Effect of preparation method on structural characteristics and propane steam reforming performance of Ni-Al2O3 catalysts

Article abstract of DOI:10.1016/j.molcata.2008.09.011

Propane steam reforming was studied over Ni-Al₂O₃ catalysts that were prepared by a conventional impregnation (IM) method and a one-step sol-gel (SG) technique. Both Ni-Al₂O₃ catalysts showed similar initial act

Full text of DOI:10.1016/j.molcata.2008.09.011

Journal of Molecular Catalysis A: Chemical 297 (2009) 26–34
Contents lists available at ScienceDirect
Journal of Molecular Catalysis A: Chemical
journal homepage: www.elsevier.com/locate/molcata
Effect of preparation method on structural characteristics and propane steam
reforming performance of Ni–Al O catalysts
2
3
∗
Lingzhi Zhang, Xueqin Wang, Bing Tan, Umit S. Ozkan
Department of Chemical and Biomolecular Engineering, The Ohio State University, 140 West 19th Avenue, Columbus, OH 43210, United States
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 6 May 2008
Received in revised form 5 September 2008
Accepted 10 September 2008
Available online 18 September 2008
Propane steam reforming was studied over Ni–Al O catalysts that were prepared by a conventional
2
3
impregnation (IM) method and a one-step sol–gel (SG) technique. Both Ni–Al2O3 catalysts showed sim-
ilar initial activity. However, IM-Ni–Al2O3 deactivated severely with time-on-stream of propane steam
reforming. The catalyst prepared using a SG technique demonstrated stable catalytic performance. The
two catalysts also showed major differences in product distribution, with SG catalyst giving much higher
yields of hydrogen. Catalysts were characterized with temperature-programmed reduction (TPR), X-
ray photoelectron spectroscopy (XPS), temperature-programmed oxidation (TPO), transmission electron
microscopy (TEM), X-ray diffraction (XRD) and Raman spectroscopy. It was revealed that, with sol–gel
preparation, highly dispersed small Ni crystallites are formed with a strong interaction with the support.
This is shown to be important for coke suppression and catalyst stability.
Keywords:
Nickel catalyst
Sol–gel preparation
Metal–support interaction
Steam reforming
Deactivation
© 2008 Elsevier B.V. All rights reserved.
Sintering
Coking
Carbon filament
1. Introduction
cost and long-proven performance of Ni-based catalysts, therefore,
still warrant the effort to optimize these catalysts for steam reform-
Steam reforming of hydrocarbons has seen increasing inter-
est due to the necessity of efficient and cost-effective reforming
technologies for hydrogen production for fuel cell applications.
In practice, steam reforming of hydrocarbons, especially that of
methane, is performed at high temperatures over Ni-based cat-
alysts where the lower cost of nickel is an important advantage.
Current Ni-based catalysts, which are used for natural gas steam
reforming processes, suffer from catalyst deactivation by coke for-
mation and sintering of metallic Ni active phase [1–3]. It is generally
agreed that coke deposition is the major cause for activity loss.
ing applications, particularly for fuel processing from existing liquid
fuels such as gasoline and diesel.
Extensive efforts have been made on development of Ni catalysts
with improved resistance to coke formation. The support material
has been found to exert a strong influence on the reactivity of CHx
surface intermediates formed during hydrocarbon reforming [7,14].
The number of hydrogen atoms in surface CHx species is relevant
to the coking resistance of Ni catalysts [15]. Lower x values tend
to result in the formation of carbonaceous deposits, a precursor of
surface coke deposition. The nature of the support materials has
been investigated in recent years in an attempt to seek supports
with the desired acid–base or redox properties for the catalysts
or modify existing supports by addition of promoters to enhance
the reactivity of surface intermediates or suppress coke dissolution
Boudouard reaction (2CO → C + CO ) and methane/hydrocarbon
2
decomposition [4] are two major side reactions during hydrocarbon
steam reforming, which could lead to filamentous carbon accumu-
lation at the surface of Ni catalysts [5]. The coking process expedites
active metal degradation and activity loss. Carbon growth is sig-
nificantly decelerated on noble metals because carbon could not
dissolve in them [2,6,7]. Therefore, supported precious metals such
as Pd, Pt, and Rh are adopted for hydrocarbon reforming reactions
and have exhibited superior activity and stability [8–13]. However,
the cost of the precious metals is still a major drawback. The low-
[16]. Alkali metal oxides such as K O and CaO have been shown to
2
improve coking resistance by enhancing carbon gasification, but at
the expense of catalytic activity [1,17]. The introduction of a small
amount of molybdenum or tungsten (0.5 wt% MoO or WO ) into Ni
3
3
catalysts has been shown to increase the coking resistance without
loss in catalytic activity [18–22]. Alternatively, lanthanides emerge
to be promising promoters because they can inhibit coke forma-
tion without sacrificing the catalytic activity [23–25]. Zhuang et
al. [23] investigated the effect of cerium oxide as the promoter in
MgAl2O4-supported Ni catalysts for methane steam reforming at
∗ Corresponding author. Tel.: +1 614 292 6623; fax: +1 614 292 3769.
E-mail address: ozkan.1@osu.edu (U.S. Ozkan).
1381-1169/$ – see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.molcata.2008.09.011

L. Zhang et al. / Journal of Molecular Catalysis A: Chemical 297 (2009) 26–34
27
◦
5
50 C. Cerium promotion showed a beneficial effect by not only
Ni–Al O3 catalysts in propane steam reforming were examined
2
decreasing the rate of carbon deposition, but also increasing the
catalytic activity. Wang and Lu [24] stated that the addition of CeO2
[26,27].
into Ni/Al O catalysts enhanced the nickel dispersion and reactiv-
2
3
2. Experimental
ity of carbon deposits, leading to an improvement in the catalytic
activity and stability in CO2 reforming of methane. Earlier work
from our group has shown that incorporation of lanthanides into
the Ni matrix reduces coking by inhibiting the diffusion of carbon
into Ni particles [26,27].
