Metal-promoter interface in Pd/(Ce,Zr)Ox/Al2O3 catalysts: Effect of thermal aging

Article abstract of DOI:10.1016/j.jcat.2003.07.010

Three-way catalysts (TWC) constitute the most satisfactory and efficient solution to diminish pollutant emissions from gasoline-engine-powered vehicles. The behavior of Pd TWC promoted with Ce-Zr mixed oxides and supported on alumina subjected to thermal degradation treatments was studied during light-off tests under stoichiometric CO + NO + O₂, focusing on the region of aging temperature 1273-1373 K, where phase segregation of the ceria-zirconia promoter occurs. Thermal degradation sintered both the noble metal and promoter components of the catalyst and stabilized oxidized Pd entities in contact with the promoter prior to the occurrence of the Ce-Zr phase segregation reaction. The interface between Pd and ceria-zirconia was critical for CO oxidation and less important for NO conversion, although the enrichment in cerium detected during promoter aging after the phase segregation phenomenon could have a positive effect on the NO dissociation process.

Full text of DOI:10.1016/j.jcat.2003.07.010

Journal of Catalysis 221 (2004) 148–161
www.elsevier.com/locate/jcat
Metal–promoter interface in Pd/(Ce,Zr)Ox/Al2O3 catalysts:
effect of thermal aging
A. Iglesias-Juez, A. Martínez-Arias, and M. Fernández-García ∗
Instituto de Catálisis y Petroleoquímica, CSIC, Campus Cantoblanco, 28049 Madrid, Spain
Received 7 May 2003; revised 23 July 2003; accepted 29 July 2003
Abstract
The behavior of Pd three-way catalysts promoted with Ce–Zr mixed oxides and supported on alumina subjected to thermal degradation
treatments has been examined during light-off tests under stoichiometric CO+NO+O . Attention was paid to the region of aging temperature
2
1273–1373 K, where phase segregation of the ceria–zirconia promoter occurs. Characterization of the samples was performed using X-ray
diffraction and Raman spectroscopy. In situ XANES/DRIFTS studies of the Pd chemical state at the surface and in the bulk of the materials
were conducted to determine the noble metal response to the gas atmosphere during light-off. Electron paramagnetic resonance was used to
analyze the promoter surface and to study the redox response of the materials in contact with the reactant gases. Thermal degradation appears
to sinter both the noble metal and promoter components of the catalyst but also stabilizes oxidized Pd entities in contact with the promoter
prior to the occurrence of the Ce–Zr phase segregation reaction. Additionally, it enriches the promoter surface in the lanthanide cation. The
influence of these physicochemical phenomena on the catalytic properties of the Pd-based TWC is discussed.
2003 Elsevier Inc. All rights reserved.
Keywords: Pd TWC; CeO –ZrO ; CO oxidation; NO reduction; XRD; Raman; in situ DRIFTS; in situ XANES; EPR
2
2
1. Introduction
tance to poisons [1,3]. The use of Pd in TWCs is restricted
by the presence of lead in the fuel, which may strongly de-
activate the noble metal by poisoning [4]. On the other hand,
ceria–zirconia (CZ) materials having Ce:Zr atomic ratios
close to 1 and tetragonal or pseudo-cubic structures replace
the classical ceria component due to their better thermal and
oxygen-handing properties [5,6].
During operation, TWCs display a general loss of cat-
alytic activity by the effect of a diversity of phenomena,
which include fouling and mechanical, chemical, and ther-
mal effects [7]. Among them, it is well established that ther-
mal ones are the major problem due to their irreversible
character [1,8,9]. The latter category comprises a series of
phenomena affecting the different components of TWCs. As
is already well known [1–6], a key factor limiting thermal
aging effects is maintenance of catalyst surface area, a fact
directly related to the stabilization of the gamma alumina
particle size and, more importantly, to the inhibition/delay
of phase transition(s) leading to theta or delta polymorphs
and finally to the low-surface-area alpha (alumina) phase.
