of the Pd/Fe₃O₄ catalyst was attributed to the enhanced synergistic effect between small Pd nanoparticles and magnetite in promoting the water-gas shift reaction, the rate-determining step in APR of EG over Pd-based catalysts.

Full text of DOI:10.1016/j.jcat.2010.07.014

Journal of Catalysis 274 (2010) 287–295
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Aqueous-phase reforming of ethylene glycol to hydrogen on Pd/Fe3O4 catalyst
prepared by co-precipitation: Metal–support interaction and excellent
intrinsic activity
Jun Liu a, Bo Sun a, Jiye Hu a, Yan Pei a, Hexing Li b, Minghua Qiao a,
*
a Department of Chemistry and Shanghai Key Laboratory of Molecular Catalysis and Innovative Materials, Fudan University, Shanghai 200433, PR China
b Department of Chemistry, Shanghai Normal University, Shanghai 200234, PR China
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 14 May 2010
Revised 9 July 2010
Accepted 14 July 2010
Available online 11 August 2010
A high-performance Pd/Fe3O4 catalyst for aqueous-phase reforming (APR) of ethylene glycol (EG) to H2
was prepared facilely by the co-precipitation method. After proper activation, the Pd was present as
highly dispersed metallic nanoparticles with dimension of <3 nm, and the Fe was present as magnetite.
When compared to Pd catalyst supported on Fe2O3, NiO, Cr2O3, Al2O3, or ZrO2 prepared by incipient
wetness impregnation, the Pd/Fe3O4 catalyst displayed superior catalytic performance in terms of activ-
ity, selectivity, and stability. The intrinsic activity of the Pd/Fe3O4 catalyst was about three times of that of
the second most active Pd/Fe2O3 catalyst under the same reaction conditions. In addition, the Pd/Fe3O4
catalyst retained ꢀ80% of its initial activity after reaching the steady-state. Notably, the Pd/Fe3O4 catalyst
possessed the highest turnover frequency of H2 (109 minꢁ1) reported so far, showing its promise as a new
practical catalyst for APR of biomass-derived oxygenates to H2. The excellent catalytic performance of the
Pd/Fe3O4 catalyst was attributed to the enhanced synergistic effect between small Pd nanoparticles and
magnetite in promoting the water–gas shift reaction, the rate-determining step in APR of EG over Pd-
based catalysts.
Keywords:
APR
Pd/Fe3O4
Co-precipitation
Ethylene glycol
H2
Ó 2010 Elsevier Inc. All rights reserved.
1. Introduction
The mechanism of H2 production from APR of EG has been de-
scribed in the literature [10]. According to that mechanism, a good
Hydrogen is a clean fuel that emits only water when combusted
or used in polymer electrolyte membrane fuel cells. However, the
production of H2 nowadays is far from clean. Approximately 95%
of H2 is produced from carbonaceous fossil fuels, which inevitably
results in tremendous CO2 emissions. Greener production routes of
H2 from water by solar- or wind-driven electrolysis and photo-bio-
logical water splitting, however, are less competitive due to the
present-day high costs to utilize these renewable energies [1–3].
Therefore, much attention has been paid to the production of H2
from biomass, which in principle leads to net zero CO2 emissions
[4]. Among these endeavors, Dumesic and coworkers demon-
strated that the catalytic aqueous-phase reforming (APR) of bio-
mass-derived oxygenates such as glucose, sorbitol, glycerol, and
ethylene glycol (EG) is thermodynamically and kinetically feasible
in generating H2-rich fuel gas [5–8]. The APR process is promising
owing to its advantages of higher energy efficiency and lower CO
concentration in the product gas than the conventional steam
reforming processes [9].
APR catalyst should not only be active for the cleavage of the C–C
bond but also be active for the water–gas shift (WGS) reaction [11].
Moreover, the catalyst should be inert to parallel and series com-
peting reactions such as the cleavage of the C–O bond and the
methanation reaction which greatly deteriorate the productivity
of H2. Based on the works of Sinfelt and Yates [12] and Grenoble
et al. [13], Davda et al. deduced that Pt and Pd could meet these cri-
teria and would be promising catalysts for APR of EG to H2 [7]. It
was verified that the Pt/Al2O3 catalyst is both active and selective
in APR of EG to H2 [5]. On the other hand, although the Pd/SiO2 cat-
alyst is also highly selective to H2, its activity expressed in terms of
the rate of CO2 production is about one order of magnitude inferior
to that of the Pt/SiO2 catalyst [7]. Since the rate of the WGS reac-
tion over Pd catalysts is lower than that over Pt catalysts, while
the rate of C–C bond cleavage in ethane hydrolysis is similar for
Pd- and Pt-based catalysts, Huber et al. deduced that the rate-
determining step for APR of EG on Pd could be the WGS reaction
[9].
The activity of the supported catalysts critically depends both
on the metal and on the support itself [14], and high-performance
catalyst can be designed by taking into account the significant
contribution of the support [15,16], which is also true for the
* Corresponding author. Fax: +86 21 65641740.
E-mail address: mhqiao@fudan.edu.cn (M. Qiao).
0021-9517/$ - see front matter Ó 2010 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2010.07.014

288
J. Liu et al. / Journal of Catalysis 274 (2010) 287–295
APR catalyst. By supporting Pd on Fe2O3 (Nanocat) using incipient
wetness impregnation (IWI), Huber et al. found that the production
rates of CO2 and H2 were surprisingly elevated to about 10 times
higher than those on the Pt/Al2O3 catalyst [9]. Because Fe2O3 did
not promote the WGS activity of Pt [9] and vice versa [17], and
Pd/Fe2O3 catalysts were several orders of magnitude more active
than the Pd/SiO2 catalyst for WGS [18], Huber et al. concluded that
the high activity of the Pd/Fe2O3 catalyst in APR of EG was caused
by the synergistic combination of Pd and Fe2O3 which boosts the
WGS reaction [9]. However, it is unfortunate that the Pd/Fe2O3 cat-
alyst lost 50% of its activity after being heated to 483 K under APR
reaction conditions.
