Mechanistic study of the unusual catalytic properties of a new Ni{single bond}Ce mixed oxide for the CO2 reforming of methane

Article abstract of DOI:10.1016/j.jcat.2007.01.011

The origin of the exceptional properties of a Ni₁₀Ce₉₀ mixed oxide relative to a Ni/Al₂O₃ catalyst (i.e., immediate activity without catalyst prereduction, high resistance toward coking, high catalytic activity) and its catalytic deactivation for the CO₂ reforming of methane was investigated by XRD, XPS, pulse experimentation, and regeneration tests. Based on those results, a reaction mechanism is proposed. First, the catalytic property with no prereduction step relies on the reactivity of oxygen contained in the Ni phase. The rapid activation of Ni₁₀Ce₉₀ is based on the active interfacial oxygen in the Ni{single bond}Ce boundary, which leads to the complete oxidation of methane and the creation of active sites for the subsequent reforming reaction, whereas a slow decomposition of NiAl₂O₄ spinel results in a long induction period of the Ni/Al₂O₃ catalyst. Second, the coking resistance as well as the catalytic activity rely on the reactivity of lattice oxygen in the ceria. The ceria lattice oxygen participates in the reforming reaction through migration, which leads to the formation of oxygen vacancies. Subsequently, the vacancies are completely supplemented by oxygen arising from the dissociative adsorption of CO₂. Finally, the stability of Ni₁₀Ce₉₀ relies on the balance between the rate of generation and supplement of oxygen vacancies. Unbalanced rates of these processes result in deactivation.

Full text of DOI:10.1016/j.jcat.2007.01.011

Journal of Catalysis 247 (2007) 101–111
www.elsevier.com/locate/jcat
Mechanistic study of the unusual catalytic properties
of a new Ni–Ce mixed oxide for the CO2 reforming of methane
Do Kyoung Kim a,1, Klaus Stöwe a, Frank Müller b, Wilhelm F. Maier a,∗
a
Lehrstuhl für Technische Chemie, Universität des Saarlandes, Gebäude C 4.2, 66123 Saarbrücken, Germany
Institut für Experimentalphysik, Universität des Saarlandes, Gebäude C 6.3, 66123 Saarbrücken, Germany
b
Received 27 October 2006; revised 16 January 2007; accepted 20 January 2007
Abstract
The origin of the exceptional properties of a Ni Ce mixed oxide relative to a Ni/Al O catalyst (i.e., immediate activity without catalyst
10 90
2 3
prereduction, high resistance toward coking, high catalytic activity) and its catalytic deactivation for the CO reforming of methane was investi-
2
gated by XRD, XPS, pulse experimentation, and regeneration tests. Based on those results, a reaction mechanism is proposed. First, the catalytic
property with no prereduction step relies on the reactivity of oxygen contained in the Ni phase. The rapid activation of Ni Ce is based on
10 90
the active interfacial oxygen in the Ni–Ce boundary, which leads to the complete oxidation of methane and the creation of active sites for the
subsequent reforming reaction, whereas a slow decomposition of NiAl O spinel results in a long induction period of the Ni/Al O catalyst.
2
4
2 3
Second, the coking resistance as well as the catalytic activity rely on the reactivity of lattice oxygen in the ceria. The ceria lattice oxygen partici-
pates in the reforming reaction through migration, which leads to the formation of oxygen vacancies. Subsequently, the vacancies are completely
supplemented by oxygen arising from the dissociative adsorption of CO . Finally, the stability of Ni Ce relies on the balance between the rate
2
10 90
of generation and supplement of oxygen vacancies. Unbalanced rates of these processes result in deactivation.
© 2007 Elsevier Inc. All rights reserved.
Keywords: Methane reforming; CO ; Ceria; Nickel catalysts; Ni–Ce interface; Autoactivation; Oxygen storage and release capacity (OSC); Pulse study; Sol–gel;
2
Heterogeneous catalysis
1. Introduction
though a large parameter space was sampled (more than 5000
mixed oxides containing 61 elements), no totally new catalyst
was discovered. The study was conducted against a Rh/alumina
reference catalyst, and rare noble metals were excluded. The
best catalysts discovered were Ni-based mixed oxides; among
these, a new Ni10Ce90 mixed oxide catalyst was identified that
exhibits several unusual features: (1) lack of coking; (2) imme-
diate activity (no need for prereduction), and (3) high catalytic
activity. Continuous deactivation of the Ni10Ce90 catalyst could
be strongly reduced by doping with 15 mol% Al or Zr. This is a
significant improvement to a state-of-the-art Ni/Al2O3 catalyst,
which tends to coke and deactivate and requires a prereduction
step to reach the desired catalytic activity. Because of these un-
usual properties, it was of interest to investigate the nature of
this ceria-based Ni catalyst in more detail.
For applications of reforming catalysts in mobile fuel cells,
high catalytic activity, no tendency toward coking, immedi-
ate catalyst startup, and long stability are desirable proper-
ties [1–3]. State-of-the-art Ni-catalysts, such as Ni on alumina,
show a tendency for coking and deactivation and require a
prereduction time to reach catalytic activity. The quest for al-
ternative catalysts is difficult, because the possible number of
compositions is almost infinite [4].
A previous study [3] used high-throughput synthesis and
screening techniques in the search for and optimization of new
autoactivation catalysts for the CO2 reforming of methane. Al-
*
Corresponding author.
Although cerium is one of the rare earth elements, the
generic term “rare earth” can be misleading, because cerium
is the 25th most abundant element in the earth’s crust [5]. It
E-mail address: w.f.maier@mx.uni-saarland.de (W.F. Maier).
Current address: Department of Chemical Engineering, University of Cali-
1
fornia, Berkeley, CA 94720, USA.
0021-9517/$ – see front matter © 2007 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2007.01.011

102
D.K. Kim et al. / Journal of Catalysis 247 (2007) 101–111
is well known that in cerium oxide the elimination of lattice
oxygen under reducing conditions leads to the creation of an-
ionic vacancies (CeO2−x) correlated to a high lattice oxygen
mobility, and such oxygen vacancy defects are proposed to be
the reactive sites on the surfaces of metal oxides [6]. Moreover,
the nonstoichiometric CeO2−x (0 ꢀ x ꢀ 0.28) easily restores its
oxygen deficiency under oxidation conditions to form stoichio-
metric CeO2 [7–9]. Recently, the structure, distribution, and
formation of oxygen vacancies on a ceria surface were studied
by means of scanning tunneling microscopy (STM) and density
functional theory (DFT), and it was noted that the linear clusters
of the surface oxygen vacancies formed on CeO2 (111) expose
exclusively Ce3+ ions to the gas-phase reactants [10]. Ceria is
an important component of three-way catalysts in automobile
applications, where it provides oxygen storage and release ca-
pacity (OSC) that enables control of the oxygen stoichiometry
in the exhaust [7–9]. The same properties account for the ex-
ceptional properties of ceria-based oxides in various reactions,
including steam reforming, hydrogenation, water–gas shift, and
CO or hydrocarbon oxidation [7,11–15].
For comparison, a 10 mol% Ni-containing catalyst sup-
ported on commercial γ -Al2O3 (Johnson Matthey, SBET
=
255 m2 g−1), designated Ni/Al2O3, was prepared by the wet-
impregnation method as follows. After 1.268 g Ni(NO3)2·6H2O
in 10 ml water was stirred for 30 min in a flask, 2.0 g γ -Al2O3
was added into the solution. Then the mixture in the flask was
stirred and evaporated at 50 ◦C in a rotary evaporator system
(IKA Labortechnik). After the solvent was removed, the wet
powder was dried in an oven at 120 ◦C for 24 h. The catalyst
powder thus obtained was ball-milled for 30 min, then calcined
in an oven, heated to 700 ◦C at a rate of 2 ◦C min−1, and kept
there for 2 h.