Among the TWC components, the CZ promoter is known to
help in stabilizing alumina; however, for Ce:Zr atomic ra-
tios close to 1, this component may suffer phase segregation
Regulations concerning automotive emissions are becom-
ing more stringent in the most advanced countries (see, as an
example, Directive 98/69/EC of the European Parliament),
driving the industry to an effort to diminish the environ-
mental impact of pollutants emitted by vehicles. In this con-
text, the development of more efficient and resistant catalytic
converters, able to maintain maximum durability while de-
creasing pollutant emissions to minimum levels, is neces-
sary. Three-way catalysts (TWC) constitute, up to now, the
most satisfactory and efficient solution to diminish pollu-
tant emissions from gasoline-engine-powered vehicles [1,2].
These mainly act through simultaneous oxidation of carbon
monoxide (CO) and hydrocarbons (HC) and reduction of ni-
trogen oxides (mainly NO). In recent years, typical TWC
formulations have included Pd as the active metal, fluorite-
type oxides, such as ceria–zirconia, as promoters, and alu-
mina as support, as well as other minor components, mainly
present in order to enhance thermal stability and/or resis-
*
Corresponding author.
E-mail address: m.fernandez@icp.csic.es (M. Fernández-García).
0021-9517/$ – see front matter
doi:10.1016/j.jcat.2003.07.010
2003 Elsevier Inc. All rights reserved.

A. Iglesias-Juez et al. / Journal of Catalysis 221 (2004) 148–161
149
above 1273 K, leading to a significant decrease in catalyst
surface area [5,6]. Additionally, the CZ catalytic role may
be altered by this phase segregation reaction, as significant
growth of promoter particle size and strong modification of
its surface composition and redox properties have been re-
ported [5,6]. Thermal aging can also influence the number
of noble metal sites as a consequence of Pd sintering and/or
encapsulation [9–11], affecting catalytic activity mainly by
the concomitant loss of metal–promoter interface sites [1–4,
9–11]. All of these phenomena are interrelated but to a large
degree depend on the interface interactions between the no-
ble metal and promoter as well as between the promoter and
the alumina carrier, which are in turn largely responsible for
the stabilization of the corresponding components during ag-
ing of TWCs.
Although thermal aging has been also studied in classi-
cal, ceria-promoted automotive catalysts [12], much less is
known about the physical phenomena occurring in the more
complex case of CZ-promoted TWCs. As above mentioned,
a basic understanding of the interface effects between com-
ponents may help to understand the chemicophysical phe-
nomena associated with the thermal aging and their catalytic
consequences. To this end, we have studied a typical Pd-
based TWC, paying attention to the effects of thermal degra-
dation in the high-temperature region 1273–1373 K, where
experience shows the occurrence of an abrupt degradation
of the CZ component properties [5,6]. A model stoichiomet-
ric CO + NO + O2 mixture will be used to test the effect of
aging in two of the main toxic exhaust emission gases, CO
and NO [1–3]. In an attempt to separate some of the above
mentioned effects, model systems in which Pd has been im-
pregnated over an aged CZ component supported on alumina
have been included in the study. These Pd–TWC systems
have been analyzed using a multitechnique approach involv-
ing XRD, TEM-XEDS, Raman, and EPR for catalyst char-
acterization and catalytic activity in combination with in situ
DRIFTS and XANES techniques to evaluate the catalytic
performance in the CO + NO + O2 reaction.
of the CZ and CZA materials gave Zr/Ce atomic ratios of
1.0 0.1.