Metal–support interaction is another phenomenon that affects
carbon deposition in steam reforming. The metal–support interac-
tion was found to correlate with Ni particle size, Ni dispersion and
morphology [28]. It has been discussed in literature [29–31] that
smaller Ni particle size and higher dispersion can minimize Ni sin-
tering at high temperatures or long-term operation and improve
coking resistance, therefore enhancing catalyst stability during
hydrocarbon reforming. The propensity to coking, to a large extent,
depends on the size of Ni particles. It is reported by Edwards and
Maitra [14] that it requires an ensemble of 6 or 7 atoms of Ni to cat-
alyze carbon formation and growth whereas only 3 or 4 atoms of Ni
2.1. Catalyst preparation
Ni–Al O3 catalysts were prepared using a SG technique and a
2
conventional IM method. Nickel precursor (Aldrich) used during
catalyst preparation was in nitrate form. Aluminum tri-sec-
butoxide (ATB; Aldrich) was used as aluminum precursor. For
SG synthesis, ethanol (Alfa Aesar) was used as the solvent. The
following variables were controlled during the synthesis: nickel
content = 20 wt%, H O/ATB molar ratio = 3.6, and pH of the result-
2
ing gels = 4.8. Initially, ATB was dissolved in ethanol. The aqueous
solution with the desired amount of nickel was added drop-wise
into the ethanol solution using a syringe pump at a flow rate of
3
0
.5 cm /min. The pH of the resulting green gels was measured and
adjusted by adding HNO (Fisher) or NH OH (Fisher). The gels were
stirred for an additional 15 min and were kept at room temperature
3
4
are needed for the CH reforming reaction. Stronger metal–support
4
for 30 min. The samples were then dried overnight in the oven at
interaction has been proven to enhance Ni dispersion dramatically,
which would prevent the formation of large Ni clusters not only
during catalyst preparation and activation, but also in the high
temperature reaction process. It was reported in Liu et al.’s work
◦
1
10 C to remove the remaining water and ethanol. The dry sam-
ples were ground to fine powders and were calcined under O2 at
◦
4
50 C for 4 h. For the IM synthesis, the Al O support was prepared
2
3
first with the SG technique as described above. Ni species were
introduced onto the Al O support using impregnation in aqueous
[
32] that nickel oxides interacting weakly with the support tend
2
3
to aggregate and sinter during catalyst reduction and reforming
and lead to activity loss. Similarly in Wang et al.’s studies [33],
suppression of coke formation and Ni sintering were attributed to
solid solution formation between active metals and the support.
However, on the other side, the interaction between Ni and the sup-
port leads to the formation of solid solution or other mixed oxide
compounds and sharply decreases the amount of available Ni for
the targeted steam reforming reaction. This can be compensated
by increasing the doping amount of Ni. Magnesia has been found
solution. The IM catalysts were then dried and calcined under the
same conditions as for the SG catalysts.
2
2
.2. Characterization
.2.1. BET surface area and H chemisorption
The surface areas of oxidic samples calcined at 450 C in O were
2
◦
2
measured using the BET method with a Micromeritics ASAP 2010
instrument, using N2 as the adsorbent at liquid N2 temperature
capable of controlling the size of metallic clusters for Ni–Al O sys-
◦
2
3
(77 K). Samples were degassed at 130 C for 12 h under vacuum
tem [14]. The presence of mobile Ti–O in Ni/TiO2 could break the
large ensembles of Ni atoms, which stimulate coke formation [7].
Other promoters including B, Ge, Ga and Zn were also examined for
modifying metal–support interactions [34].
In addition to doping other elements to enhance the interaction,
another strategy is development of novel preparation techniques to
improve metal dispersion and coking resistance. Co-precipitation
method was adopted in Pelletier et al.’s work [16] and it is evidenced
that this method is effective in achieving high dispersion and
improving coking resistance. Creation of solid solutions between
NiO and MgO during catalyst synthesis yields crystals of smaller
size with high reaction activity and stability [32,33,35].
before surface area measurements. The volumetric measurement
of H2 chemisorption was conducted using a Micromeritics ASAP
2010 Chemisorption system. Prior to adsorption measurements,
calcined samples were reduced in situ under H2 at the desired
◦
reduction temperature (600 C) for 2 h followed by evacuation
−5
◦
to 10 mmHg and cooling down to 35 C. Adsorption isotherms
were measured at equilibrium pressures between 50 mmHg and
5
00 mmHg. The first adsorption isotherm was established by mea-
suring the amount of gas adsorbed as a function of pressure. After
completing the first adsorption isotherm, the system was evacu-
−5
ated for 1 h at 10 mmHg. Then a second adsorption isotherm
was obtained. The amount of probe molecule chemisorbed was
calculated by taking the difference between the two isothermal
adsorption amounts.
In this study, structural characteristics and propane steam
reforming performance of Ni–Al O catalyst prepared by a sol–gel
2
3
technique were examined and compared to those of Ni–Al O3
2
catalyst prepared by a conventional impregnation. Although the
low temperatures chosen for the propane reforming can classify
the reaction as “pre-reforming”, the term “steam reforming” is
used in view of the significant levels of hydrogen yields obtained.
Characterization was performed using the BET method with N2
adsorption, H2 chemisorption, X-ray photoelectron spectroscopy,
temperature-programmed reduction, temperature-programmed
oxidation, X-ray diffraction, Raman spectroscopy and trans-
mission electron microscopy. It was found that, with the SG
2
.2.2. TPR studies
Temperature-programmed reduction of catalysts was con-
ducted using a laboratory-made gas flow system described in detail
elsewhere [36]. 50 mg sample was loaded in a 1/4 in. O.D. quartz U-
tube reactor. The sample was then heated under 10% O /He at 450 C
for 1 h followed by cooling to room temperature under Ar. The
reduction was performed with 10% H /Ar (40 cm (STP)/min). The
reactor temperature was raised using a ramp rate of 10 C/min to
900 C and held at that level for 10 min. To remove moisture formed
◦
2
3
2
◦
◦
technique, Ni species interact strongly with the Al O3 support,
during reduction, effluents from the reactor were sent through a
2
which could resist sintering of Ni species and retard the forma-
tion of carbon fibers. This study is a continuation of the work
where the effect of lanthanide promotion on catalytic perfor-
mance and deactivation characteristics of lanthanide-promoted
silica gel water trap before reaching the detector. H consumption
2
was measured continuously as a function of sample temperature
using a thermal conductivity detector (TCD) connected to a data-
acquisition system.