2.2. Reactivity testing of catalysts
Catalytic performance tests were carried out in a gas-phase
quartz reactor (7 mm i.d., 400 mm long) at atmospheric pres-
sure. Typically, 100 mg of the sample in fine-powder form
(sieved fraction with 25–50 µm) was placed in the center of
the reactor and held by quartz wool plugs. The reactor tube
was heated inside an electric furnace. A quartz tube-sheathed
K-type thermocouple was placed at the top of the catalyst bed to
monitor the reaction temperature. The reactant gas [52.8 vol%
CH4 (quality 3.5) and 47.2 vol% CO2 (4.8)], diluted with ar-
gon (5.0), was used without further purification. The follow-
ing experimental conditions were typical for a single run. The
catalyst was oxidized or reduced in situ at 600 ◦C for 2 h in
flowing air or hydrogen (50 ml min−1). After purging with Ar
(50 ml min−1) for 30 min, the gas flow was switched to an up-
stream reactant gas mixture (CH4/CO2/Ar = 16.9/15.1/68; total
flow, 65 ml min−1). The gas hourly space velocity (GHSV) of
each catalyst was calculated on the basis of the volume of the
gas contacted with the catalyst per hour at standard conditions
divided by the bulk volume of the catalyst particles. The vol-
ume of each catalyst was calculated as the ratio of the catalyst
weight to its packed density determined before the experiments.
The GHSV of Ni10Ce90 was 139,000 h−1, and that of Ni/Al2O3
was 36,000 h−1. The gas composition of products was ana-
lyzed by gas sensors (GfG-mbH) with ranges of 0–20 vol%
CH4, 0–25 vol% CO2, and 0–10 vol% CO, along with a micro-
gas chromatograph (model CP 4900, Varian). Each sample was
tested for 10 h monitored with 60 GC measurements. A blank
reaction test without catalyst was carried out to confirm that the
reactor and thermocouple itself were not catalytically active.
For regeneration tests, the gas-phase composition over
Ni10Ce90 was analyzed for 150 min by GC after every 10 min
within a measurement run. After the reaction test, the catalyst
was regenerated in situ at 600 ◦C for 2 h in a flow of air, Ar,
or H2 (50 ml min−1). After purging with Ar (50 ml min−1) for
30 min, the reaction test was conducted again. The regeneration
cycle was performed once more.
In the present study, the origin of the exceptional proper-
ties of a Ni10Ce90 mixed oxide relative to a Ni/Al2O3 catalyst
(i.e., immediate activation in the absence of catalyst prereduc-
tion, high resistance toward coking, and high catalytic activity)
and its catalytic deactivation for the CO2 reforming of methane
was investigated by XRD, XPS, pulse experiments, and regen-
eration tests.
2. Experimental
2.1. Catalyst preparation
The NixCe100−x (x = 0, 10, 20, 100) mixed oxides were
prepared by the modified sol–gel method as described in de-
tail previously [3]. The molar ratio of metal (1 M nitrate
solution in methanol):complexing agent (4-hydroxy-4-methyl-
pentanone):acid (propionic acid) was 100:300:2. For example,
in the preparation of Ni10Ce90, 3.7 ml (30 mmol) of complex-
ing agent was placed in a 100-ml glass beaker. Then 9.0 ml
(9 mmol) of 1 M Ce(NO3)3·6H2O solution in methanol and
1.0 ml (1 mmol) of 1 M Ni(NO3)2·6H2O solution in methanol
were added, and the resulting solution was magnetically stirred
for 30 min, after which 15.0 µl (0.2 mmol) of propionic acid was
added dropwise under continuous stirring. The solution was
stirred for another 3 h while covered with Parafilm, to obtain
a homogeneous solution. After the Parafilm was removed, the
solution was transferred to a wide, shallow beaker and kept for
5 days at 40 ◦C for gelation. Then it was calcined in an oven at
65 ◦C for 5 h (at a heating rate of 0.2 ◦C min−1) and at 250 ◦C
for 5 h (at a heating rate of 0.2 ◦C min−1). The catalyst pow-
der thus obtained was manually ground in an agate mortar for
30 min, then recalcined in an oven, heated to 700 ◦C at a rate
of 2 ◦C min−1, and kept there for 2 h. Cerium oxide and nickel
oxide were prepared by the same method. An artificial sample
for XPS measurements was set by physical mixing of the nickel
oxide and cerium oxide (the molar ratio of Ni:Ce = 1:9).
2.3. Pulse experiment
The transient pulse experiments were performed in a gas-
phase quartz reactor (3 mm i.d., 520 mm long) at atmospheric
pressure. Typically, the sample in fine powder form (25–50 µm)
was placed in the center of the reactor and held by quartz wool

D.K. Kim et al. / Journal of Catalysis 247 (2007) 101–111
103
plugs. The reactor tube was heated inside an electric furnace.
Table 1
Catalytic behavior and coking rate on Ni/Al O and Ni Ce depending on
2
3
10 90
The temperature of the catalyst bed was monitored by a ther-
mocouple attached to the outside of the reactor tube. Before
the pulse data were obtained, the catalyst was treated in situ at
600 ◦C for 150 min in flowing Ar (30 ml min−1). During the
pulse experiment, gas pulses [CH4 (quality 3.5), CO2 (4.8), or
H2 (5.0)] with a volume of 100 µl (ca. 4.07 µmol) were passed
at 600 ◦C over the catalyst sample using Ar (5.0) as carrier gas
(30 ml min−1). The time between consecutive pulses was 5 min.
The pulses of a different gas were introduced after stabiliza-
tion for 30 min in flowing Ar (30 ml min−1) at 600 ◦C. The
products were monitored on-line by quadrupole mass spectrom-
etry (QMS, GSD 300 T2, Balzers). The argon intensity of each
pulse was chosen for internal standardization as representative
within the chronological progress recording the detection sen-
sitivity over the whole experiment. The normalized areas with
respect to argon intensity under a peak were integrated, and then
the integrated area was converted to a molar fraction, using a
conversion factor determined after the experiments. The areas
under the peaks were reproducible within about 8%. The con-
version factors were determined under experimental conditions
as close as possible to the product gas compositions in a blank
reactor at 600 ◦C.
the different pretreatments
a
b
c
Catalyst
Treatment
Conversion (%)
10 min 600 min
CH CH
TGA coke
−1
−1
(g
C
g
h
× 100)
cat
CO
CO
2
4
2
4
d
d
Ni/Al O
2
Oxidation
Reduction
Oxidation
Reduction
4.6
7.9 30.7
38.8 48.5 38.0
32.3 48.7 27.6
36.9
45.4
38.3
6.6
2.28
1.00
0.08
0.01
3
Ni Ce
10 90
6.2 15.2
4.9
a
◦
−1
Pretreated in situ at 600 C for 2 h in flowing air or hydrogen (50 ml min ).
b
◦
−1
Operating conditions: T = 600 C; p = 1 bar; total flow 65 ml min
(CH /CO /Ar = 16.9/15.1/68); catalyst weight m = 0.100 g.
4
2
c
Operating conditions: catalyst weight m ≈ 0.050 g; T increasing rate
◦
−1
−1
15 C min ; air total flow 15 ml min
d
.
Measured after 380 min due to reactor clogging.
the program TOPAS [16] using the fundamental parameters ap-
proach [17].
X-ray photoelectron spectra (XPS) were recorded by a
VG ESCALAB MK II spectrometer using AlKα radiation
(1486.6 eV). The reaction or reduction of samples was per-
formed in a gas-phase quartz reactor with an on/off valve on
the top and bottom of the reactor. The reactor with the sample
under Ar was transferred to a glove box, in which the sample
in powder form was mixed with zirconium (about one-third of
total weight) to compensate for charging of the nonconducting
samples during the measurement, and pressed into a pellet. The
resulting pellet was placed in a transfer chamber with a stain-
less steel sample holder under the inert conditions to avoid any
exposure to the atmosphere before the analysis. The XPS data
from the regions related to the C 1s, O 1s, Ce 3d, and Ni 2p
core levels were recorded for each sample. The binding ener-
gies were adjusted relative to C 1s at 284.5 eV.