A, CZ, CZA, CZA-m1273, and CZA-m1373 supports
were impregnated by the incipient wetness method with
aqueous solutions of Pd(NO3)2 to give a 1.0 wt% metal load-
ing. The resulting catalysts were calcined following the same
drying/calcination procedure described above for the sup-
ports (773 K, 2 h) and are referred to as fresh PdA, PdCZ,
and PdCZA samples or PdCZA-m1273 and PdCZA-m1373
model systems. Oven aging of the fresh PdA, PdCZ, and
PdCZA materials was performed by successive treatments
at 1273 K (samples PdA-a1273, PdCZ-a1273, and PdCZA-
a1273) and 1373 K (PdA-a1373, PdCZ-a1373, and PdCZA-
a1373) under air for a total of 12 h for each step. Therefore,
the fresh sample for each support (PdA, PdCZ, PdCZA),
as well as model PdCZA-m and aged PdA-a, PdCZ-a, and
PdCZA-a series, have been prepared. Note that only the sup-
port underwent high-temperature (1273–1373 K) aging in
materials denoted with an -m suffix, while both the noble
metal and support are jointly aged for materials denoted with
an -a suffix.
2.2. Catalytic tests
Catalytic tests using a 1% CO + 0.1% NO + 0.45% O2
(N2 balance) stoichiometric mixture at 30,000 h−1 were
performed in a pyrex glass reactor system. After grinding
and sieving, catalyst particles in the range 0.12 to 0.25 nm,
for which internal diffusion effects were minimized, were
employed for these tests. Gases were regulated with mass
flow controllers and analyzed on-line using a Perkin–Elmer
1725X Fourier transform infrared (FTIR) spectrometer cou-
pled with a multiple reflection transmission cell (Infrared
Analysis Inc.; path length 2.4 m). Oxygen concentrations
were determined using a paramagnetic analyzer (Servomex
540A). Prior to catalytic testing, in situ calcination was per-
formed in diluted oxygen (2.5% O2 in N2) at 773 K, fol-
lowed by cooling under the same atmosphere and subsequent
N2 purge at room temperature (RT). A standard test con-
sisted of increasing the temperature from 298 to 823 K at
2. Experimental
5 K min−1
.
2.1. Catalyst preparation
2.3. Characterization techniques
A ceria–zirconia/alumina support (CZA) of 10 wt% of
mixed oxide, expressed as Ce0.5Zr0.5O2, was prepared us-
ing γ -Al2O3 Condea Puralox as carrier and an adaptation
of the microemulsion method used previously for prepar-
ing the unsupported oxide, hereafter called CZ [13]. Full
details of the procedure can be found elsewhere [14]. Af-
ter drying overnight at 393 K, portions of these alumina,
ceria–zirconia, and ceria–zirconia/alumina supports (here-
after called A, CZ, and CZA, respectively) were calcined at
773 K for 2 h under dry air flow or aged under air by a treat-
ment at 1273 (CZA-m1273) or 1373 (CZA-m1373) K for a
total of 12 h for each step. Chemical analysis (by ICP-AES)
Powder XRD patterns were recorded on a Seifert diffrac-
tometer using nickel-filtered Cu-Kα radiation, operating at
40 kV and 40 mA, and with a 0.02◦ step size, using a con-
tinuing time of 5 s. Fitting of XRD data was performed
using Pearson VII functions and the Winfit! Program (by
S. Kumm; http://www.geol.unerlangen.de). BET surface ar-
eas were determined from N2 adsorption isotherms in a Mi-
cromeritics 2100 automatic apparatus.
Raman spectra were acquired with a Renishaw Micro-
Raman System 1000 equipped with a cooled OCD detector
(200 K) and a holographic super-Notch filter that removes
the elastic scattering. The samples were excited with the

150
A. Iglesias-Juez et al. / Journal of Catalysis 221 (2004) 148–161
514 nm Ar line. The spectra acquisition consisted of 60 ac-
cumulations of 10 s acquisition time for each sample, with a
typical running power of 25 mW.
tion (varimax rotation) which should align factors (as close
as possible) along the unknown concentration profiles, fol-
lowed by iterative transformation factor analysis (ITFA).
ITFA starts with delta function representations of the con-
centration profiles, associated with each chemical species,
located at temperatures predicted by the varimax rotation,
which are then subjected to refinement by iteration until er-
ror in the resulting concentration profiles is lower than the
statistical error extracted from the set of raw spectra [15–17].