Enlightened by these works, we suggest that if a more intimate
interaction between Pd and iron oxide can be achieved, in principle
the synergistic effect could be enhanced further, and an even more
active and stable iron oxide-supported Pd catalyst could be devel-
oped. Following this idea, in this work we prepared a Pd/Fe3O4 cat-
alyst using the co-precipitation method which usually leads to a
better interaction between metal and support than the impregna-
tion method [19–22] and evaluated its catalytic performance in
APR of EG to H2. Although the APR performance of the Pd/Fe2O3
catalyst prepared by IWI has been investigated previously [9], this
catalyst was not fully characterized. So, in the present work we
also prepared the Pd/Fe2O3 catalyst by IWI and systematically
characterized and evaluated this catalyst for the purpose of com-
parison. The excellent catalytic performance of the Pd/Fe3O4 cata-
lyst was discussed and correlated with the characterization results.
radiation (0.15418 nm). The X-ray tube was operated at 40 kV
and 40 mA. The mean crystallite size of Fe3O4 was calculated from
the full width at the half maximum (FWHM) of the Fe3O4(3 1 1)
diffraction peak according to the Scherrer equation.
Temperature programmed reduction (TPR) was carried out on a
home-made apparatus. About 50 mg of the catalyst precursor was
degassed at 473 K under Ar for 2 h. After cooling down to room
temperature in the blanket of Ar, the gas was switched to 5 vol.%
H2/Ar (50 ml minꢁ1), and the sample was heated to 873 K at a
ramping rate of 10 K minꢁ1. The amount of H2 consumed was mon-
itored by a thermal conductivity detector (TCD).
The Brunauer–Emmett–Teller (BET) surface area (SBET) and
porosity were acquired by N2 physisorption at 77 K on a Micro-
meritics TriStar3000 apparatus. The microstructure was observed
on a JEOL JEM 2011 transmission electron microscope (TEM) oper-
ating at 200 kV.
The active surface area (Sact) was measured based on CO desorp-
tion. To remove H2 adsorbed on the catalyst reduced in H2/Ar, the
catalyst was heated at 498 K in Ar (deoxygenated by an Alltech
Oxy-trap filter) for 1 h. After the catalyst was cooled down to room
temperature under Ar, CO pulses were injected until the eluted
peak areas of consecutive pulses were constant. The maximum
desorption temperature, 973 K, was achieved at a heating rate of
10 K minꢁ1. Sact was calculated from the volume of CO desorbed
by assuming CO/Pds stoichiometry of 1 and a surface area of
7.87 ꢃ 10ꢁ20 m2 per Pd atom [24].
The surface chemical state was determined by X-ray photoelec-
tron spectroscopy (XPS, Perkin–Elmer PHI5000C) using Mg K
a
radiation (1253.6 eV). The catalyst covered by ethanol was
mounted on the sample plate, degassed in the pretreatment cham-
ber at 393 K for 4 h in vacuo, and then transferred to the analyzing
chamber where the background pressure was <2 ꢃ 10ꢁ9 Torr. All
binding energy (BE) values were calibrated by C 1s peak of contam-
inant carbon at 284.6 eV with an uncertainty of 0.2 eV.
2. Experimental
2.1. Catalyst preparation
The Pd/Fe3O4 catalyst was prepared by the co-precipitation
method [23]. Under stirring, 4.1 g of Fe(NO3)3ꢂ9H2O was dissolved
in 14 ml of PdCl2 solution (3.2 ꢃ 10ꢁ2 M) at room temperature.
Then, 30 ml of Na2CO3 solution (1.0 M) was added dropwise to
the mixed solution, and the pH of the final solution was ca. 8.5. After
stirring and standing for 3 h respectively, the resulting precipitate
was filtered, washed with deionized water several times, air-dried
at 373 K overnight, and calcined in air at 573 K for 3 h. The resulting
brown powder was denoted as Pd/Fe(OH)x. Prior to activity testing
and characterizations, the Pd/Fe(OH)x precursor was reduced in
5 vol.% H2/Ar (50 ml minꢁ1) at 473 K for 3 h at a heating rate of
1 K minꢁ1. The as-reduced catalyst was denoted as Pd/Fe3O4.
The Pd/Fe2O3 catalyst was prepared by the IWI method. Fe2O3
was prepared by thermal decomposition of Fe(NO3)3ꢂ9H2O at
673 K. Then, 4.2 ml of PdCl2 solution (0.11 M) was impregnated
onto 1.0 g of the as-prepared Fe2O3 and stayed at room tempera-
ture overnight, followed by drying at 373 K overnight and calcina-
tion at 573 K for 2 h. The Al2O3- (Shanghai Super), ZrO2-, Cr2O3-,
and NiO-supported (Shanghai Chemical Reagents) Pd catalysts
were also prepared by IWI to expand our study on the effect of sup-
port on the APR performance of Pd. These catalyst precursors were
reduced similarly to that of Pd/Fe3O4. The nominal Pd loading was
fixed at 5.0 wt.% on all the supports, and the practical Pd loading in
all the reduced catalysts was in the range of 4.7 0.1 wt.%, as
determined by inductively coupled plasma–atomic emission spec-
troscopy (ICP–AES, IRIS Intrepid). The amount of Na retained in the
Pd/Fe3O4 catalyst was ꢀ1 ppm. So, the influence of Na on the reac-
tivity of the Pd/Fe3O4 catalyst should be minimal, if any.
2.3. Activity testing
The reactor system for APR of EG was established according to
the set-up described by Shabaker et al. [10]. The catalyst precursor
(ꢀ1.30 g for Pd/Fe(OH)x and 1.06 g for others) was loaded in the
stainless steel tubular reactor (i.d. 6 mm) and reduced on site with
5 vol.% H2/Ar at 473 K for 3 h, followed by cooling down to 303 K
under the same atmosphere before exposure to an aqueous solu-
tion of 5.0 wt.% EG. Argon was used to regulate the system pres-
sure. The reforming was typically conducted with catalyst
containing 0.05 g of Pd, weight hourly space velocity (WHSV
(weight flow rate of the feed solution) ꢃ (weight fraction of EG in
the feed)/(weight of Pd in the catalyst)) of 3.6 hꢁ1, temperature
of 498 K, and system pressure of 2.58 MPa, unless otherwise spec-
ified. The stability test was conducted under above reaction condi-
tions up to 130 h on stream.
During the reaction, the gas products were analyzed by an on-
line gas chromatograph (GC122). H2, CO, CH4, and CO2 were sepa-
rated by a 2-m TDX-01 packed column and examined by TCD.
Methane, ethane, propane, and butane were separated by a DM
PoraPlot Q capillary column (10 m ꢃ 0.53 mm ꢃ 20
lm) and exam-
ined by a flame ionization detector (FID). Liquid phase effluent was
condensed and analyzed gas chromatographically using the same
capillary column and FID detector. The liquid products were also
qualified by GC-MS (Finnigan Voyager) fitted with an HP-5 capil-
lary column. The carbon balance was within 5% for all the cata-
lytic runs, indicating negligible carbon deposition on the Pd/
Fe3O4 catalyst.