2.4. Catalyst characterization
Total BET surface areas were measured by nitrogen adsorp-
tion (Sorptomatic 1990, Carlo Erba) at the temperature of liquid
nitrogen. The samples were outgassed overnight under vacuum
at 200 ◦C before adsorption.
Temperature-programmed oxidation (TPO) was done by
thermogravimetric analysis (TGA) using a Shimadzu TGA-50
thermogravimetric analyzer. After 10 h of reaction time, the cat-
alyst sample was cooled to room temperature in a flow of Ar
and then transferred to the TGA apparatus. About 50 mg of
sample was loaded in a sample pan. The weight change of the
catalyst was recorded in a flow of air (15 ml min−1) as temper-
ature was increased from room temperature to 900 ◦C at a rate
3. Results and discussion
3.1. Catalyst performance and carbon deposition
of 15 ◦C min−1
.
The chemical composition of each catalyst was verified by
XRF. According to BET measurements of each material after
calcination at 700 ◦C, fresh ceria (19.7 m2 g −1), nickel oxide
(29.4 m2 g−1), and Ni10Ce90 mixed oxide (36.9 m2 g −1) had
a low surface area, and γ -Al2O3 (255.0 m2 g−1) and Ni/Al2O3
(205.9 m2 g−1) had a relatively high surface area.
The chemical composition of catalysts was analysed semi-
quantitatively by standardless X-ray fluorescence (XRF) spec-
troscopy (EAGLE II μ-Probe, EDAX) with RhKα radiation
using the fundamental parameter model. The X-ray beam was
focused to a spot size of 300 µm on the sample surface.
Structural characterization of the N10Ce90 catalysts was per-
formed by powder X-ray diffraction (XRD) with a Huber G
670 Guinier-camera using CuKα1 radiation (λ = 1.54056 Å).
The powder diffraction patterns of the Ni/Al2O3 catalysts were
recorded by a D5000 diffractometer (Bruker–Nonius Corp.)
in focusing Bragg–Brentano reflection geometry with a Cu
fine-focus tube (U = 40 kV, I = 25 mA) and a Johannson-
type quartz monochromator [λ(CuKα1) = 1.54056 Å]. The dif-
fracted intensity was measured by a linear position sensitive
detector with Ar/10% methane filling at 8.5 bar pressure and an
aperture of 10 mm to improve spatial resolution. In step-scan
mode, the step width was ca. 0.015◦ in 2θ. Lattice parameter
refinement was done by full-pattern Rietveld refinement with
Equilibrium conversions of CH4 and CO2, calculated by the
equiTherm software [18], were 53.1 and 69.4%, respectively,
with a H2/CO ratio of 0.85 at 600 ◦C. Only at 800 ◦C the equi-
librium conversions reached almost total consumption (87.8%
CH4 and 99.1% CO2, respectively) with a H2/CO ratio of unity.
Catalytic activity tests of Ni10Ce90 and Ni/Al2O3 were per-
formed for the CO2 reforming of methane after in situ oxidation
or reduction at the reaction temperature (600 ◦C). The reac-
tion conditions (residence time, catalyst amount) were chosen
to stay in the kinetic regime by keeping the conversion signifi-
cantly below thermodynamic equilibrium.
Table 1 summarizes the catalytic performance and tendency
toward coking on Ni/Al2O3 and Ni10Ce90 depending on the dif-

104
D.K. Kim et al. / Journal of Catalysis 247 (2007) 101–111
ferent pretreatments. The oxidized Ni/Al2O3 showed low initial
activity that did not start to increase until after almost 80 min,
with catalytic activity stabilizing only after about 240 min
(long induction period) [3]. Moreover, Ni/Al2O3 led to reac-
tor plugging after only 380 min due to strong carbon deposi-
tion (2.28 wt% h−1) [3]. The prereduced Ni/Al2O3 revealed an
immediate activity and relatively high rate of coke formation
(1.00 wt% h−1), although reactor plugging was not detected.
For the Ni/Al2O3 catalyst, a prereduction step or long induc-
tion period was required to reach catalytic activity, and carbon
deposition was favored regardless of the pretreatment condi-
tions. In strong contrast, the oxidized Ni10Ce90 exhibited a
rapid startup operation and high coking resistance (0.08 wt%
h
−1) [3], whereas prereduced Ni10Ce90 demonstrated low ac-
tivity and negligible tendency toward coking (0.01 wt% h−1).
The oxidized Ni10Ce90 was much more active than the reduced
Ni10Ce90, and both samples showed high resistance to carbon
deposition and a continuous decrease of conversions. Wang and
Lu reported that in catalytic activity tests after prereduction,
Ni/CeO2 showed much lower activity and carbon deposition
relative to Ni/Al2O3 [19], in agreement with our results, pre-
sented in Table 1.
(a)
The strong tendency toward carbon deposition of Ni/Al2O3
also could originate from compacting problems of the high-
surface area Al2O3, leading to a pressure increase. But this pos-
sibility was excluded, because testing of 400 mg of Ni10Ce90
(GHSV 35,000 h−1, with the same bed volume and particle size
distribution as that of Ni/Al2O3) under identical conditions also
demonstrated high coking resistance.
To elucidate the effect of treatment conditions (oxidation,
reduction, or inertness) on the catalytic activity in more detail,
we tested Ni10Ce90 catalysts with different sequences of regen-
eration procedures, as shown in Fig. 1. As shown in Fig. 1a,
fresh Ni10Ce90 (after in situ oxidation) always exhibited an im-
mediate high catalytic activity, which then slowly decreased.
After about 150 min of reaction time, the catalyst was regener-
ated with air, in the other two experiments with Ar, and in the
last with hydrogen; the regeneration time was always 30 min
(not shown in Fig. 1). The CO2-conversion results, depicted in
Fig. 1b, also prove that the treatment with air led to full recovery
of the initial activity. Deactivation continued after Ar treatment.
H2 treatment led to a dramatic reduction of catalytic activity.
After another 150 min of reaction time, the catalysts were
regenerated again (see Figs. 1aC and 1bC). When the air-
regenerated catalyst was regenerated with air again, most of
its activity was recovered. The catalyst regenerated originally
with Ar recovered most of its original activity after regeneration
with air, whereas deactivation continued when it was regener-
ated again with Ar. The sample deactivated by the regeneration
with hydrogen recovered only little of its original activity by
the final air regeneration.
(b)
Fig. 1. Time dependence of the catalytic activity and stability of 100 mg
Ni Ce in a variety of regeneration conditions: (a) CH conversion and
10 90
4
◦
(b) CO conversion. Operating conditions: T = 600 C; P = 1 bar; total flow
2
−1
65 ml min (CH /CO /Ar = 16.9/15.1/68). The different regeneration or pre-
4
2
treatment conditions are stated in the legend. (A), (B) and (C) indicate succes-
sive experiments.
of reaction. Fig. 2 shows the XRD pattern of as-prepared, re-
duced, and tested Ni on γ -Al2O3 in relation to that of the pure,
undoped support. The XRD pattern of as-prepared Ni/Al2O3
shows main peaks at 2θ = 37.3◦, 45.5◦, and 65.3◦, correspond-
ing to the reflections (311), (400), and (440) of cubic Al2O3.