2
8
L. Zhang et al. / Journal of Molecular Catalysis A: Chemical 297 (2009) 26–34
2
.2.3. XRD studies
In situ X-ray diffraction patterns during reduction of catalysts
ionization detector (FID) and TCD detectors. Separations were per-
formed under Ar using two columns: Porapak Q (12 ft × 1/8 in.
SS, 80/100 mesh) and Molecular Sieve 13× (5 ft × 1/8 in. SS,
60/80 mesh). A GOW-MAC 069-50 ruthenium methanizer oper-
ated at 350 C was used with the FID for accurate quantification
of CO down to 10 ppm concentration level. The post-reaction sam-
were obtained with a Bruker D8 Advance X-Ray diffractometer
equipped with an atmosphere and temperature control stage and
using Cu K␣ radiation (ꢀ = 1.542 Å) operated at 40 kV and 50 mA.
The powder diffraction patterns were recorded in the 2Â range
◦
o
o
from 20 to 80 . Reduction was performed in situ under 5% H /N
ples were passivated with 1% O in He for 2 h at room temperature
2
2
2
gas flow using a linear temperature-program between isothermal
steps (50 C). The catalysts were kept at isothermal steps for 0.5 h
before being unloaded. Then they were saved in a desiccator for
TEM experiments.
◦
before data collection and the ramp rate between the consecutive
isothermal steps was 10 C/min.
Below are definitions used for catalytic performance calcula-
tions.
◦
moles of propane converted
total moles of propane in the feed
%
C3H8 conversion =
× 100
2
.2.4. XPS studies
X-ray photoelectron spectroscopy analysis was performed using
an AXIS Ultra XPS spectrometer, operated at 13 kV and 10 mA with
2
× moles of H produced
2
monochromator Al K␣ radiation (1486.6 eV). Samples were reduced
%H2 yield =
× 100
total moles of hydrogen in the feed
◦
with 20% H /N at 600 C for 2 h in a reactor. The reduced samples
2
2
were then sealed using the valves located at both ends of the reac-
tor. The reactors were then taken to an argon–atmosphere glove
box, where the catalysts were removed out of the reactors and
mounted onthe XPSsample holders usingaconductivecarbon tape.
The sample holders were transferred to the XPS chamber without
exposing them to the atmosphere using a controlled-atmosphere
transporter. All spectra were corrected using the C 1 s signal located
at 284.5 eV.
%
C-containing product selectivity for A
moles of C in product A
=
× 100
moles of C in all C-containing products
3
. Results and discussion
.1. Physical surface area and metallic Ni surface area
Table 1 lists the physical surface area of calcined catalysts and
3
2.2.5. TPO studies
Temperature-programmed oxidation of post-reaction catalysts
was performed using a laboratory-made gas flow system. Initially,
0 mg of post-reaction catalyst was placed in a 1/4 in. O.D. quartz U-
tube reactor and was heated under He (40 cm (STP)/min) at 400 C
for 30 min to remove adsorbed moisture and other species (e.g.
the bare support and the metallic Ni surface area of reduced
1
catalysts. IM-Ni–Al O3 catalyst was prepared by impregnating
2
3
◦
Ni(NO3)2 onto a sol–gel prepared Al O3 support. SG-Ni–Al O3
2
2
catalyst was prepared by a sol–gel technique with precursors
of Ni(NO3)2 and aluminum tri-sec-butoxide in an ethanol sol-
CO ). During TPO, the reactor temperature was raised under 5%
2
3
◦
◦
O /He (40 cm (STP)/min) with a ramp rate of 10 C/min to 900 C.
vent. IM-Ni–Al O3 catalyst has a lower surface area compared
2
2
Effluents from the reactor were continuously monitored as a func-
tion of sample temperature using a mass spectrometer (Thermo
Finnigan Trace GC ultra-Trace DSQ MS) with He as the carrier gas.
with SG-Ni–Al O3 catalyst, which is possibly due to loss of
2
pore size caused by Ni impregnation [16]. This is supported
by the pore volume measurement that IM-Ni–Al O has a pore
2
3
3
3
volume of 0.26 cm /g in comparison with 0.63 cm /g for SG-
Ni–Al2O3.
2.2.6. Raman studies
Raman spectroscopy was conducted with a Horiba Jobin-Yvon
H2 uptakes at mono-layer coverage of the metallic Ni were used
to estimate surface Ni atom area for reduced catalysts, assum-
ing that each surface Ni atom chemisorbs one hydrogen atom
(H/Nisurface = 1). It should be noted that H2 chemisorption is a
measurement of reducible Ni species on the surface. Reduced SG-
Ni–Al2O3 catalysts demonstrate a lower metallic surface area than
reduced IM-Ni–Al2O3 catalysts. This can be possibly attributed to
two reasons. First, the impregnation method may lead to more
Ni species staying on the surface instead of entering the bulk
as in sol–gel preparation. Another reason is that for Ni species
exposed on the surface in both catalytic systems, variation in
extents of reduction may exist. In IM catalysts, large Ni particles
of weak interaction with neighbouring support could make these
Ni species easy to reduce. However, in SG catalysts, as would be
evidenced by XPS or XRD results in later sections, there are much
smaller Ni crystallites and stronger interaction between Ni and the
LabRam HR800 spectrometer. The illumination source employed
is an internal HeNe laser with 633 nm wave length. The instru-
ment is equipped with a CCD detector and a standard Olympus
microscope. The laser power at the sample was controlled below
3
mW to avoid laser heating effects. Post-reaction samples used for
TPO and Raman analysis were kept on-line using a feed stream of
◦
C H /H O/N = 1/4/95 at 500 C for 28 h.