2.2. Catalyst characterization
Powder X-ray diffraction (XRD) was executed on a Bruker AXS
According to Shabaker et al. [25], the selectivity to H2 was de-
fined as:
D8 Advance X-ray diffractometer in a step mode using Cu K
a

J. Liu et al. / Journal of Catalysis 274 (2010) 287–295
289
H2 selectivity ð%Þ ¼ ½ðmoles of H2 producedÞ=
ðmoles of C in gas phaseÞꢄ ꢃ ð2=5Þ ꢃ 100
The XRD pattern of the Pd/Fe2O3 catalyst prepared by IWI
clearly presents diffraction peaks at 2h of 24.0°, 33.1°, 35.5°,
40.8°, 49.5°, 53.9°, 62.4°, and 64.0° ascribable to hematite. How-
ever, the appearance of small features at 2h of 30.0°, 43.0°, and
56.9° and the lowered (1 0 4)/(1 1 0) and improved (2 1 4) to
(3 0 0) intensity ratios with respect to those of the standard diffrac-
togram of hematite strongly indicate the presence of a small
amount of magnetite in this catalyst, although the H2–TPR result
presented below showed that this reduction temperature was
not sufficient to transform bulk hematite to magnetite. Compared
with the XRD pattern of Fe2O3 without supporting of Pd, the stron-
gest (1 1 1) reflection of metallic Pd at 2h of ca. 40.2° was readily
discerned on this catalyst, which evidences that the crystallite size
of metallic Pd was larger than that in the Pd/Fe3O4 catalyst, infer-
ring that the interaction between Pd and iron oxide in the Pd/Fe3O4
catalyst prepared by co-precipitation is much stronger than that in
the Pd/Fe2O3 catalyst prepared by IWI.
which takes into account of the occurrence of the WGS reaction;
and the selectivity to alkane was defined as:
Alkane selectivity ð%Þ ¼ ½ðmoles of C in gaseous alkanesÞ=
ðtotal moles of C in gas productsÞꢄ ꢃ 100
According to these definitions, the summation of the selectivi-
ties to H2 and alkane does not lead to unity, for they were calcu-
lated based on independent hydrogen and carbon balances,
respectively.
3. Results and discussion
3.1. Physicochemical properties
The H2–TPR profiles of Pd/Fe(OH)x and Pd/Fe2O3 precursors are
shown in Fig. 2. For comparison, the reduction profiles of hematite
without supporting of Pd as well as Fe(OH)x are also presented.
Fe(OH)x showed two broad H2 consumption peaks at ꢀ694 K and
890 K. After incorporation of Pd into Fe(OH)x, the peak at ꢀ694 K
disappeared, and a broad peak centered at ꢀ430 K emerged. Since
the XRD pattern of the Pd/Fe3O4 catalyst reduced at 473 K indi-
cated the formation of magnetite, the H2 consumption peak at
ꢀ430 K was assigned to the simultaneous reduction of Pd2+ to
metallic Pd and Fe(OH)x to magnetite. The high-temperature H2
consumption peak is therefore assigned to the reduction of magne-
tite to wüstite and metallic iron [26]. Considering the fact that the
peak areas of the low-temperature H2 consumption peaks of the
Pd/Fe(OH)x precursor and Fe(OH)x are similar, the H2 consumption
peak at 694 K for Fe(OH)x is assigned to the transition to magnetite
accordingly.
In the H2–TPR profiles of hematite and the Pd/Fe2O3 precursor,
the first H2 consumption peak was found at 675 K for hematite and
at 575 K for the Pd/Fe2O3 precursor, showing that Pd is also effec-
tive in promoting the reduction of hematite to magnetite. A similar
decrement in the reduction temperature of hematite has been ob-
served by Chang et al. on the Au/Fe2O3, Ru/Fe2O3, and Au–Ru/Fe2O3
catalysts prepared by deposition–precipitation [27]. The small H2
consumption peak at 451 K for the Pd/Fe2O3 precursor can be
attributed to the reduction of Pd2+ to metallic Pd based on the area
of this peak. Musolino et al. prepared a Pd/Fe2O3 precursor with the
same Pd loading and method and found that the reduction temper-
ature of hematite to magnetite was virtually identical to our
Fig. 1 shows the XRD patterns of the Pd/Fe(OH)x precursor, the
corresponding Pd/Fe3O4 catalyst, Fe(OH)x prepared in the way sim-
ilar to that of the Pd/Fe(OH)x precursor but in the absence of PdCl2
and subjected to the same reduction treatment as Pd/Fe(OH)x
(abbreviated as Fe(OH)x–R), Fe2O3 without supporting of Pd, as
well as the Pd/Fe2O3 catalyst. It is interesting that no diffraction
peak could be detected for the Pd/Fe(OH)x precursor, indicating
that Fe(OH)x was in the amorphous state, and PdO was highly dis-
persed with the crystallite size beyond the detection limit of XRD.
After reduction, weak diffraction peaks of magnetite (Fe3O4, JCPDS
85–1436) emerged for the Pd/Fe3O4 catalyst. In contrast, there
were only diffraction peaks ascribable to hematite (a-Fe2O3, JCPDS
85–0986) for Fe(OH)x–R, indicating that Pd was so active that
Fe(OH)x could be reduced to magnetite at 473 K. No diffraction
peaks of maghemite (
c-Fe2O3, JCPDS 39–1346) such as (1 1 0),
(2 1 0), and (2 1 1) reflections, which are stronger than the
c
-
Fe2O3(1 1 1) reflection at the same position of the Fe3O4(3 1 1)
reflection, were observed for the Pd/Fe3O4 catalyst, indicating the
exclusive transformation of Fe(OH)x to magnetite. Based on the
Scherrer equation and the most intensive Fe3O4(3 1 1) diffraction
peak, the average crystallite size of magnetite was estimated to
be ca. 11 nm. Again, no diffraction peaks of metallic Pd were dis-
cernable, indicating that Pd remained highly dispersed in the Pd/
Fe3O4 catalyst after reduction.
Fe O
Fe2O34
Pd3
675
Pd/Fe2O3
Fe2O3
575
Fe2O3
Fe(OH)x-R
451
694
Pd/Fe3O4 after reaction
Pd/Fe2O3 precursor
Pd/Fe3O4
Pd/Fe(OH)x
430
Fe(OH)x
20
40
60
80
2θ (degree)
Pd/Fe(OH)x
Fig. 1. XRD patterns of the Pd/Fe(OH)x precursor, the Pd/Fe3O4 catalyst before and
after 130 h on stream in APR of 5 wt.% EG at 498 K, 2.58 MPa, and WHSV of 3.6 hꢁ1
Fe(OH)x prepared in the way similar to that of the Pd/Fe(OH)x precursor but in the
absence of PdCl2 and subjected to the same reduction treatment as Pd/Fe(OH)x
(abbreviated as Fe(OH)x–R), Fe2O3 without supporting of Pd, and the Pd/Fe2O3
catalyst.