The absence of reflections at 2θ = 43.3◦ and 63.0◦ indicates
that no NiO was formed in the Ni/Al2O3 catalyst. A possible
reaction product of the wet-impregnation method is the spinel-
phase NiAl2O4, which crystallizes isotypically to the γ -Al2O3
support in a fcc oxygen packing with partial occupation of
voids. As a result of this structure relation, a solid solution
between the two boundary phases is formed, which can be de-
scribed by the formula Ni3xAl8−2xO12 (0 ꢀ x ꢀ 1). Due to the
very similar lattice dimensions, the composition of the product
formed cannot be deduced from inspection of the diffraction
pattern alone; thus, a Rietveld refinement of the complete dif-
fraction pattern would seem to be advisable. On the other hand,
3.2. Characterization of Ni/Al2O3 catalyst
To see the effect of catalyst structure on the activity, XRD
measurements were performed for the Ni/Al2O3 catalysts in as
prepared state, after 2 h of reduction, and after 30 and 380 min

D.K. Kim et al. / Journal of Catalysis 247 (2007) 101–111
105
Fig. 3. XRD patterns (a) of the as prepared artificial sample (physical mixing
of nickel oxide and cerium oxide both prepared by the sol–gel method, the
molar ratio of Ni:Ce = 1:9); (b) after 2 h reduction of sample (a); (c) of as
Fig. 2. XRD patterns of as prepared, reduced and tested Ni/Al O relative to
2
3
the fine γ -Al O support: (a) γ -Al O , (b) as prepared Ni/Al O , (c) reduced
2
3
2
3
2 3
prepared Ni Ce ; (d) after 10 h reaction of sample (c); (e) after 2 h reduction
10 90
Ni/Al O (after 2 h reduction of sample (b)), and (d) tested Ni/Al O (after
2
3
2 3
of sample (c); (f) after 10 h reaction of sample (e).
380 min reaction of sample (b)).
tivation period is attributed to the continuous reduction of the
Ni3xAl8−2xO12 species by methane, which can be promoted by
product gas (hydrogen) according to the following equation:
the diffraction patterns shown in Fig. 2 are characterized by
very broad reflections due to the poor crystallinity and small
crystallite size of the calcined reaction product, which affected
all results from Rietveld refinements because of the low sig-
nificance level of the refinement results. A major difference
between the diffraction pattern of NiAl2O4 and γ -Al2O3 is the
intensity of the (222) reflection at about 2θ = 39.4◦ (compare
Fig. 2, patterns a and b). Based on the low intensity of the (222)
reflection in the diffraction pattern of the reaction product, we
deduce a spinel formation according to the formula given above
with undetermined Ni content in agreement with previous in-
vestigations [20].
Ni3xAl8−2xO12 + 3xH2 → 3xNi + (4 − x)Al2O3 + 3xH2O.
XRD characterization of Ni/Al2O3 indicates a correlation be-
tween catalytic activity change and Ni oxidation state, and the
long induction period of Ni/Al2O3 catalyst is due to a slow de-
composition of NiAl2O4 spinel to metallic nickel and Al2O3.
Severe coke formation on the tested Ni/Al2O3 was supported
by XRD (Fig. 2d) and TPO studies (see Table 1). Pulse ex-
periments for nickel oxide (see Section 3.4) demonstrated that
reduced nickel decomposes methane to H2 and carbon, and the
surface carbon on metallic nickel is not removed by CO2. In ad-
dition, it is known that γ -Al2O3 is an irreducible support which
does not provide lattice oxygen [21]. Thus, a lack of oxygen
supplied from gas phase or catalyst support capable of elimi-
nating the surface carbon on Ni0 contributes to the high coking
tendency of Ni/Al2O3.
For the reduced Ni/Al2O3 catalyst, a new peak at 2θ = 51.8◦
as well as a line broadening at the lower angle side of the spinel
(400) reflection at 2θ = 45.5◦ is interpreted as formation of
metallic nickel. Analysis of a diffraction pattern for the sam-
ple after the reaction of fresh catalyst revealed no significant
change after 30 min of reaction (not shown) compared with the
as-prepared Ni/Al2O3. After 380 min of reaction, the peak at
2θ = 51.8◦ and line broadening at the lower angle side of the
spinel (400) reflection at 2θ = 45.5◦ were detectable similarly
as in the reduced catalyst, and a very broad diffraction inten-
sity was seen around 2θ = 26.4◦, assigned to the formation of
carbon. It is well known that carbon is readily deposited on the
Ni/Al2O3 catalyst under our experimental conditions, as was
demonstrated in the TPO study (see Table 1).
In catalytic testing, the oxidized Ni/Al2O3 maintained a low
initial activity until almost 80 min. In the XRD pattern of the
Ni/Al2O3 catalyst, the nickel in as-prepared state was present
in the form of NiAl2O4 spinel, which maintained its state for
at least 30 min of reaction. The Ni/Al2O3 catalyst reduced for
2 h or tested for about 240 min demonstrated high and sta-
ble catalytic conversions. Here the Ni oxidation state was Ni0,
which represents a presumably active center for the reforming
reaction. Thus, the slow increase of conversion during the ac-
3.3. Characterization of Ni–Ce mixed oxide
XRD measurements were performed for the Ni10Ce90 cat-
alyst in its as-prepared state, after 2 h of reduction, and after
30 and 600 min of reaction (Fig. 3, patterns c–f). The patterns
are dominated by the diffraction peaks of the ceria substrate,
revealing a much lower peak-width FWHM, that is, a much
better crystallinity and larger crystallite size compared with
γ -Al2O3, but lower crystallinity than the pure oxide as pre-
pared by the sol–gel method. The reflections of oxidized nickel
(I100 at 2θ = 43.4◦) and metallic nickel (I100 at 2θ = 44.5◦) on
the catalysts are very weak; this, along with the line broaden-
ing, indicate high dispersion of the Ni. The small metal par-
ticles and their high dispersion in Ni10Ce90 are attributed to
an intrinsic result of the acid-catalyzed sol–gel preparation,
which is known to provide atomically dispersed mixed oxides

106
D.K. Kim et al. / Journal of Catalysis 247 (2007) 101–111
[22,23]. Moreover, a linear decrease of the CeO2 lattice parame-
ter with increasing Ni amount was identified for the pure CeO2
a = 5.4135(1) Å, as-prepared Ni10Ce90 a = 5.4124(3) Å, and
Ni20Ce80 a = 5.411(1) Å, which follows Vegard’s law [24].
This suggests an oxide solid solution was formed in the Ni–
Ce mixed oxide system with incorporation of some Ni2+ ions
into the ceria lattice, in agreement with previous investigations
[25,26]. Formation of the solid solution is attributed to the acid-
catalyzed sol–gel process used for synthesis of the material.
The lack of a carbon peak (at 2θ = 26.4◦) in the XRD patterns
of tested samples confirms the high coking resistivity of this
Ni10Ce90 mixed oxide, which is consistent with the results of
our TPO study (see Table 1).
A control experiment with identical composition (shown in
Fig. 3, patterns a and b) found that metallic Ni and Ni oxide can
be clearly identified when present as bulk materials. For fresh
CeO2 and NiO prepared by the sol–gel method, the XRD pat-
tern revealed a fluorite-type structure [a = 5.4135(1) Å] and a
cubic rock-salt structure [a = 4.1793(1) Å], respectively. These
data were used as reference. The XRD patterns of the artificial
sample (physical mixing of the nickel oxide and cerium oxide
prepared by the sol–gel method, at a molar ratio of Ni:Ce = 1:9)
in as-prepared and reduced state are shown in Fig. 3, patterns a
and b. Clearly, the reflections of oxidized nickel (i.e., NiO [I100
at 2θ = 43.4◦]), on the as-prepared sample and metallic nickel
(I100 at 2θ = 44.5◦) on the reduced sample are detectable, con-
firming the foregoing conclusions on the nature of the Ni in the
mixed oxide.