3
8
2
2
2
.2.7. TEM studies
TEM images were collected from Tecnai TF-20 operated at
00 kV. Post-reaction samples were ultrasonically dispersed in
ethanol. A drop of the suspension was then deposited on a Cu grid
for TEM observation.
2
2
.3. Reaction studies
Table 1
Propane steam reforming reaction experiments were performed
Surface area and metallic Ni surface area.
using a stainless steel fixed-bed flow reactor (1/4 in. O.D.). Cat-
alytic performance of different catalysts were compared using
2
2
Sample
Surface area (m /g)
Metallic Ni surface area (m /g)
2
◦
an equal surface area loading (5 m ) in the reactor at 500 C
SG-Al2O3
O-450-20% IM-Ni–Al2O3
O-450-20% SG-Ni–Al2O3
R-600-20% IM-Ni–Al O
3
R-600-20% SG-Ni–Al2O3
300
204
238
3
and C H /H O/N = 1/4/80 with a total flow rate at 170 cm /min.
3
8
2
2
Prior to a reaction, the catalyst was reduced in situ under 20%
◦
9.4
6.8
H /N at 600 C for 2 h. Effluents from the reactor were ana-
2
2
2
lyzed using an automated Shimadzu GC-14A equipped with flame

L. Zhang et al. / Journal of Molecular Catalysis A: Chemical 297 (2009) 26–34
29
Table 2
Content of Ni species for catalysts prepared by impregnation and sol–gel methods:
calcined and reduced samples.
Sample
Ni species (%)
Ni-852.7
Ni-854.9
Ni-856.1
O-450-IM-Ni–Al2O3
O-450-SG-Ni–Al2O3
R-600-IM-Ni–Al2O3
R-600-SG-Ni–Al2O3
–
–
70
53
73
51
–
27
49
30
47
–
◦
After a reduction step at 600 C, the NiO peak at 854.9 eV disap-
peared, replaced by a new peak centered at 852.7 eV (2p3/2). This
corresponds to metallic Ni species as a result of NiO reduction. It
was noted that, over both catalysts, NiO species were completely
converted to metallic Ni species under our reduction conditions.
However, there were negligible changes for the peak at 856.1 eV.
One can conclude that NiO on the support surface can be reduced
more easily whereas the Ni species at Ni 2p3 of 856.1 eV is hardly
/2
◦
reduced at 600 C, which is most likely NiAl O species instead of
2
4
Ni O mentioned earlier. Studies from literature [37,39] reveal that
2
3
NiAl O species is much more difficult to be reduced compared to
2
4
NiO species.
From the Ni 2p region of reduced catalysts (Fig. 1 and Table 2),
for IM-Ni–Al O , a larger portion of metallic Ni (70% compared with
2
3
5
3% for that of SG-Ni–Al O ) was detected on the surface, which
2 3
Fig. 1. X-ray photoelectron spectra (Ni 2p region) of Ni–Al2O3 catalysts. O-450:
calcined at 450 C; R-600: reduced at 600 C.
explains the H2 chemisorption measurement that there are more
reducible Ni species on the impregnated catalyst surface and the
sol–gel catalyst has a substantially lower degree of reduction due
to spinel formation.
◦
◦
support. Therefore, complete reduction of Ni species on the sur-
face is difficult for the reduction conditions employed during H2
chemisorption.
3.3. X-ray diffraction
3.2. XPS of Ni species in calcined and reduced catalysts
In situ X-ray diffraction patterns were collected during a
step-wise temperature-programmed reduction process where a
calcined sample was reduced in 5% H /N2 (Fig. 2). 5% H /N2 was
Fig. 1 presents the Ni 2p region of the X-ray photoelectron spec-
2
2
◦
◦
tra of Ni–Al O catalysts calcined at 450 C and reduced at 600 C
chosen instead of 20% H /N2 used in reaction testing because
2
3
2
(
O-450 and R-600). For peak identification, the spectra were decon-
of safety regulations imposed by the facility. Diffraction patterns
◦ ◦
were collected at different temperatures including 450 C, 500 C,
voluted based on the constraints of equal spin-orbit splitting for
Ni 2p peaks, the height ratio of Ni 2p3/2 to Ni 2p1/2 being con-
stant at 2 (theoretical value), and the full width at half maximum
◦
◦
◦
600 C, 700 C and 800 C. At each temperature, a 30 min hold-
ing time was used before data collection. The International Center
for Diffraction Data (ICDD) library was used for phase identifi-
cation. The major diffraction features observed over the oxidized
(FWHM) of these peaks being equal. For calcined catalysts, the peak
located between 855 eV and 857 eV can be interpreted as a com-
bination of two Ni 2p3/2 peaks with binding energies of 854.9 eV
and 856.1 eV. There have been previous investigations of the XPS
◦
samples (R-27 C) are characteristic of ␥-Al O and correspond
2
3
to (3 1 1), (4 0 0) and (4 4 0) diffraction lines. Other possible crys-
talline phases present in the calcined catalysts could be bulk NiO
Ni 2p spectral region in the Ni–Mo–␥-Al O3 system by our group
2
[
37] as well as by others [38,39]. The peak at 854.9 eV suggests the
or NiAl O4 because they have overlapping diffraction lines with
2
presence of NiO on the surface, which can be supported by strong
shake-up lines around 860.7 eV (2p3/2), indicative of Ni2+ species.
There has been no affirmative assignment of the peak at 856.1 eV
reported in the literature, with the possibility of NiAl O or Ni O .
␥-Al O3 phase. With an increase in reduction temperature, the
2
(4 4 0) diffraction line shifts to a higher angle. The diffraction lines
◦
from metallic Ni appeared after reduction above 450 C for both
Ni–Al O3 catalysts prepared by the IM method and SG technique.