305
405
505
605
705
805
,
Temperature (K)
Fig. 2. H2–TPR profiles of the Pd/Fe(OH)x precursor, Fe(OH)x, the Pd/Fe2O3 precur-
sor, and hematite.

290
J. Liu et al. / Journal of Catalysis 274 (2010) 287–295
Table 1
observation [28]. However, they observed a lower reduction
temperature of Pd2+ to metallic Pd (ca. 360 K), which is probably
a result of changes in palladium salt and thermal treatment condi-
tions. The more drastic decrement in the reduction temperature of
Physicochemical properties of Fe(OH)x and Fe2O3 supports, Pd/Fe(OH)x and Pd/Fe2O3
precursors, and Pd/Fe3O4 and Pd/Fe2O3 catalysts.
a
Sample
SBET
Vpore
dpore
(nm)
Sact
(m2 gꢁ1
)
(cm3 gꢁ1
)
(m2 gꢁ1
)
ferric ions to magnetite in Fe(OH)x
(DT = 264 K) than in Fe2O3
Fe(OH)x
110
229
190
23
0.16
0.24
0.20
0.13
0.067
3.9
3.6
3.3
16.4
11.4
–
–
4.0
–
–
(
D
T = 100 K) in the presence of Pd is a clear sign of the much stron-
Pd/Fe(OH)x
Pd/Fe3O4
Fe2O3
Pd/Fe2O3
precursor
Pd/Fe2O3
ger interaction between Pd species and iron oxide in the Pd/
Fe(OH)x precursor than in the Pd/Fe2O3 precursor.
Fig. 3 shows the N2 isotherms and pore size distribution (PSD)
curves of the Pd/Fe3O4 and Pd/Fe2O3 catalysts. Notably, the Pd/
Fe3O4 catalyst exhibited a type IV isotherm and a type H2 hyster-
esis loop at the relative pressure range of 0.5–0.8, which is charac-
teristic of a mesoporous structure [29]. The corresponding PSD
curve displayed a relatively uniform pore size centered at ca.
3.4 nm. However, for the Pd/Fe2O3 catalyst the hysteresis loop
was shifted to higher relative pressure, indicating the existence
of larger pores as confirmed by its corresponding PSD curve.
As summarized in Table 1, the BET surface areas of the Pd/
Fe(OH)x precursor and Fe(OH)x were 229 and 110 m2 gꢁ1, respec-
tively, which were much larger than the values of 18 and
23 m2 gꢁ1 of the Pd/Fe2O3 precursor and hematite. It is remarkable
that the SBET of the Pd/Fe(OH)x precursor was about two times of
that of Fe(OH)x prepared in the same manner, showing that PdCl2
also effectively improves the dispersion of Fe(OH)x. In TEM images
of Fe(OH)x and the Pd/Fe(OH)x precursor (not shown), the better
dispersion of the nanoparticles in the latter is readily observed.
After reduction, the SBET of the Pd/Fe3O4 catalyst was slightly
decreased to 190 m2 gꢁ1, which was close to the value of 170
m2 gꢁ1 for a similarly prepared Pd/Fe3O4 catalyst by Musolino
et al. [28], but still being one order of magnitude larger than that
of the Pd/Fe2O3 catalyst (17 m2 gꢁ1).
18
17
0.057
12.0
2.1
a
Calculated from the amount of CO desorbed after saturation adsorption of CO.
to the (1 1 1) plane of fcc Pd (JCPDS 46–1043) was observed, and
Pd nanoparticle with dimension of less than 3 nm was clearly iden-
tified, which accounts for the absence of the diffraction peaks of
metallic Pd for the Pd/Fe3O4 catalyst. The interplanar spacing of
0.24 nm in Fig. 4b corresponds well to the (3 1 1) plane of magne-
tite [23]. For the Pd/Fe2O3 catalyst, Fig. 4c shows that hematite had
much larger dimension of about 16 nm and broad particle size dis-
tribution (standard deviation of 3 nm), and Pd nanoparticles had a
dimension of about 10 nm and tended to aggregate.
3.2. APR of EG to H2
Fig. 5 compares the catalytic activities in terms of the turnover
frequencies (TOFs) of CO2 over the Pd/Fe2O3, Pd/NiO, Pd/Cr2O3, Pd/
Al2O3, and Pd/ZrO2 catalysts prepared by IWI. The conversions of
EG to gas products over these catalysts were all below 7.5%. It is
evident that the Pd/Fe2O3 catalyst is intrinsically the most active
one among the catalysts prepared by IWI. The activities of the
Pd/Cr2O3 and the Pd/NiO catalysts were about one-sixth of that
of the Pd/Fe2O3 catalyst. When Al2O3 and ZrO2 were used as sup-
ports, the activities were further decreased by about one order of
magnitude, which may be ascribed to the absence of synergistic
effect between Pd and the support in these catalysts [11,30].
The TEM image in Fig. 4a shows that the particle size of magne-
tite in the Pd/Fe3O4 catalyst was mainly around 20 nm, and magne-
tite particles were uniform in size and well dispersed. In contrast,
only after close inspection one can find uniform and small Pd nano-
particles highly dispersed on magnetite, as indicated by the arrows
drawn in Fig. 4a. Fig. 4b shows the HRTEM image of the Pd/Fe3O4
catalyst, in which interplanar spacing of 0.22 nm corresponding
100
Pd/Fe3O4
0.20
3.4
80
60
0.15
Pd/Fe3O4
0.10
0.05
40
Pd/Fe2O3
20
0
12.4
Pd/Fe2O3
0.00
0.0
0.2
0.4
0.6
0.8
1.0
5
10
15
20
P/P
Pore Diameter (nm)
o
Fig. 3. N2 adsorption–desorption isotherms and pore size distribution curves of the Pd/Fe3O4 and Pd/Fe2O3 catalysts.

J. Liu et al. / Journal of Catalysis 274 (2010) 287–295
291
Fig. 4. TEM image (a) and HRTEM image (b) of the Pd/Fe3O4 catalyst, and TEM images of (c) the Pd/Fe2O3 catalyst and (d) the Pd/Fe3O4 catalyst after 130 h on stream in APR of
5 wt.% EG at 498 K, 2.58 MPa, and WHSV of 3.6 hꢁ1
.