Fig. 4. Ce 3d XPS spectra for (a) as prepared Ni Ce , (b) reduced Ni Ce
20 80
20 80
(after 2 h reduction of sample (a)), and (c) tested Ni Ce (after 2 h reaction of
20 80
sample (a)). The spectra for Ce O (d) and CeO (e) are taken from Ref. [28],
2
3
2
the original data were published in Ref. [27].
XPS was used to obtain further information about the va-
lency/oxidation state of elements and surface composition of
the Ni–Ce mixed oxide by inspecting the spectral line shape and
the intensities of the Ce 3d and Ni 2p core-level electrons. Due
to the surface sensitivity of XPS, special precautions were taken
to avoid contact of the samples with air during introduction into
the XPS analysis chamber. In addition, the samples were mixed
with zirconium to reduce charging of the nonconducting sam-
ples during the measurement, as described in Section 2. Despite
the comparable photoemission cross-sections of Ce 3d and Ni
2p, the intensity of Ni 2p dropped below the nominal 10:90 ra-
tio for the Ni10Ce90 catalysts; due to the very weak intensities,
even a qualitative interpretation of the Ni 2p line shape was not
possible. Even in the case of Ni20Ce80 catalysts with the higher
Ni/Ce ratio, Ni 2p intensity was very weak. Fig. 4 compares
the Ce 3d spectra of the Ni20Ce80 catalysts in as prepared state,
after reduction for 2 h, and after reaction for 2 h with the Ce
3d spectra of Ce2O3 and CeO2 [27,28]. According to literature
data [25,29,30], the CeO2 spectrum (representing Ce4+) is la-
belled with v, vꢀꢀ, vꢀꢀꢀ and u, uꢀꢀ, uꢀꢀꢀ, describing the 3d5/2 and
3d3/2 photoelectron intensities for 4f2, 4f1, and 4f0 occupation,
and vꢀ and uꢀ, representing the lower binding energy intensities
of the 3d5/2 and 3d3/2 electrons for the 4f1/4f2 configuration
in Ce2O3 (representing Ce3+). Comparing the observed 3d en-
velope with those previously reported for Ce4+ and Ce3+ [25,
27–30] shows that the Ce 3d spectra of the catalysts in Fig. 4,
a–c, are very similar to the Ce 3d spectrum of CeO2 (cf., e.g.,
the strong 3d3/2 (4f0) contributions (uꢀꢀꢀ), which are missing
in Ce2O3 [28]); therefore, the Ce is mainly of Ce4+ charac-
ter. But in Fig. 4, b and c (i.e., after reduction as well as after
reaction), the spectra also exhibit clearly visible vꢀ and uꢀ contri-
butions, indicating that the catalysts also contain small amounts
of Ce3+. Thus, the surfaces of both the reduced and used sam-
ples are formed by a mixture of Ce4+ and Ce3+, whereas the
surfaces of the fresh catalysts seem to be mainly formed by
Ce4+. This supports the involvement of lattice oxygen in the
reforming reaction.
From XPS analysis, as a surface-sensitive technique (with a
probing depth of about 1–2 nm), the distributions of Ce and Ni
seem to be not homogeneous within the catalyst, at least within
an area of a few atomic layers. Because of the nearly vanishing
Ni intensities (not shown), the probed surface area of a single
grain must be strongly enriched with Ce, or the inner core of
the grain must be enriched with Ni.
A slow but continuous deactivation during the reaction and
dramatic loss of catalytic activity after hydrogen treatment were
observed for the Ni10Ce90 catalyst. On BET measurements, the
surface area of Ni10Ce90 showed no obvious change (within ex-
perimental error) after 10 h of reaction (a decrease from 36.9
to 35.2 m2 g−1) or 2 h of reduction (an increase from 36.9
to 38.2 m2 g−1), indicating that no sintering effect occurs un-
der our conditions. For the used or reduced Ni10Ce90 catalyst,
XRD patterns of oxidized nickel and metallic nickel were still
very weak compared to those of the fresh catalyst (see Fig. 3).
This does not support significant Ni phase sintering as potential
cause of catalyst deactivation. Moreover, XRD characterization
(Fig. 3) and TPO study (Table 1) of tested Ni10Ce90 exhibited

D.K. Kim et al. / Journal of Catalysis 247 (2007) 101–111
107
Table 2
Table 3
a
Conditions of pulse experiments
The amount of oxygen consumed during the pulse experiments
Catalyst
Empty
m
(mg)
Pulse sequence
Pulse No.
Catalyst
m
Theor.
oxygen
amount
(×10 mol)
Consumed oxygen amount
for complete oxidation of the
Pulse
No.
cat
0
cat
(mg)
CH –CO
15–15
15–15
8–8–7
15–15
15–19–15
15–15–15
20–20–20
15–15–15
4
2
2
−5
b
reducing agent (×10 mol)
CeO
100
100
20
CH –CO
2
4
−5
a
H –CH
2
Sum
H –CH –CO
4
2
4
2
NiO
CH –CO
4
2
NiO
20
26.8
6.1 8.7
5.6 0.9
6.5
14.8 15–19
20
H –CH –CO
H –CH –CO
2 4
2
4
2
2
4.60 6.2
6.4
6.5
6.6
15–15
0–20
15–15
4.60
Ni Ce
10 90
100
100
6.2
6.2
Ni Ce
10 90
100
100
CH –CO –CH
4 2 4
5.7 0.9
H –CH –CO
2
2
4
a
The amount of oxygen contained in Ni–O is calculated on the basis of the
a
◦
−1
Operating conditions: T = 600 C; p = 1 bar; Ar flow 30 ml min ; pulse
catalysts weight.
volume 100 µl (4.07 µmol).
b
The amount of consumed oxygen for complete oxidation is calculated on the
basis of the hydrogen consumption for H pulses and the CO production for
2
2
CH pulses.
4
negligible coke formation. Therefore, the deactivation observed
during the reaction or reduction is more likely due to a loss
of active sites. Because the XRD reflections corresponding to
the Ni phase of the Ni10Ce90 catalyst were very weak, together
with line broadening due to its very small particles of high dis-
persion or the formation of Ni–Ce solid solution, the oxidation
state of nickel could not be determined from the XRD measure-
ment alone. From XPS analysis, the mixture of Ce4+ and Ce3+
was identified in the surface of both reduced and tested Ni–Ce
mixed oxide, but the valance state of nickel was unclear due
to its very weak intensities. To shed more light on the species
of active site, we carried out a series of pulse experiments, as
reported next.
3.4.2. Cerium oxide
Application of CH4 or CO2 pulses to 100 mg ceria pro-
duced no detectable product formation exceeding background.
Attempts to activate the ceria by H2 pulses also failed. This in-
dicates a negligible interaction of ceria with CH4, CO2, or H2 at
600 ◦C, which is in perfect agreement with literature data. The
oxidation of CH4 by CeO2 becomes thermodynamically viable
only at temperatures above 650 ◦C [35], and the reduction peak
of pure CeO2 in the H2-TPR profile is observed at about 830 ◦C
[26]).
3.4.3. Nickel oxide
An experiment involving application of 15 CH4 pulses fol-
lowed by 15 CO2 pulses on 20 mg of nickel oxide again re-
vealed no detectable catalytic activity, whereas 15 pulses of
H2 over the NiO resulted in quantitative conversion. This indi-
cates that hydrogen is a stronger reducing agent for NiO than
methane. The amount of oxide oxygen calculated in 20 mg
NiO was 26.8 × 10−5 mol, and the amount of oxygen con-
sumed by the H2 pulses was 6.1 × 10−5 mol for quantitative
conversions of the reducing agent, as summarized in Table 3.