2
4
2
3
2
We tend to ascribe this feature to a spinel NiAl O4 structure on
The intensity of metallic Ni diffraction lines from Ni–Al O -IM cata-
2
2
3
the surface, which can be supported by our preparation methods
lyst grows faster than those from Ni–Al O -SG catalyst. Meanwhile
2 3
creating a stronger bonding between Ni and the Al O3 support.
the diffraction lines become sharper with increasing reduction
temperatures. These results indicate that the Ni species in impreg-
nation catalysts can be reduced more easily and form larger metallic
2
This assignment is also in agreement with findings obtained from
other characterization results presented in the paper. The “sur-
face spinel” model proposed by Jacono et al. [38] where Ni2 ions
+
particles. Because of the overlap between Al O3 (4 0 0) diffrac-
2
occupy the tetrahedral sites in ␥-Al O3 again supports this possi-
tion line and the most intensive metallic Ni (1 1 1) diffraction line,
Ni (2 0 0) diffraction line broadening was used to calculate the
metallic Ni crystallite size with Scherrer equation. The FWHM was
obtained with GRAM software. The calculated results are reported
2
bility. The two deconvoluted peaks described earlier (NiO–854.9 eV
and NiAl O –856.1 eV) were integrated and the quantified per-
2
4
centages were shown in Table 2. SG-Ni–Al O3 exhibits a larger
2
portion of NiAl O4 species, 49% in comparison to 27% for that of
in Table 3. For both Ni–Al O catalysts, they started with the same
Ni crystallite size at 500 C (9 nm). With increase in reduction tem-
perature, the crystallite Ni size tends to grow larger, evidenced
2
2 3
◦
IM-Ni–Al O . Impregnation catalysts have more NiO species on the
2
3
surface.

3
0
L. Zhang et al. / Journal of Molecular Catalysis A: Chemical 297 (2009) 26–34
Fig. 3. Temperature-programmed reduction profiles of Ni–Al2O3 catalysts.
Fig. 4. Time-on-stream performance of Ni–Al2O3 catalysts during steam reforming
Fig. 2. In situ X-ray diffraction patterns of Ni–Al2O3 catalysts prepared by (a)
impregnation method and (b) sol–gel technique during temperature-programmed
reduction.
reaction: %propane conversion.
by sharper and more intense diffraction lines. For sol–gel cata-
lysts, metallic Ni size stabilized at 13 nm whereas for impregnated
catalysts, it continued to grow with increase in reduction tem-
◦
peratures. By 800 C, this size was more than double the size
◦
at 500 C. These findings are consistent with other characteris-
tics of the sol–gel preparation technique (high BET surface area,
improved thermal stability, small particle size) summarized by
Gonzalez et al. [40]. The weaker metal–support interaction in
impregnated catalysts makes the Ni particles aggregate more easily
into polymeric structures and form bigger Ni particles after reduc-
tion.
Table 3
Ni crystallite size after reduction at different temperatures.
◦
Catalyst
d (nm) at different reduction temperatures ( C)
500
600
700
800
IM-Ni–Al2O3
SG-Ni–Al2O3
9
9
15
12
17
13
20
13
Fig. 5. Comparison of H2 yields over Ni–Al2O3 catalysts during propane steam
Note: Calculation based on broadening of Ni (2 0 0) diffraction lines.
reforming.

L. Zhang et al. / Journal of Molecular Catalysis A: Chemical 297 (2009) 26–34
31
Fig. 7. Temperature-programmed oxidation profiles of carbon deposited on post-
reaction Ni–Al2O3 catalysts.
alysts, which is consistent with our statement that there is stronger
metal–support interaction for sol–gel catalysts and this structure
difference yields more reducible NiO species for impregnated cat-
alysts.
3
.5. Steam reforming of propane over reduced Ni–Al O catalysts
2 3
◦
Reduced (600 C) 20% Ni–Al O catalysts prepared by sol–gel
2
3
and impregnation methods were evaluated with regard to activ-
ity, product distribution and stability in propane steam reforming
◦
at 500 C. The feed composition used in the experiments was
C H /H O/N = 1/4/80. The products are H and C-containing
3
8
2
2
2
species (CO, CO , and CH4 with CO2 being the main one) from
2
propane steam reforming.
Fig. 4 presents the C H conversion over the catalysts prepared
3
8
by the two different techniques. The reaction conditions are chosen
so that complete conversion is not reached in these experiments. As
seen in Fig. 4, the conversion levels over the two catalysts are com-
parable initially. However, the IM-Ni–Al O catalyst shows a much
Fig. 6. C-containing product distribution during steam reforming reaction of
propane: (a) IM-Ni–Al2O3 catalyst; (b) SG-Ni–Al2O3 catalyst.
2
3
faster decline in activity compared to the SG-Ni–Al O catalyst. By
2
3
3
.4. TPR profiles of Ni–Al O catalysts
2 3
Temperature-programmed reduction experiments on Ni–Al O3
2
catalysts were performed using 10% H /Ar as the reducing agent. As
2
shown in Fig. 3, TPR profiles exhibit two distinct reduction features
accompanied by a shoulder peak at high temperature. A small low
◦
temperature reduction peak around 400 C could be assigned to the
reduction of trace amounts of NiO clusters or bulk NiO weakly inter-
acting with the support, which was reported in Song et al.’s work
[
41] as the approximate reduction temperature for unsupported
NiO. Also in their publication, Ni species supported on ␥-Al O
show a reduction feature at 500–700 C. Correspondingly in our
2
3
◦
TPR profiles, the broad high temperature reduction peak around
◦
6
00 C can be ascribed to the reduction of NiOx species that have
stronger binding to the support. Sol–gel catalyst displays a slight
◦
shift of this reduction peak (at 620 C) compared with impregnated
◦
catalyst (at 596 C). Further interaction between NiO and the sup-
port would lead to the formation of NiAl O , which could not be
2
4
◦
◦
reduced until 800 C [41]. The tail reduction peak after 800 C for
SG-Ni–Al O might have contributions from NiAl O4 reduction.