Electron paramagnetic resonance (EPR) spectra were
recorded at 77 K with a Bruker ER 200 D spectrometer
operating in the X-band and calibrated with a DPPH stan-
dard (g = 2.0036). Portions of the samples were placed
inside a quartz probe cell with greaseless stopcocks, using
a conventional dynamic high-vacuum line for the different
treatments. Oxygen adsorption treatments were initiated by
admitting a dose of 210 µmol of O2 per gram of sample at
77 K. This treatment was followed by thorough outgassing at
77 K (aiming to minimize magnetic interactions with non-
chemisorbed oxygen) and referred to as O2 adsorption at
77 K. Subsequent 30 min warming to room temperature
(RT) will be denoted as O2 adsorption at RT. Some of the
experiments were continued by subsequent outgassing at
RT. These experiments were designed on the basis of for-
mer work, in order to get maximum intensity and optimum
resolution of superoxide signals formed upon oxygen inter-
action [6,19,20]. Pretreatments under CO or CO + NO were
performed by admitting at RT, respectively, 100 Torr of CO
or, in a consecutive way, 100 Torr of CO and 10 Torr of
NO, followed by heating at 453 K for 1 h and outgassing
at 453 K for 0.5 h. Computer simulations were employed in
some cases to determine spectral parameters or contributions
of individual signals.
DRIFTS analysis of adsorbed species present on the cata-
lyst surface under reaction conditions was carried out using a
Perkin–Elmer 1750 FTIR fitted with a MCT detector. Analy-
sis of the NO conversion at the outlet of the IR cell was
performed by chemiluminescence (Thermo Environmental
Instruments 42C). The DRIFTS cell (Harrick) was fitted
with CaF2 windows and a heating cartridge that allowed
samples to be heated to 773 K. Portions of ca. 65 mg were
calcined in situ with synthetic air (20% O2 in N2) and then
cooled to 298 K in the same atmosphere before introduction
of the reaction mixture and subsequent heating at 5 K min−1
up 573 K, recording one spectrum (4 cm−1 resolution, av-
erage of 50 scans) generally every 15 K. The gas mixture
(1% CO + 0.1% NO + 0.45% O2, N2 balance) was prepared
using a computer-controlled gas blender with 60 cm3 min−1
passing through the catalyst bed. Following runs under re-
action conditions, the samples were cooled under inert gas
flow. Subsequently, CO (3% in N2) was flowed at RT and
finally CO(g) was flushed out with N2.
XANES experiments at the Pd K-edge were performed
on line 9.2 at SRS, Daresbury. A Si(220) double-crystal
monochromator was used in conjunction with a rejection
mirror to minimize the harmonic content of the beam. Addi-
tionally, a detuning of 30% was used. Transmission experi-
ments were carried out using Kr/Ar-filled ionization cham-
bers. The energy scale was simultaneously calibrated by
measuring the Pd foil using a third ionization chamber. Sam-
ples as self-supported discs (absorbance ca. 0.5) were placed
in a controlled-atmosphere cell for treatment. XANES spec-
tra were taken every 15 K during a 5 K min−1 temperature
ramp up to 723 K, in the presence of the CO + NO + O2
flowing mixture (similar to that employed for catalytic ac-
tivity tests).
3. Results
Results obtained in XANES experiments were subjected
to normalization using standard procedures and analyzed
using principal component analysis (PCA) [15–17], which
assumes that the absorbance in a set of spectra can be mathe-
matically modeled as a linear sum of individual components,
called factors, which correspond to each one of the palla-
dium species present in a sample, plus noise [18]. To deter-
mine the number of individual components, an F-test based
on the variance associated with factor k and the summed
variance associated with the pool of noise factors is per-
formed. A factor is accepted as a “pure” chemical species
(i.e., a factor associated with signal and not noise) when
the percentage of significance level of the F-test, %SL, is
lower than a test level set in previous studies at 5% [15,16].