Rodriguez et al. [36] evaluated the mechanism of NiO reduc-
tion by hydrogen and found that crystalline or powdered forms
of NiO are directly reduced to metallic nickel at 400–460 ◦C
without detectable formation of intermediate phases. When the
partly reduced NiO was exposed to additional 19 pulses of
CH4, H2O and CO2 were produced with no decrease in activ-
ity and negligible formation of H2 and CO, consuming another
8.7 × 10−5 mol of the remaining oxygen in the NiO. When an
additional 15 CO2 pulses were subsequently applied to this re-
duced Ni0/NiO, no products were detectable, implying that the
partially reduced nickel oxide cannot dissociate CO2 for com-
plete or partial reoxidation.
The experiment was repeated with only 4.6 mg of NiO,
corresponding to the amount of nickel oxide in 100 mg of
Ni10Ce90. The amount of oxide oxygen calculated in 4.6 mg
of NiO was 6.2 × 10−5 mol. Fig. 5 gives the amounts of prod-
ucts during the experiment of consecutive 15 H2–15 CH4–
15 CO2 pulses. In Fig. 5, the amount of water released is de-
termined indirectly by the amount of H2 consumed in panel A
and by [(the amount of CH4 consumed)−0.5×(the amount of
H2 produced)] × 2 in panel B.
3.4. Pulse experiments
3.4.1. General
Pulse experiments have been useful tools in mechanistic
studies of methane dry reforming on prereduced Pt/ZrO2 [31,
32], Pt/Al2O3 [32], Pd/CeO2 [33], Pd/ZrO2 [33], Rh/Al2O3
[34], and fresh CeO2 [35] catalysts. In preliminary studies on
the Ni10Ce90 catalyst, in situ oxidation at 600 ◦C for 2 h re-
sulted in an instable material, with mass spectrometry response
signals of nitrogen and oxygen stabilizing only after about 7 h
of stationary experiment. Because no difference in activity be-
tween in situ air or Ar pretreatment at 600 ◦C was found in 10-h
reaction tests (data not shown), all samples were treated in flow-
ing Ar for 150 min before pulse experiments. Thus, in the CH4
or H2 pulsed experiments, the catalyst itself was the sole source
of oxygen. Among the products (i.e., H2, H2O, CO, and CO2),
a detection time of 5 min was not sufficient for H2O, due to
interaction with the walls of the capillary connecting the reac-
tor with the QMS. Therefore, the amount of water released was
determined indirectly based on the amount of educt pulses con-
sumed. A blank reaction test without catalyst confirmed that
the reactor itself was not catalytically active. The formation
of trace amounts of products during the blank tests (<0.5% of
that during the blank test of the product gases) is attributed to
thermal reaction and registered as background. The conditions
of pulse experiments are summarized in Table 2. The carbon
balance was determined by [(the sum amount of CO and CO2
produced)/(the amount of CH4 consumed)] for CH4 pulses and
[(the amount of CO produced)/(the amount of CO2 consumed)]
for CO2 pulses.

108
D.K. Kim et al. / Journal of Catalysis 247 (2007) 101–111
(a)
Fig. 5. The amount of products from 15 consecutive pulses of (A) H , (B)
2
CH , and (C) CO over 4.60 mg NiO. The amount of H O is determined in-
4
2
2
directly by the amount of H consumed in panel A and by [(the amount of
2
CH consumed) − 0.5 × (the amount of H produced)] × 2 in panel B. Oper-
4
2
◦
−1
ating conditions: T = 600 C; p = 1 bar; Ar flow 30 ml min ; pulse volume
100 µl (4.07 µmol).
With the prereduced NiO, H2 conversion began to decrease
after four pulses (see Fig. 5A), indicating that most, but not
all, NiO was reduced. An estimated 5.6 × 10−5 mol of oxygen
was consumed by the 15 H2 pulses, corresponding to approx-
imately 90% of the oxygen content of the NiO. When the 15
CH4 pulses were applied to this highly reduced NiO, only the
first 5 CH4 pulses produced small amounts of H2O and CO2
apart from H2 and CO, all of which then rapidly declined (see
Fig. 5B). The formation of H2O and CO2 is attributed to the
complete oxidation of methane with the oxygen remaining in
NiO, as described in the study of 20 mg of NiO. The sum
amount of oxygen consumption for the formation of H2O and
CO2 in Figs. 5A and 5B is 6.4 × 10−5 mol, as summarized in
Table 3. H2/CO ratios were about 3 until pulse 5. A H2/CO ra-
tio above 2 indicates carbon formation on the catalyst surface.
After pulse 6, hydrogen remained the only significant product.
Apparently, the remaining oxygen in the NiO was consumed by
the first five methane pulses, and methane decomposition dom-
inated on the reduced nickel to generate carbon and H2. When
the experiment was continued with CO2 pulses, no product was
detectable (see Fig. 5C). This proves that the carbon on metal-
lic nickel derived from methane decomposition does not react
with CO2 from the gas phase, further confirming that the metal-
lic nickel is incapable of dissociating CO2 for reoxidation of Ni
under these conditions. No interaction between the carbon on
metallic nickel and gas-phase CO2 can explain the formation
of nonremovable carbon on Ni/Al2O3 shown in Fig. 2d and Ta-
ble 1. This is quite different from the result with prereduced
Rh/Al2O3, where most of the surface carbon on Rh is success-
fully oxidized to CO by CO2 from the gas phase [34].
(b)
Fig. 6. (a) The amount of products from 20 consecutive pulses of (A) CH ,
4
(B) CO , and (C) CH over 100 mg Ni Ce . (b) The ratio of (2) H O/
2
4
10 90
2
CO , (") H /CO, (Q) carbon determined by [(the sum amount of CO and
2
2
CO produced)/(the amount of CH consumed)] for CH pulses, and (a)
2
4
4
carbon determined by [(the amount of CO produced)/(the amount of CO
2
consumed)] for CO pulses. The amount of H O is determined indirectly
2
2
by [(the amount of CH consumed)−0.5×(the amount of H produced)]×2.
4
2
◦
−1
Operating conditions: T = 600 C; p = 1 bar; Ar flow 30 ml min ; pulse vol-
ume 100 µl (4.07 µmol).
produced)] × 2. When the first eight CH4 pulses were intro-
duced to Ni10Ce90, H2O, and CO2 were the main products.
H2O/CO2 ratios of about 2 and carbon balances of unity were
found (see panel A of Fig. 6b), implying that the complete oxi-
dation of methane occurred without carbon formation. Further-
more, there seems to be a formation period, because the highest
activity is reached not until five pulses. Because this catalyst
was not reduced, the question of where this reaction occurs can
then be asked. Neither NiO nor CeO2 activates CH4, as shown
previously. Therefore, it is assumed that the interfacial oxygen
in the Ni–Ce boundary (Ni–O–Ce site) is the active component
for the complete oxidation of methane. This is not unreason-
able, because several studies point to the formation of M–O–Ce
surface complexes with metals such as Pd, Pt, Rh, and Au at the
interface [7,37–39]. As summarized in Table 3, 6.5 × 10−5 mol
of oxygen is consumed by total methane oxidation on Ni10Ce90,
which corresponds to about the total oxygen content of the
Ni–oxygen in the catalyst. Moreover, for the Ni20Ce80 mixed
3.4.4. Ni10Ce90 mixed oxide
The products from consecutive 20 CH4–20 CO2–20 CH4
pulses on 100 mg of Ni10Ce90 at 600 ◦C are depicted in Fig. 6.
In Fig. 6a, again the amount of H2O is determined indirectly
by [(the amount of CH4 consumed) − 0.5 × (the amount of H2

D.K. Kim et al. / Journal of Catalysis 247 (2007) 101–111
109
oxide, the amount of oxygen consumed by total methane oxi-
dation (15.5 × 10−5 mol) is increased by factor of 2.3 relative
to Ni10Ce90 [24].