2
3
2
The H2 consumption can be determined by integrating the peak
area under the TPR curve for the two Ni–Al O3 catalysts. An area
2
ratio of 1.3 was obtained for IM-Ni–Al O3 over SG-Ni–Al O . This
2
2
3
suggests that more Ni species can be reduced over impregnated cat-
Fig. 8. Raman spectra of calcined and post-reaction Ni–Al2O3 catalysts.

3
2
L. Zhang et al. / Journal of Molecular Catalysis A: Chemical 297 (2009) 26–34
Fig. 9. Post-reaction TEM images of IM-Ni–Al2O3 catalysts.
the 19th hour on-stream, the IM catalyst is seen to lose more than
half of its activity. The stability difference presented in Fig. 4 could
be associated with the Ni cluster size. There has been discussion in
the literature about the relationship between metallic Ni size and
catalyst stability [2,14,16,28,42,43]. Larger Ni particle size catalyzes
the coke formation reaction, which is detrimental for catalyst sta-
seen over the IM catalyst. When the carbon-containing product
selectivities are examined, there appear to be important differ-
ences as well. Over the IM-Ni–Al O3 catalyst, CH4 is the main
2
C-containing product at the beginning of the run. As the activity
of the catalyst declines, the CH4 selectivity also decreases, with
accompanying increases in CO and CO selectivities. The opposing
2
bility. Also, as shown by the in situ X-ray diffraction, IM-Ni–Al O3
trends between the COx selectivities and CH selectivity can partly
2
4
catalyst is more likely to lead to Ni particle size growth through sin-
tering. TEM images to be discussed later provide visual evidence of
Ni sintering taking place during reaction. It is clear that catalyst
preparation plays an important role in determining Ni dispersion
and particle size, metal–support interaction and morphology. Sim-
ilar observations have been reported earlier for Ni-based catalytic
systems [32,33,35].
explain why H2 yield remains constant in spite of the decreasing
propane conversions.
The selectivities over the SG-Ni–Al O catalyst, are much more
2
3
constant compared to their counter-parts observed over the IM cat-
alyst. The major C-containing product is CO , followed by CH and
2
4
CO. There is a slight decrease in CO2 selectivity accompanied by a
slight increase in CH selectivity. These changes correspond to the
4
The hydrogen yields and the selectivities of carbon-containing
slight decline in activity after the 14th hour on stream over this
catalyst.
products (CO, CO , CH ) are presented in Figs. 5 and 6, respectively.
2
4
It should be noted that under these conditions, thermodynamic
equilibrium would lead to complete propane conversion and a H2
3.6. Coke deposition characterization on post-reaction catalysts
yield of 62%. The equilibrium selectivities for CO , CO and CH4
2
would be 43%, 25% and 32%, respectively. So, it is clear that the
product distributions in our experiments are not dictated by ther-
modynamic equilibrium. Although the propane conversion levels
are comparable at the beginning of the run, the H2 yields show a
major difference between the two catalysts right from the begin-
ning, with the SG catalyst giving a much higher hydrogen yield.
Another interesting feature seen in Fig. 5 is the fact that hydrogen
yields remain fairly constant in spite of the decline in conversion
In this section, temperature-programmed oxidation and Raman
spectroscopy were used to characterize the propane steam reform-
ing post-reaction catalysts. TPO experiments were performed with
3
◦
5% O /He (40 cm (STP)/min) at a ramp rate of 10 C/min from
2
◦
room temperature to 900 C. Two peaks were observed from the
TPO profile (Fig. 7). A low temperature peak around 350 C is
◦
likely caused by combustion of highly reactive monoatomic car-
bon species deposited on Ni surfaces (Type I). Another peak of

L. Zhang et al. / Journal of Molecular Catalysis A: Chemical 297 (2009) 26–34
33
Fig. 10. Post-reaction TEM images of SG-Ni–Al2O3 catalysts.
◦
much higher intensity around 630 C can be assigned to filamentous
is developing under steam reforming reaction conditions used in
this study.
carbon (Type II), which is typically stable and oxidized at higher
temperatures [27,44,45]. Because of large intensity scale differences
for these two carbon oxidation peaks, the low temperature oxida-
tion peak was plotted as an inset in Fig. 7. It is shown that sample
preparation leads to differences in this low temperature feature. For
3.7. TEM characterization of post-reaction samples
TEM images acquired of the post-reaction samples show major
differences in the way carbon was deposited on these catalysts.
◦
◦
SG catalysts, there is a plateau between 330 C and 380 C resulting
from Type I carbon oxidation before the major Type II carbon com-
bustion takes off. This peak for IM catalyst has lower intensity and
it quickly evolves into Type II carbon combustion stage. Although
the trends of CO evolution and O consumption are similar for the
Figs. 9 and 10 show the images of post-reaction IM-Ni–Al O3 and
2
SG-Ni–Al O3 catalysts, respectively. A comparison of the images
2
point to major differences in the way carbon is deposited over these
two catalysts. Over the impregnated post-reaction sample, carbon
deposition is exclusively in the form of long filaments, some being
as long as 500 nm or more. Some fibers have hollow cores with Ni
particles at the tip. Although most of the fibers have a diameter of
about 15 nm, there are some with much larger diameters. Some of
the particles found at the tip of the fibers are no longer spherical,
but elongated along the axis of the fibers, some of them exhibit-
ing dimensions over 35 nm in one direction (Fig. 9(d)). Because
the size of Ni particles determines the carbon filament diameter
[27,44], the carbon filament diameter distribution is suggestive of
a non-uniform distribution of Ni particle sizes. There are two pos-
sible reasons that could lead to a non-uniform size distribution.
The impregnation method used for the preparation may generate
different dimensions of Ni ensembles. Additionally, Ni particles on
impregnated catalysts may tend to aggregate and form larger parti-
cles under reduction and reaction conditions. The presence of large
Ni particles accelerates the coke formation reaction, as evidenced
by the heavily carbon-covered catalysts from the TEM images.