The ratio between the reduced eigenvalues, R(r), which ap-
proaches one for noise factors, was also used in determin-
ing the number of factors. Once the number of individual
components was set, XANES spectra corresponding to in-
dividual palladium species and their concentration versus
temperature profiles were generated by an orthogonal rota-
3.1. Catalytic tests
Results obtained during the light-off tests for the Pd-
CZA, PdCZA-m1273, PdCZA-a1273, PdCZA-m1373, and
PdCZA-a1373 systems are depicted in Fig. 1. Comparison
with the reference PdA and PdCZ systems is presented in
Table 1 on the basis of conversion values at a 50% level (T50)
for fresh and 1273-K-aged materials.
Aging by thermal effects seems to have a significant im-
pact on the CO and NO conversions displayed by the PdCZA
system (Figs. 1A and B). Thus, an increase in the T50 tem-
perature of 60/25 and 25/35 K is detected in the CO/NO
elimination reactions in comparing, respectively, PdCZA
and PdCZA-a1273 samples and the two aged PdCZA-a cat-
alysts (Fig. 1; Table 1). The physicochemical effects taking
place during aging of the CZA support seem of lesser impor-
tance in defining catalytic performance; this follows from the
fact that both model systems, PdCZA-m1273 and PdCZA-
m1373, show close or improved performance in both CO

A. Iglesias-Juez et al. / Journal of Catalysis 221 (2004) 148–161
151
Fig. 1. CO (A) and NO (B) conversion profiles, NO production (C), and N selectivity (D) during CO + NO + O light-off tests.
2
2
2
Table 1
The strongest deactivation is clearly produced for the PdCZ
system. Comparison between PdCZA and PdA shows mod-
erate differences in CO oxidation activity. These are much
stronger for NO reduction activity, in particular due to the
relatively strong increase in the T50 value displayed by PdA
after aging.
50% conversion temperatures (K) observed for fresh and 1273-K-aged Pd-
CZA, PdCZ, and PdA systems
Treatment
PdCZA
PdCZ
PdA
CO
NO
CO
NO
CO
NO
Fresh
Aged: 1273 K
Difference
435
495
60
475
500
25
360
550
190
465
565
100
470
515
45
420
515
95
3.2. Characterization of the calcined samples
Powder X-ray diffraction patterns are displayed in Fig. 2.
Reflections are here indexed for Ce-rich phases (pseudo-
cubic tꢀꢀ or cubic c phases; Ce:Zr ꢀ 1) as derived from the
cubic Fm3m cation lattice symmetry, while Zr-rich ones
(tetragonal tꢀ, t) have been indexed using a P42/nmc space
group symmetry. It can be noted that the poor resolution of
the XRD data, inherent in the heterogeneity of the CZA-
related materials, does not allow a simple resolution of
the tetragonal splitting of the corresponding Ce–Zr mixed
oxides, as discussed previously for the CZA support [6].
Analyses of the cell parameters and particle size obtained
from the XRD data as well as the BET area of the sup-
ports/catalysts are summarized in Table 2. The XRD data
indicate the presence of several transitional alumina phases
as a function of the aging treatment suffered by the CZA-
and NO elimination reactions with respect to the fresh Pd-
CZA catalyst. The samples do not display significant dif-
ferences in the NO reduction selectivity during the light-off
tests, reaching 100% N2 yield a few degrees after the tem-
perature of full NO conversion (Fig. 1D). Small differences
are, however, encountered in the production of NO2, which
is typically negligible for Pd-based systems operating under
stoichiometric conditions [21] but here visible for PdCZA-
a1373 and, more clearly, for PdCZ-a1273. NO2 formation
thus appears to be related to systems containing CZ particles
of significant dimensionality (vide infra).
The influence of the support nature on the CO and NO
elimination catalytic activity is briefly illustrated in Table 1.
The third row for each type of catalyst reports the difference
in T50 values between the fresh and 1273-K-aged catalysts.