After the eighth pulse, CO2 and H2O formation ceased,
and the product was dominated by H2 and CO, the formation
of the latter also continuously decreased during the follow-
ing pulses. Here the oxygen consumed must originate from
the ceria lattice. It is believed that storage oxygen in the ce-
ria is provided to the active Ni–Ce sites, which are now de-
pleted of the original oxygen, leading to the creation of oxy-
gen vacancies in the ceria. The amount of syngas formed
is conceivably based on the diffusion rate of lattice oxygen
correlated to oxygen concentration gradient. The H2/CO ra-
tios produced remained close to 2, and carbon balances were
about unity (see panel A of Fig. 6b). Therefore, no indica-
tion for carbon deposition was found on the Ni10Ce90 by CH4
pulses.
Fig. 7. The amount of products from 15 consecutive pulses of (A) H ,
2
(B) CH , and (C) CO over 100 mg Ni Ce . The amount of H O is
4
2
10 90
2
determined indirectly by the amount of H consumed in panel A and by
2
When the catalyst was subjected to CO2 pulses, CO2 was
converted exclusively into equivalent of CO, as shown in
Figs. 6aB and 6bB. The total amount of CO obtained from CO2
during the 20 pulses was 2.51×10−5 mol. Because CO2 pulses
resulted no activity with the as-prepared Ni10Ce90 (not shown)
or with reduced nickel (see Fig. 5c), the CO formation indi-
cates either that reduced Ni–Ce sites are reoxidized or that the
ceria oxygen vacancies created during the methane pulses 8–
20 are supplemented by CO2 dissociation and vacancy filling.
The subsequent 20 CH4 pulses (see below) seem to indicate
that the CO formation arises from reoxidation of the ceria oxy-
gen vacancies. Compared with the total amount of CO released
from the CH4 pulses in panel A (2.55 × 10−5 mol), the oxygen
vacancies are fully replenished by CO2 dissociation. Compar-
ing the CO production rate of the two processes (creation and
recovery of oxygen vacancies) in Fig. 6a demonstrates that the
supplementation of oxygen vacancies in panel B is clearly faster
than their generation in panel A, indicating that the oxygen
vacancies react very rapidly with CO2 to allow ready replenish-
ment.
The interaction of hydrogen with cerium-based oxides has
been widely studied, and ceria is known to be able to store hy-
drogen [25,40]. Therefore, the CO formation could have origi-
nated from the RWGS reaction between hydrogen stored in the
ceria and subsequent CO2 pulses. However, this effect was neg-
ligible here for two reasons. First, if hydrogen storage occurred
during the methane pulses, then the H2/CO ratios produced
should be <2, due to the additional consumption of H2. At
the same time, H2O should be detectable. However, the H2/CO
ratios produced remained close to 2, and the amount of H2O re-
leased was negligible after the 9th methane pulse. Second, if the
RWGS reaction occurred during the CO2 pulses, then H2O as
well as CO should be produced. However, as shown in panel B
of Fig. 6a, CO was the only product produced when CO2 pulses
were applied.
[(the amount of CH consumed) − 0.5 × (the amount of H produced)] × 2 in
4
2
◦
−1
;
panel B. Operating conditions: T = 600 C; p = 1 bar; Ar flow 30 ml min
pulse volume 100 µl (4.07 µmol).
that the oxygen used for the initial formation of H2O and CO2
were not regenerated, and only ceria oxygen vacancies were
reoxidized by the CO2 treatment. CO2 dissociation apparently
allows recharging of the oxygen vacancies in the ceria lattice,
but not direct recharging of the consumed interfacial oxygen of
the Ni–Ce boundary. H2/CO ratios were about 2, but increased
slightly compared with those of the first 20 CH4 pulses.
After the full supplementation of oxygen vacancies by CO2
dissociation, the deactivation for syngas production slowed,
but the amount of H2 and CO released decreased significantly
(compare Figs. 6aA and 6aC). This might be due to microstruc-
tural changes during the depletion of reactive oxygen by the
subsequent methane pulses.
Another series of pulse experiments was conducted, start-
ing with H2 pulses. The results of the 15 H2–15 CH4–15 CO2
pulses on 100 mg of Ni10Ce90 are shown in Fig. 7. Although
the first 10 pulses of hydrogen were completely converted, con-
version dropped over the subsequent 5 pulses. Total oxygen
consumption was 5.7 × 10−5 mol. During the next 15 CH4
pulses, the small amount of CO2 initially formed corresponds
to the 0.9 × 10−5 mol of oxygen remaining in the NiO. The
sum amount of oxygen consumption for the formation of H2O
and CO2 shown in Figs. 7A and 7B is 6.6 × 10−5 mol—very
close to the total oxygen content of the NiO in the catalyst,
as summarized in Table 3. In panel B, the product is domi-
nated by H2 and CO formation in amounts comparable to those
formed during the CH4 pulses in Fig. 6a. Considering the to-
tal amount of oxygen consumption for the complete oxidation
of methane in Fig. 6a (6.5 × 10−5 mol), the similar behavior
and amount of production seem reasonable because of its gen-
eration on similar amounts of the active Ni–Ce sites. When
the 15 CO2 pulses were applied to Ni10Ce90, CO2 was con-
verted exclusively again to CO. The sum amount of CO ob-
tained was 2.01 × 10−5 mol. Compared with the sum amount
of CO released by CH4 (2.07 × 10−5 mol), the ceria oxygen
The subsequent 20 CH4 pulses produced essentially H2 and
CO, with only trace amounts of H2O and CO2. The amounts of
H2O and CO2 released in Fig. 6aC are significantly lower than
those shown in Fig. 6aA. Because regeneration of the Ni–O–Ce
sites should have led to total oxidation, this supports the view

110
D.K. Kim et al. / Journal of Catalysis 247 (2007) 101–111
vacancies were completely replenished again by CO2 dissocia-
tion.
dency toward coking, and high catalytic activity) and its slow
deactivation for the CO2 reforming of methane was explored
and understood at a fundamental level by XRD, XPS, pulse
experiments and regeneration tests. Based on the results, a re-
action mechanism was suggested.
3.5. Suggested mechanism
Based on the results of this investigation, the interfacial oxy-
gen in the Ni–Ce boundary of the Ni10Ce90 catalyst seems to
be an active component for the complete oxidation of methane,
because (i) nickel oxide or cerium oxide itself shows no activ-
ity with methane, (ii) the amount of oxygen consumed by the
total oxidation of methane on Ni10Ce90 is close to the total oxy-
gen content of the Ni-oxygen in the catalyst (see Table 3), and
(iii) for Ni20Ce80, the amount of oxygen consumed by the com-
plete oxidation of methane increases by factor of 2.3 compared
with Ni10Ce90. Therefore, the first 1 mol CH4 is converted to
2 mol H2O and 1 mol CO2 at the Ni–Ce boundary accompa-
nying the consumption of the interfacial oxygen (Fig. 6, pulses
1–8). The resulting site is considered the active site for the re-
forming reaction, where subsequently 1 mol CH4 is reformed
through the reaction with the oxygen migrated from ceria lattice
to release 2 mol H2 and 1 mol CO. The reaction is concomi-
tant with the creation of oxygen vacancies of the ceria lattice.
Thereafter, the oxygen vacancies are supplemented by oxygen
arising from the dissociative adsorption of 1 mol CO2 to pro-
duce 1 mol CO. However, the consumed interfacial oxygen in
the Ni–Ce boundary cannot be recovered by CO2 dissociation.