Table 2
40
Comparison of catalytic properties over Pd catalysts supported on different metal
oxides in APR of 5 wt.% EG at 498 K, 2.58 MPa, WHSV of 3.6 hꢁ1, and after 6 h on
stream.
Catalyst
Pd/Fe3O4
Pd/Fe2O3
Pd/Cr2O3
Pd/NiO
Conversion (%)
Conv. to gas (%)
H2 sel. (%)
101.1
99.6
94.2
0.33
51.6
41.8
96.0
0.29
52.3
40.1
97.5
0.28
61.8
50.0
15.6
54.7
Alkane sel. (%)
20
Gas products (mol.%)
H2
CO2
CO
Methane
Ethane
Propane
Butane
70.2
29.5
0.24
0.10
0.00
0.00
0.00
70.4
29.2
0.10
0.08
0.00
0.00
0.00
70.9
28.3
0.68
0.09
0.00
0.00
0.00
28.5
32.9
0.00
37.2
1.10
0.17
0.02
Liquid products (mol.%, excluding unconverted EG)
Methanol
Ethanol
Acetone
88.8
11.1
0.10
0.00
97.7
2.30
0.00
0.00
83.2
16.5
0.20
0.02
84.3
15.4
0.20
0.10
0
2-Propanol
Fig. 5. TOFs of CO2 in APR of 5 wt.% EG over supported Pd catalysts at 498 K,
2.58 MPa, WHSV of 36.1 hꢁ1 for Fe3O4-, Fe2O3-, Cr2O3-, and NiO-supported catalysts,
3.6 hꢁ1 for Al2O3- and ZrO2-supported catalysts, and after 6 h on stream.
NiO in the Pd/NiO precursor can be reduced to the metallic state
at 512 K, which is close to the reduction temperature used here
to activate the catalysts. Metallic Ni is an excellent catalyst for
the methanation reaction that consumes H2 [31], resulting in lower
selectivity to H2 but higher selectivity to alkanes on the Pd/NiO cat-
alyst. On the other hand, Fe2O3 [18] and Cr2O3 [32] were beneficial
for the WGS reaction that generates H2. However, Fig. 6 shows that
the Pd/Cr2O3 catalyst suffered from rapid deactivation: it lost about
a half of its initial activity at 8 h on stream, which could be caused
by the instability of the Cr2O3 structure in the aqueous environ-
For the Pd/Fe2O3, Pd/NiO, and Pd/Cr2O3 catalysts showing
appreciable APR activities, the selectivities to H2 and alkanes as
well as gas and liquid product distributions under the same reac-
tion conditions are presented in Table 2. Although the Pd/NiO cat-
alyst displayed higher conversion of EG to gas products than other
two catalysts, the selectivity to H2 was only 16%, while the selectiv-
ities to H2 over the Pd/Fe2O3 and Pd/Cr2O3 catalysts were as high as
96.0% and 97.5%, respectively. According to Musolino et al. [28],

292
J. Liu et al. / Journal of Catalysis 274 (2010) 287–295
Table 3 summarizes the intrinsic activities of the Pd/Fe3O4 and
Pd/Fe2O
H2 sel.
3
Pd/Fe2O3 catalysts. The absence of transport limitations under the
reaction conditions specified in Table 3 was verified following the
method proposed by Koros and Nowak [34]. It has been claimed
that the reaction rate per gram of catalyst is proportional to S,
the number of active sites per unit volume of the reactor, in the ab-
sence of any transport limitations. Thus, for the Pd/Fe3O4 catalyst,
the reaction rate per gram of catalyst should remain constant at
various catalyst weights in the kinetically controlled regime. As
shown in Fig. 7, the H2 production rates per gram of the Pd/
Fe3O4 catalyst, which exhibited the highest activity, were almost
identical irrespective of the weight of the catalyst used, strongly
indicating that the data in Table 3 as well as in Fig. 5 are intrinsic.
Table 3 shows that in compliance with the results at high con-
versions, the conversion of EG to gas products on the Pd/Fe3O4 cat-
alyst was about 3.2 times of that on the Pd/Fe2O3 catalyst. When
using the Pd/Fe3O4 catalyst in the APR reaction, the production
rates of H2 and CO2 were accelerated, while the production rates
of CO and alkanes were suppressed. It is worth noting that on
100
80
60
40
20
0
Conv. of EG to gas
Alkane sel.
H2 sel.
Pd/Cr2O
3
100
80
60
40
20
0
the Pd/Fe3O4 catalyst the production rate of H2 amounted to
9:21 mmol minꢁ1
g
, corresponding to a H2 production rate of
ꢁ1
Pd
ꢁ3
ꢀ 700
lmol cm
minꢁ1. According to Cortright et al. [5], such
reactor
a H2 production rate is sufficient to sustain a 1 kW fuel cell assum-
ing 50% efficiency if a 1-l capacity reformer is used.
Conv. of EG to gas
Alkane sel.
The higher conversion and production rates of H2 and CO2 were
not caused exclusively by the higher dispersion of Pd nanoparticles
on the Pd/Fe3O4 catalyst leading to higher Sact. Table 3 shows that
the intrinsic activities of the Pd/Fe3O4 catalyst expressed as the
TOFs of H2 and CO2 were also dramatically improved, indicating
that the active sites on the Pd/Fe3O4 catalyst are not the same as
those on the Pd/Fe2O3 catalyst. The TOF of H2 on the Pd/Fe3O4 cat-
alyst was 109 minꢁ1, which was substantially higher than the best
value (60.1 minꢁ1) ever reported in the literature obtained on a Pd/
Fe2O3 catalyst prepared by IWI [9], and 48.8 minꢁ1 obtained on our
2
3
4
5
6
7
8
Time on Stream (h)
Fig. 6. A comparison of the catalytic performances of the Pd/Fe2O3 and Pd/Cr2O3
catalysts during 8 h on stream in APR of 5 wt.% EG at 498 K, 2.58 MPa, and WHSV of
3.6 hꢁ1
.
ment [33]. Although the initial conversion of EG of the Pd/Fe2O3
catalyst was lower than that of the Pd/Cr2O3 catalyst, it was more
stable. During the same time span, the Pd/Fe2O3 catalyst essentially
remained its initial activity. Thus, as far as the activity, selectivity,
and stability are concerned, Pd/Fe2O3 is the most suitable catalyst
for APR of EG to H2 among the supported Pd catalysts prepared by
IWI.
The catalytic performance of the Pd/Fe3O4 catalyst prepared by
co-precipitation was also presented in Fig. 5 and Table 2. It is evi-
dent that the APR activity of the Pd/Fe2O3 catalyst was far inferior
to that of the Pd/Fe3O4 catalyst, while the selectivity to H2 on the
latter was only slightly lower under the same reaction conditions.