2
2
two post-reaction catalysts, the amount of CO2 produced and O2
consumed during the oxidation of carbon deposits on 20% Ni–Al O3
catalysts are significantly different. A much larger amount of carbon
is deposited on Ni–Al O3 catalysts prepared by the impregnation
method during steam reforming of propane. The ratio of CO pro-
duced for IM/SG catalysts is about 1.5 as calculated by integrating
the peak area under the m/z = 44 evolution curve.
Fig. 8 shows the Raman spectra of Ni–Al O catalysts in fresh and
spent form. Two distinctive bands are observed for post-reaction
IM-Ni–Al O catalyst, at 1320 cm
on SG-Ni–Al O catalyst, they are very weak and broad. The
200–1700 cm Raman spectra region is widely used to study dis-
order in carbon structures, with two characteristic bands: D bands
around 1339–1345 cm and G band around 1592–1597 cm [46].
The peak broadening for sol–gel sample indicates a very poorly
graphitized structure. The more visible peak for impregnated cat-
alysts suggests that a certain ordering of the carbonaceous deposit
2
2
2
2
3
−1
−1
and 1597 cm . However,
2
3
2
3
−
1
1
−1
−1

34
L. Zhang et al. / Journal of Molecular Catalysis A: Chemical 297 (2009) 26–34
For the sol–gel post-reaction sample, fiber growth is markedly
binding between metal and support on sol–gel catalysts minimizes
Ni sintering at high temperatures and suppresses carbon dissolu-
tion and diffusion into Ni particles, protecting them from lifting up
from the surface during carbon filament growth. Therefore, a much
more stable catalytic performance is achieved with sol–gel catalytic
system.
suppressed (Fig. 10). Although there are filaments, they are signifi-
cantly shorter and thinner. Most of the particles still seem to be on
the surface and no large Ni ensembles are seen. In fact most of the
particles seem to be less than 10 nm. This is in agreement with the
high Ni dispersion and small particle size in sol–gel preparation.
It also suggests that there is no sintering of Ni during reduction or
reaction as seen over the impregnated sample. Some of the fibers
appear to be solid as seen in Fig. 10(d). Most fibers are well crys-
tallized, but there is amorphous carbon as well, suggesting that
the weak ordering revealed by the Raman spectroscopy might be
because of the small percentage of graphite carbon fibers in the
post-reaction sample.
Acknowledgments
The financial support provided by Honda Research Institute, USA
Inc., the National Science Foundation (NSFDMR grant # 0114098),
the Ohio Department of Development (342-0561 and OCRC4-06-
3-C3.31) is gratefully acknowledged.
Combining the surface and structural examination on Ni–Al O3
2
catalysts prepared by the sol–gel and impregnation techniques,
reaction tests, post-reaction sample characterization and earlier
studies by our group [26,27], it appears that the preparation tech-
nique leads to significant differences in the way coking takes place
on the surface. The fiber growth observed over the impregnated
sample is typical of a mechanism that involves dissolution and
diffusion of carbon into the metal, which results in lifting of the
metal particles from the support surface as fibers grow [47]. Car-
bon formation mechanisms, the effectofdiffusion pathsandsurface
morphologies over supported metal catalysts have been discussed
extensively in the literature, with representative discussions in
References
[
[
1] D.E. Ridler, M.V. Twigg, Catalyst Handbook, 2nd ed., ICI, Wolf Publishing Ltd.,
London, 1989, p. 225.
2] F. Pompeo, N.N. Nichio, M.M.V.M. Souza, D.V. Cesar, O.A. Ferretti, M. Schmal,
Appl. Catal. A: Gen. 316 (2007) 175.
[3] V. Modafferi, G. Panzera, V. Baglio, F. Frusteri, P.L. Antonucci, Appl. Catal. A: Gen.
34 (2008) 1.
3
[
4] K.M. Hardiman, T.T. Ying, A.A. Adesina, E.M. Kennedy, B.Z. Dlugogorski, Chem.
Eng. J. 102 (2004) 119.
[5] Q. Ming, T. Healey, L. Allen, P. Irving, Catal. Today 77 (2002) 51.
[6] J.R. Rostrup-Nielsen, I. Alstrup, Catal. Today 53 (1999) 311.
[7] M.C.J. Bradford, M.A. Vannice, Appl. Catal. A: Gen. 142 (1996) 73.
[8] J.T. Richardson, S.A. Paripatyadar, Appl. Catal. A: Gen. 61 (1990) 293.
[
48–50].
During the steam reforming process, propane initially cracks
[9] A. Igarashi, T. Ohtaka, S. Motoki, Catal. Lett. 13 (1991) 189.
10] J.R. Rostrup-Nielsen, J.H. Hansen, J. Catal. 144 (1993) 38.
[
[
11] D.L. Qin, J. Catal. Today 21 (1994) 551.
12] R. Craciun, B. Shereck, R.J. Gorte, Catal. Lett. 51 (1998) 149.
[13] X. Wang, R.J. Gorte, Catal. Lett. 73 (2001) 15.
into CHx fragments and dissociatively adsorbs on the surface of
metallic Ni particles. Subsequently, the adsorbed CHx fragments
react with oxygen-containing surface species to produce CO and
[
[14] J.H. Edwards, A.M. Maitra, Fuel Process. Technol. 42 (1995) 269.
[15] Y.-G. Chen, K. Tomishige, K. Yokoyama, K. Fujimoto, J. Catal. 184 (1999) 479.
[16] L. Pelletier, D.D.S. Liu, Appl. Catal. A: Gen. 317 (2007) 293.