Based on these suggestions, the origin of unusual Ni10Ce90
mixed-oxide properties for the CO2 reforming of methane can
be discussed. First, the catalytic property in the absence of a
prereduction step relies on the reactivity of oxygen contained in
the Ni species. The immediate activity of Ni10Ce90 is based on
the active interfacial oxygen in the Ni–Ce boundary (Ni–O–Ce
site), which leads to the complete oxidation of methane and the
creation of active sites for the subsequent reforming reaction,
whereas the strong interaction of the NiAl2O4 spinel retards
the activity of Ni/Al2O3. Second, the coking resistance relies on
the reactivity of lattice oxygen in the catalyst support. The ceria
lattice oxygen, with high mobility, participates in the reforming
reaction through migration, leading to the formation of CO and
oxygen vacancies of ceria lattice. On the other hand, the lack of
interaction between the surface carbon on metallic nickel and
lattice oxygen of γ -Al2O3 results in the formation of nonre-
movable carbon on the Ni/Al2O3 catalyst. Finally, based on the
negligible amount of surface area loss and coke formation dur-
ing reaction, the continuous deactivation behavior of Ni10Ce90
is due mainly to a loss of active sites. In the pulse experiment
on Ni10Ce90, the continued consumption of reactive oxygen in
the condition of no supplementation or delayed replenishment
resulted in deactivation. Thus, in the CO2 reforming of methane
a balanced rate between the generation and supplement of oxy-
gen vacancies seems to be needed to avoid deactivation.
Several conclusions can be drawn from this investigation:
1. The catalytic property with no prereduction step relies on
the reactivity of oxygen contained in the Ni phase. Active
interfacial oxygen in the Ni–Ce boundary leads to a rapid
startup operation of Ni10Ce90, whereas slow decomposition
of NiAl2O4 spinel results in the long induction period of the
Ni/Al2O3 catalyst.
2. Coking resistance and catalytic activity rely on the reac-
tivity of lattice oxygen in the Ni-catalyst support. The ce-
ria lattice oxygen participates in the reforming reaction
through migration, leading to the formation of oxygen va-
cancies in the ceria lattice. Subsequently, the vacancies are
completely supplemented by oxygen arising from CO2 dis-
sociation.
3. The stability of Ni10Ce90 relies on the balance between
the rate of generation and supplement of oxygen vacancies.
Unbalanced rates of these processes result in deactivation.
Acknowledgments
Financial support was provided by the Deutsche Forschungs-
gemeinschaft (grant GRK 232). The authors thank H. Höltzen
for assisting with product analysis, R. Richter for construct-
ing the reactor, R. Haberkorn for performing XRD studies, and
R. Nagel for helping with the BET measurements.
References
[1] www.fuelcells.umicore.com/de/downloads/fcprotonics032004.pdf.
[2] J.A.C. Dias, J.M. Assaf, J. Power Sources 139 (2005) 176.
[3] D.K. Kim, W.F. Maier, J. Catal. 238 (2006) 142.
[4] J. Scheidtmann, P.-A.W. Weiss, W.F. Maier, Appl. Catal. A 222 (2001) 79.
[5] http://en.wikipedia.org/wiki/Rare_earth_element.
[6] C.T. Campbell, C.H.F. Peden, Science 309 (2005) 713.
[7] A. Trovarelli, Catal. Rev. Sci. Eng. 38 (1996) 439.
[8] J. Kasˆpar, P. Fornasiero, M. Graziani, Catal. Today 50 (1999) 285.
[9] E. Rocchini, A. Trovarelli, J. Llorca, G.W. Graham, W.H. Weber, M. Ma-
ciejewski, A. Baiker, J. Catal. 194 (2000) 461.
[10] F. Esch, S. Fabris, L. Zhou, T. Montini, C. Africh, P. Fornasiero, G. Co-
melli, R. Rosei, Science 309 (2005) 752.
[11] Y. Li, Q. Fu, M. Flytzani-Stephanopoulos, Appl. Catal. B 27 (2000) 179.
[12] P. Bera, K.C. Patil, V. Jayaram, G.N. Subbanna, M.S. Hegde, J. Catal. 196
(2000) 293.
[13] J. Barrault, A. Alouche, V. Paulboncour, L. Hilaire, A. Percheronguegan,
Appl. Catal. 46 (1989) 269.
[14] S. Park, R.J. Gorte, J.M. Vohs, Appl. Catal. A 200 (2000) 55.
[15] X. Wang, R.J. Gorte, Appl. Catal. A 224 (2002) 209.
[16] TOPAS, Version 2.1, Bruker AXS, Karlsruhe.
[17] R.W. Cheary, A.A. Coelho, J. Appl. Crystallogr. 25 (1992) 109.
[18] equiTherm, Version 3.02, Scienceware/VCH, 1997.
[19] S. Wang, G.Q. Lu, Appl. Catal. B 19 (1998) 267.
[20] Z. Xu, Y. Li, J. Zhang, L. Chang, R. Zhou, Z. Duan, Appl. Catal. A 210
(2001) 45.
[21] J.M. Rynkowski, T. Paryjczak, M. Lenik, M. Farbotko, J. Goralski,
J. Chem. Soc. Faraday Trans. 91 (1995) 3481.
[22] S. Klein, S. Thorimbert, W.F. Maier, J. Catal. 163 (1996) 476.
4. Conclusions
In this study, the origin of the exceptional Ni10Ce90 mixed-
oxide properties relative to the Ni/Al2O3 catalyst (i.e., imme-
diate activation in the absence of catalyst prereduction, no ten-

D.K. Kim et al. / Journal of Catalysis 247 (2007) 101–111
111
[23] M. Stöckmann, F. Konietzni, J.U. Notheis, J. Voss, W. Keune, W.F. Maier,
Appl. Catal. A 208 (2001) 343.
[24] D.K. Kim, Ph.D. thesis, Universität des Saarlandes, 2006.
[25] G. Wrobel, M.P. Sohier, A. D’Huysser, J.P. Bonnelle, Appl. Catal. A 101
(1993) 73.
[32] J.H. Bitter, K. Seshan, J.A. Lercher, J. Catal. 176 (1998) 93.
[33] S. Sharma, S. Hilaire, J.M. Vohs, R.J. Gorte, H.-W. Jen, J. Catal. 190
(2000) 199.
[34] M.F. Mark, W.F. Maier, Angew. Chem. Int. Ed. 33 (1994) 1657.
[35] K. Otsuka, Y. Wang, E. Sunada, I. Yamanaka, J. Catal. 175 (1998)
152.
[36] J.A. Rodriguez, J.C. Hanson, A. Frenkel, J.-Y. Kim, M. Perez, J. Am.
Chem. Soc. 124 (2002) 346.
[37] J.Z. Shyu, K. Otto, W.L.H. Watkins, G.W. Graham, R.K. Belitz, H.S.
Gandhi, J. Catal. 114 (1988) 23.
[38] M.S. Brogan, T.J. Dines, J.A. Cairns, J. Chem. Soc. Faraday Trans. 90
(1994) 1461.
[26] N. Yisup, Y. Cao, W.-L. Feng, W.-L. Dai, K.-N. Fan, Catal. Lett. 99 (2005)
207.
[27] J.W. Allen, J. Magn. Magn. Mater. 47–48 (1985) 168.
[28] K. Gschneider, L. Eyring, S. Hüfner, Handbook on the Physics and Chem-
istry of Rare Earths, vol. 10, North-Holland, Amsterdam, 1987.
[29] F.B. Noronha, E.C. Fendley, R.R. Soares, W.E. Alvarez, D.E. Resasco,
Chem. Eng. J. 82 (2001) 21.
[30] D.H. Kim, S.I. Woo, Catal. Lett. 98 (2004) 23.
[31] S.M. Stagg, F.B. Noronha, G. Fendley, D.E. Resasco, J. Catal. 194 (2000)
240.
[39] Q. Fu, H. Saltsburg, M. Flytzani-Stephanopoulos, Science 301 (2003) 935.
[40] C. Lamonier, A. Ponchel, A. D’Huysser, L. Jalowiecki-Duhamel, Catal.
Today 50 (1999) 247.