On the Pd/Fe3O4 catalyst, the conversion of EG to gas products and
selectivity to H2 could achieve 99.6% and 94.2%, respectively. In a
control experiment, since Fig. 2 indicates that magnetite is formed
by reduction of Fe(OH)x at 694 K, we synthesized this magnetite
and used it to support Pd by IWI in the way identical to that of
the Pd/Fe2O3 catalyst. The conversion of EG to gas products on
the IWI-derived Pd/Fe3O4 catalyst was only 42.0% under the same
reaction conditions, unequivocally demonstrating that co-precipi-
tation is more effective than IWI in preparing iron oxide-supported
Pd catalyst for APR of EG to H2.
800
600
400
200
0.3
0.4
0.5
0.6
0.7
0.8
Weight of Pd/Fe3O4 (g)
Fig. 7. H2 production rate per gram of catalyst as a function of the weight of Pd/
Fe3O4 used in APR of 5 wt.% EG at 498 K, 2.58 MPa, and WHSV of 36.1 hꢁ1
.
Table 3
Kinetic results of APR of 5 wt.% EG on the Pd/Fe3O4 and Pd/Fe2O3 catalysts at 498 K, 2.58 MPa, WHSV of 36.1 hꢁ1, and after 6 h on stream.
ꢀ
ꢁ
Production rate mmol minꢁ1
g
=TOF ðminꢁ1
Þ
Catalyst
Conv. to gas (%)
Sel. (%)
H2
ꢁ1
Pd
H2
CO2
CO
CH4
Alkane
Pd/Fe3O4
Pd/Fe2O3
14.4
4.4
9.21/109
2.16/48.8
2.76/32.6
0.80/18.0
9.99 ꢃ 10ꢁ3/0.12
8.29 ꢃ 10ꢁ4/9.8 ꢃ 10ꢁ3
133
105
0.03
0.1
1.90 ꢃ 10ꢁ2/0.43
8.59 ꢃ 10ꢁ4/1.9 ꢃ 10ꢁ2

J. Liu et al. / Journal of Catalysis 274 (2010) 287–295
293
Pd/Fe2O3 catalyst also prepared by IWI. The lower TOF of H2 on our
Pd/Fe2O3 catalyst than that on the literature catalyst may be due to
experimental differences between different laboratories.
According to Shabaker et al. [10], the ratios of the partial
pressures QP (QP = PCO2PH2/PCO) under reaction conditions indicated
100
80
60
40
20
0
Pd/Fe3O4
in Table 3 were calculated, and the quotient QP/KWGS (KWGS
=
PCO2PH2/PCOaH2O) denoting the extent of the WGS reaction was
determined to be 0.25 and 0.067 for the Pd/Fe3O4 and Pd/Fe2O3
catalysts, respectively. The small QP/KWGS ratio for the Pd/Fe2O3
catalyst implies that the WGS reaction was far from equilibrium,
while the extent of the WGS reaction was greatly enhanced when
using the Pd/Fe3O4 catalyst. A good APR catalyst should be active
both in the cleavage of the C–C bond and in the WGS reaction
[11], and the rate-determining step for APR on Pd catalyst could
be the WGS reaction [9], which may be rationalized by previous
experimental [35–37] and theoretical work [38,39] that water does
not dissociate on the clean Pd{1 1 1} surface. The WGS reaction not
only produces more H2 from EG but also eliminates adsorbed CO
from the catalyst surface, or high surface coverage of CO would
lead to low catalytic activity [9].
Pd/Fe2O3
0.0
0.5
1.0
1.5
2.0
Time on Stream (h)
Fig. 8. CO conversions over the Pd/Fe3O4 and Pd/Fe2O3 catalysts at 498 K,
atmospheric pressure, feed composition of 10.6% CO and 29.3% H2O (balance N2),
and gas hourly space velocity of 15,000 hꢁ1
.
For high-temperature WGS reaction, magnetite rather than
hematite is the active phase and generally obeys the multi-step re-
dox or regenerative mechanism [40,41]. Boreskov demonstrated
that the Fe2+ and Fe3+ ions located in octahedral sites in the mag-
netite-based structure function as a redox couple and that magne-
tite-based catalysts can be highly effective for the complete
dissociation of water into H2 and adsorbed O under reaction condi-
tions. Water dissociation causes the oxidation of the octahedral
Fe2+ to Fe3+ and liberates H2. The oxidized iron centers may subse-
quently be reduced by CO thereby producing CO2 to complete the
catalytic cycle [42]. Among these steps, Keiski et al. indicated that
CO adsorption is one of the rate-determining steps based on the
stationary and transient kinetics of the WGS reaction over an
industrial ferrochrome catalyst [43]. Using TPD, IRAS, and HREELS,
Lemire et al. examined the adsorption of CO and water on
Fe3O4(1 1 1) films and evidenced that water readily dissociates
and blocks the Fe3+ step sites which bind most strongly with CO
[44]. For supported metal catalysts, the WGS reaction generally oc-
curs in a bifunctional manner, with the participation of both the
dispersed metallic phase and the support. The role of noble metal
is most probably restricted to CO adsorption and supply to the me-
tal–support interface to form CO2 [15]. It is well known that Pd
readily adsorbs CO [45]. According to our XRD, H2–TPR, and BET re-
sults, in the Pd/Fe3O4 catalyst the iron oxide support was in the
form of magnetite with large BET surface area, and Pd nanoparti-
cles were less than 3 nm in diameter. In the Pd/Fe2O3 catalyst,
the iron oxide support was mainly in the form of hematite with
small BET surface area, and only weak diffraction peaks of magne-
tite were present. The Pd nanoparticles in this catalyst were about
10 nm in diameter. Therefore, in the Pd/Fe3O4 catalyst a stronger
cooperative effect between Pd and magnetite than that in the Pd/
Fe2O3 catalyst is expected, thus resulting in exceptionally higher
TOFs of H2 and CO2 and consequently a higher Qp on the Pd/
Fe3O4 catalyst prepared by the co-precipitation method. This argu-
ment is directly justified by the much higher WGS activity over the
Pd/Fe3O4 catalyst than that over the Pd/Fe2O3 evaluated on a sep-
arate WGS reactor described previously [46]. In Fig. 8, it is found
that under the same WGS reaction conditions, the conversion of
CO on the Pd/Fe2O3 catalyst was 15%, whereas it was as high as
91% on the Pd/Fe3O4 catalyst.
ture was cycled from 453 K to 483 K and back, staying at each
temperature for 15 h. They pointed out that future catalyst devel-
opment work should focus on how to prepare more stable PdFe
catalyst [9]. Thus, it will be of practical significance to know
whether the Pd/Fe3O4 catalyst prepared by co-precipitation is also
stable in APR of EG to H2.