H . The CHx fragments could further decompose to carbon and dis-
2
solve into Ni particles. With increased carbon formation on the Ni
surface, the concentration gradient induces more carbon to diffuse
through the Ni particle towards the Ni–support interface, where the
carbon interacts with the Al O surface and accumulates, leading to
carbon filament growth. The growth of carbon filament could break
the attachment between metal and the support and lift the metal
particle up. The carbon fiber growth results in a significant expan-
sion of the catalyst bed, severe operation problems and substantial
activity loss.
[17] J.R. Rostrup-Nielsen, Catalytic Steam Reforming, in: J.R. Anderson, M. Boudart
Eds.), Catalysis: Science and Technology, 5, Springer Verlag, New York, 1984,
(
p. 3.
[18] T. Borowiecki, A. Golebiowski, Catal. Lett. 25 (1994) 309.
2
3
[19] T. Borowiecki, A. Golebiowski, B. Stasinska, Appl. Catal. A: Gen. 153 (1997) 141.
[20] B. Stasinska, A. Golebiowski, T. Borowiecki, in: B. Delmon, G.F. Froment (Eds.),
Catalyst Deactivation, Elsevier, Amsterdam, 1999, p. 431.
[21] T. Borowiecki, A. Machocki, in: B. Delmon, G.F. Froment (Eds.), Catalyst Deacti-
vation, Elsevier, Amsterdam, 1999, p. 435.
[22] T. Borowiecki, G. Giecko, M. Panczyk, Appl. Catal. A: Gen. 230 (2002) 85.
[23] Q. Zhuang, Y. Qin, L. Chang, Appl. Catal. 70 (1991) 1.
[24] S. Wang, G.Q. Lu, Appl. Catal. B: Environ. 19 (1998) 267.
[25] R.M. Sambrook, J.R.H. Ross, US Patent 4 469 815 (1984).
For catalysts prepared by impregnation method, Ni particles
weakly attach to the support, which makes it easier for the parti-
cles to detach from Al O surfaces. For catalysts prepared by sol–gel
[26] S. Natesakhawat, O. Oktar, U.S. Ozkan, J. Mol. Catal. A: Chem. 241 (2005) 133.
27] S. Natesakhawat, R.B. Watson, X. Wang, U.S. Ozkan, J. Catal. 234 (2005) 496.
28] A. Slagtern, Y. Schuurman, C. Leclercq, X. Verykios, C. Mirodatosy, J. Catal. 172
1997) 118.
[29] Z. Hou, T. Yashima, Appl. Catal. A: Gen. 261 (2004) 205.
30] J.-H. Kim, D.-J. Suh, T.-J. Park, K.-L. Kim, Appl. Catal. A: Gen. 197 (2000) 191.
31] H. Hayashi, S. Murata, T. Tago, M. Kishida, K. Wakabayashi, Chem. Lett. 30 (2001)
4.
[32] B.S. Liu, C.T. Au, Appl. Catal. A: Gen. 244 (2003) 181.
[
[
2
3
technique, Ni particles bond strongly with the support and part of
the particles may be anchored into the Al O matrix. The chemical
(
2
3
[
[
bonding between Ni particles and the support prevents Ni particles
from leaving the support, thus retarding the formation of carbon fil-
ament. Part of the coking on the sol–gel sample may also be through
an “extrusion” process where carbon, instead of dissolving and dif-
fusing into the metal particle, “extrudes” over it, thus leaving the
metal particle intact on the surface. The solid filaments seen over
SG samples with a “stacked-platelet structure” with no hollow core
may be the result of such a process. Some of the “extruded” carbon
can also be in amorphous form.
3
[33] H.Y. Wang, E. Ruckenstein, Appl. Catal. A: Gen. 209 (2001) 207.
[34] B. Jongsomjit, J. Panpranot, J.G. Goodwin Jr., J. Catal. 215 (2003) 66.
[35] Y. Chen, J. Ren, Catal. Lett. 29 (1994) 39.
[36] U.S. Ozkan, Y. Cai, M.W. Kumthekar, L. Zhang, J. Catal. 142 (1993) 182.
[37] X. Wang, U.S. Ozkan, J. Phys. Chem. B 109 (2005) 1882.
[
[
[
38] M.L. Jacono, M. Schiavello, A. Cimino, J. Phys. Chem. 75 (1971) 1044.
39] K.T. Ng, D.M. Hercules, J. Phys. Chem. 80 (1976) 2094.
40] R.D. Gonzalez, T. Lopez, R. Gómez, Catal. Today 35 (1997) 293.
[41] M.H. Youn, J.G. Seo, P. Kim, I.K. Song, J. Mol. Catal. A: Chem. 261 (2007) 276.
[
[
42] V.R. Choudhary, A.S. Mamman, Appl. Energ. 66 (2000) 161.
43] J.S. Lisboa, D.C.R.M. Santos, F.B. Passos, F.B. Noronha, Catal. Today 101 (2005)
4
. Summary
1
5.
[
44] P. Wang, E. Tanabe, K. Itob, J. Jia, H. Morioka, T. Shishido, K. Takehira, Appl. Catal.
Compared with conventional impregnation methods, one pot
A: Gen. 231 (2002) 35.
sol–gel preparation yields Ni–Al O catalysts with highly dispersed
2
3
[45] S. Wang, G.Q. Lu, Ind. Eng. Chem. Res. 38 (1999) 2615.
Ni particles on the surface and a strong metal–support interac-
tion. Ni crystallite size over the sol–gel sample is relatively small
and thermally stable under reaction conditions, which suppresses
the carbon filament formation and filament growth. The chemical
[46] J. Llorca, N. Homs, J. Sales, P. Ramírez de la Piscina, J. Catal. 209 (2002) 306.
[47] N.M. Rodriguez, J. Mater. Res. 8 (1993) 3233.
[
[
48] J.-W. Snoeck, G.F. Froment, M. Fowlesz, J. Catal. 169 (1997) 240.
49] J.-W. Snoeck, G.F. Froment, M. Fowlesy, J. Catal. 169 (1997) 250.
[50] R.T.K. Baker, M.A. Barber, P.S. Harris, F.S. Feates, R.J. Waite, J. Catal. 26 (1972) 51.