Fig. 9 presents the catalytic performance of the Pd/Fe3O4 cata-
lyst versus time-on-stream during APR of EG. It should be men-
tioned that the WHSV of 3.6 hꢁ1 used here is the experimentally
determined turning point at which the conversion of EG to gas
products just reached 100%. After the first 50 h on stream, the con-
version of EG to gas products only dropped slightly from 100% to
97%. During the next 50 h on stream, the catalyst underwent a rel-
atively obvious deactivation, approaching a steady-state conver-
sion of EG to gas products of approximately 80%. It is thus clear
that co-precipitation is a highly effective method to answer the
question raised by Huber et al. [9]. Moreover, during 130 h on
stream, the selectivity to H2 increased gradually from ca. 92% to
100%, and the selectivity to alkanes was always below 0.1%. The in-
crease in selectivity to H2 with aging may be due to the incomplete
reforming reactions. Considering its lower cost than Pt and its
higher stability and selectivity to H2 than Ni-based catalysts [25],
the Pd/Fe3O4 catalyst is a very promising high-performance candi-
date for converting EG to H2.
100
H2 sel.
80
Conv. of EG to gas
60
40
20
Alkane sel.
0
3.3. Catalytic stability
0
20
40
60
80
100
120
140
Time on Stream (h)
For the Pd/Fe2O3 catalyst, Huber et al. reported that although
the catalyst had the highest TOF of H2 among the catalysts tested
in APR of EG, it lost 50% of its activity after the reaction tempera-
Fig. 9. The catalytic performance of the Pd/Fe3O4 catalyst during 130 h on stream in
APR of 5 wt.% EG at 498 K, 2.58 MPa, and WHSV of 3.6 hꢁ1
.

294
J. Liu et al. / Journal of Catalysis 274 (2010) 287–295
Pd 3d
335.4
Fe 2p
710.8
Pd/Fe3O4
after reaction
Pd/Fe3O4
Pd/Fe3O4
after reaction
Pd/Fe3O4
350
345
340
335
330
740 735 730 725 720 715 710 705 700
Binding Energy (eV)
Binding Energy (eV)
Fig. 10. Pd 3d and Fe 2p spectra of the Pd/Fe3O4 catalyst before and after 130 h on stream in APR of 5 wt.% EG at 498 K, 2.58 MPa, and WHSV of 3.6 hꢁ1
.
To better understand the change in the catalytic performance in
to H2. On the Pd/Fe3O4 catalyst prepared by the co-precipitation
the stability test, the Pd/Fe3O4 catalyst after 130 h on stream was
systematically characterized. Chemical analysis revealed that the
concentrations of Pd and Fe dissolved in the liquid products after
130 h on stream were both less than 1.0 ppm, verifying that the
activity loss should not be induced by the leaching of the active
phase and the support. Pd 3d and Fe 2p spectra shown in Fig. 10
disclosed that after the stability test, the Pd species was still in
the metallic state with the Pd 3d5/2 BE of 335.4 eV [47], and the
Fe species was still in the form of magnetite with the Fe 2p3/2 BE
of ca. 710.8 eV [48]. For comparison, on Ni-based catalysts the oxi-
dation of metallic Ni occurred during APR of EG, which can be an
important reason for the deactivation of Ni-based APR catalysts
[25,49]. In addition, the surface Pd/Fe atomic ratio derived from
Fig. 10 increased from 0.05 to 0.11 after 130 h on stream.
method, the TOF of H2 in APR of EG represents the highest value
ever reported in open literature. Based on the characterizations
and literature works, we suggest a synergistic mechanism on the
Pd/Fe3O4 catalyst involving small Pd nanoparticles and magnetite
which effectively promotes the WGS reaction that is believed as
the rate-determining step for APR of EG over Pd-based catalysts.
Moreover, the Pd/Fe3O4 catalyst retained 80% of its initial APR
activity after reaching the steady-state. Thus, the Pd/Fe3O4 catalyst
prepared by the co-precipitation method can be regarded as the
most promising candidate thus far for the APR of biomass-derived
oxygenates to H2 for fuel cell applications.
Acknowledgments
The XRD pattern of the Pd/Fe3O4 catalyst after 130 h on stream
is also illustrated in Fig. 1. When compared to the as-reduced Pd/
Fe3O4 catalyst, the diffraction peaks due to magnetite were sharp-
ened, which signifies the sintering of magnetite. According to the
Scherrer equation, the crystallite size of magnetite in the Pd/
Fe3O4 catalyst after reaction was 21 nm, which is larger than the
value of 11 nm in the as-reduced Pd/Fe3O4 catalyst. After the sta-
bility test, there was still no diffraction peak of metallic Pd
(Fig. 1), signifying again the strong interaction between Pd and
magnetite in the Pd/Fe3O4 catalyst. So, the increased surface Pd/
Fe ratio after 130 h on stream obtained above can be rationalized
by the decreased dispersity of magnetite. However, in the TEM im-
age of the Pd/Fe3O4 catalyst after the stability test (Fig. 4d), we ob-
served some Pd nanoparticles with dimension of ca. 5 nm, possibly
due to the sintering of Pd nanoparticles loosely bonding with mag-
netite. Based on these results, we tentatively attribute the 20%
deactivation of the Pd/Fe3O4 catalyst after 130 h on stream to the
sintering of magnetite and some weakly bound Pd nanoparticles
with magnetite during APR of EG.
This work was supported by the National Basic Research Pro-
gram of China (2006CB202502), the Program of New Century
Excellent Talents in Universities (NCET-08-0126), the Fok Ying
Tong Education Foundation (104022), the NSF of China
(20673025, 20703011, 20803011, J0730419), and the Science and
Technology Commission of Shanghai Municipality (06JC14009,
08DZ2270500, 09ZR1402300). The authors also thank Ms. Rui Dai
for collecting the reaction data, Dr. Songhai Xie for his assistance
in TEM and XPS characterizations, and Prof. Kangnian Fan for his
valuable help in writing this paper.
References
[1] D. Pimentel, G. Rodrigues, T. Wang, R. Abrams, K. Goldberg, H. Staecker, E. Ma,
L. Brueckner, L. Trovato, C. Chow, U. Govindarajulu, S. Boerke, BioScience 44
(1994) 536.
[2] M. Ishida, K. Otsuka, S. Takenaka, I. Yamanaka, J. Chem. Technol. Biotechnol. 80
(2005) 281.
[3] D. Preti, S. Squarcialupi, G. Fachinetti, Angew. Chem. Int. Ed. 48 (2009) 4763.
[4] J. Turner, G. Sverdrup, M.K. Mann, P.C. Maness, B. Kroposki, M. Ghirardi, R.J.
Evans, D. Blake, Int. J. Energy Res. 32 (2008) 379.
[5] R.D. Cortright, R.R. Davda, J.A. Dumesic, Nature 418 (2002) 964.
[6] G.W. Huber, J.W. Shabaker, J.A. Dumesic, Science 300 (2003) 2075.
[7] R.R. Davda, J.W. Shabaker, G.W. Huber, R.D. Cortright, J.A. Dumesic, Appl. Catal.
B 43 (2003) 13.
[8] J.W. Shabaker, D.A. Simonetti, R.D. Cortright, J.A. Dumesic, J. Catal. 231 (2005)
67.
4. Conclusion
The preparation method has a profound effect on the catalytic
performance of iron oxide-supported Pd catalysts for APR of EG

J. Liu et al. / Journal of Catalysis 274 (2010) 287–295
295
[9] G.W. Huber, J.W. Shabaker, S.T. Evans, J.A. Dumesic, Appl. Catal. B 62 (2006)
226.
[10] J.W. Shabaker, R.R. Davda, G.W. Huber, R.D. Cortright, J.A. Dumesic, J. Catal. 215
(2003) 344.
[29] M. Kruk, M. Jaroniec, Chem. Mater. 13 (2001) 3169.
[30] J.W. Shabaker, G.W. Huber, R.R. Davda, R.D. Cortright, J.A. Dumesic, Catal. Lett.
88 (2003) 1.
[31] M.A. Vannice, J. Catal. 50 (1977) 228.
[11] R.R. Davda, J.W. Shabaker, G.W. Huber, R.D. Cortright, J.A. Dumesic, Appl. Catal.
B 56 (2005) 171.
[12] J.H. Sinfelt, D.J.C. Yates, J. Catal. 8 (1967) 82.
[13] D.C. Grenoble, M.M. Estadt, D.F. Ollis, J. Catal. 67 (1981) 90.
[14] J.A. Rodriguez, P. Liu, J. Hrbek, J. Evans, M. Pérez, Angew. Chem. Int. Ed. 46
(2007) 1329.
[15] P. Panagiotopoulou, D.I. Kondarides, J. Catal. 225 (2004) 327.
[16] M. Haruta, J. Nanopart. Res. 5 (2003) 3.
[32] L. Ma, B. Gong, T. Tran, M.S. Wainwright, Catal. Today 63 (2000) 499.
[33] X.L. Hong, S.Z. Ren, Int. J. Hydrogen Energy 33 (2008) 700.
[34] R.M. Koros, E.J. Nowak, Chem. Eng. Sci. 22 (1967) 470.
[35] X.Y. Zhu, J. White, M. Wolf, E. Hasselbrink, G. Ertl, J. Phys. Chem. 95 (1991)
8393.
[36] C. Clay, L. Cummings, A. Hodgson, Surf. Sci. 601 (2007) 562.
[37] M.J. Gladys, A.A. El Zein, A. Mikkelsen, J.N. Anderson, G. Held, Surf. Sci. 602
(2008) 3540.
[17] S. Zhao, R.J. Gorte, Catal. Lett. 92 (2004) 75.
[38] A. Michaelides, A. Alavi, D.A. King, Phys. Rev. B 69 (2004) 113404.
[39] G.S. Karlberg, Phys. Rev. B 74 (2006) 153414.
[40] C. Rhodes, G.J. Hutchings, A.M. Ward, Catal. Today 23 (1995) 43.
[41] C. Ratnasamy, J.P. Wagner, Catal. Rev. 51 (2009) 325.
[42] G.K. Boreskov, Kinet. Katal. 11 (1970) 374.
[18] X. Wang, R.J. Gorte, Appl. Catal. A 247 (2003) 157.
[19] F. Schüth, M. Hesse, K.K. Unger, in: G. Ertl, H. Knözinger, F. Schüth, J.
Weitkamp (Eds.), Handbook of Heterogeneous Catalysis, Wiley-VCH,
Weinheim, 2008, p. 100.
[20] F. Liu, S.P. Xu, L. Cao, Y. Chi, T. Zhang, D.F. Xue, J. Phys. Chem. C 111 (2007)
7396.
[43] R.L. Keiski, T. Salmi, P. Niemistö, J. Ainassaari, V.J. Pohjola, Appl. Catal. A 137
(1996) 349.
[21] G. Neri, G. Rizzo, L. De Luca, A. Donato, M.G. Musolino, R. Pietropaolo, Appl.
Catal. A 356 (2009) 113.
[44] C. Lemire, R. Meyer, V.E. Henrich, Sh. Shaikhutdinov, H.J. Freund, Surf. Sci. 572
(2004) 103.
[22] B. Bridier, N. López, J. Pérez-Ramírez, J. Catal. 269 (2010) 80.
[23] B.T. Qiao, L.Q. Liu, J. Zhang, Y.Q. Deng, J. Catal. 261 (2009) 241.
[24] H. Dropsch, M. Baerns, Appl. Catal. A: Gen. 158 (1997) 163.
[25] J.W. Shabaker, G.W. Huber, J.A. Dumesic, J. Catal. 222 (2004) 180.
[26] G. Munteanu, L. Ilieva, D. Andreeva, Thermochim. Acta 291 (1997) 171.
[27] F.W. Chang, L.S. Roselin, T.C. Ou, Appl. Catal. A 334 (2008) 147.
[28] M.G. Musolino, C. Busacca, F. Mauriello, R. Pietropaolo, Appl. Catal. A 379
(2010) 77.
[45] H. Conrad, G. Ertl, J. Koch, E.E. Latta, Surf. Sci. 43 (1974) 462.
[46] P.J. Guo, L.F. Chen, Q.Y. Yang, M.H. Qiao, H. Li, H.X. Li, H.L. Xu, K.N. Fan, Int. J.
Hydrogen Energy 34 (2009) 2361.
[47] P.M.Th.M. van Attekum, J.M. Trooster, Phys. Rev. B 20 (1979) 2335.
[48] T. Choudhury, S.O. Saied, J.L. Sullivan, A.M. Abbot, J. Phys. D 22 (1989) 1185.
[49] L.J. Zhu, P.J. Guo, X.W. Chu, S.R. Yan, M.H. Qiao, K.N. Fan, X.X. Zhang, B.N. Zong,
Green Chem. 10 (2008) 1323.