152
A. Iglesias-Juez et al. / Journal of Catalysis 221 (2004) 148–161
Fig. 2. XRD diffractograms of support and Pd-containing samples. Peaks corresponding to ceria–zirconia mixed phases are indexed; C or Z prefixes refer to
Ce-rich or Zr-rich phases, respectively. Peak positions corresponding to PdO are indicated by vertical dashed lines. Greek letters indicate presence of different
∗
alumina phases ( symbol corresponds to diffraction peaks which can be associated with several transitional phases). See text for details.
Table 2
the corresponding mixed oxide compound. The XRD study
(Fig. 2 and Table 2) indicates the absence of significant dif-
ferences between the support and Pd-containing catalysts in
either the fresh or the aged states. Up to 1273 K, a growth
of CZ particle size and a small variation of the cell para-
meter are detected (Table 2); this suggests that the integrity
of the CZ component is mostly maintained, although the
asymmetry of the peaks indicates a certain degree of het-
erogeneity (see, as an example, the zone corresponding to
the c(111) reflection of Fig. 2; note also that higher index
peaks with better resolved splittings do not yield a definitive
conclusion in this respect, due to their lower intensity and/or
overlapping with peaks from other compounds). Upon ag-
ing at 1373 K, the Ce–Zr phase compositional homogeneity
abruptly breaks, giving rise to two discernible phases (Fig. 2
and Table 2), one with Ce/Zr atomic ratio close to 3.3, ac-
cording to the value of its lattice parameter [5,6,22], and
the other exhibiting Zr enrichment and composition, which
cannot be determined on the exclusive basis of XRD due
to the rough steadiness of cell parameters for Ce/Zr com-
positions below 1 [5,22]. In the case of the PdCZ system,
the phase segregation reaction is already visible after the
treatment at 1273 K. Indeed, Table 2 indicates a degree of
Ce–Zr phase segregation for PdCZ-a1273 similar to that for
PdCZA-a1373, in spite of the fact that particle size appears
to evolve in a roughly similar way with temperature for both
CZ- and CZA-supported systems. A mutual stabilization be-
tween the promoter and the alumina carrier then becomes
apparent. The tetragonal character of the Zr-rich t phase is
Main characterization results for calcined materials
2
−1
) Cell parameter (Å) Particle size (nm)
a
Sample
S
(m g
BET
A
200
84
7.5
–
–
–
–
–
–
A-a1273
A-a1373
b
b
CZ
130
5
2
5.30
5.36 , 3.63/5.26
4.5
15, 7
27, 24
c
c
Pd/ZC-a1273
Pd/ZC-a1373
b
5.37 , 3.63/5.23
b
CZA
PdCZA
CZA-a1273
PdCZA-a1273
CZA-a1373
PdCZA-a1373
203
201
83
77
37
5.30
3
3
12
10
28, 19
26, 15
b
5.30
b
5.33
b
5.34
b
c
c
5.37 , 3.62/5.29
b
38
5.37 , 3.61/5.32
a
ꢀꢀ
When adequate, first average particle size is reported for t /c phases
ꢀ
while the second correspond to t /t tetragonal phases.
b
Fluorite-type cell with Fm3m symmetry is considered for phases as-
ꢀꢀ
cribed to cubic c or tetragonal t phases assigned according to Raman
results.
c
ꢀ
Tetragonal-type with P 4 /nmc symmetry is considered for t /t
2
phases.
containing materials. Concerning these phases, the practical
absence of differences between the behavior under aging of
the support (for which data obtained under the same con-
ditions can be found elsewhere [6]) and the Pd-containing
catalyst (Fig. 2) can be noted.
Peaks related to the CZ component (discussed elsewhere
in Refs. [5,6]) are dominated by the cation sublattice of

A. Iglesias-Juez et al. / Journal of Catalysis 221 (2004) 148–161
153
ther to a localized defect or an impurity-derived vibration
mode. The ca. 470 cm−1 band is first order anion-lattice
associated vibration [23], whose characteristic frequency
displays low sensitivity to changes in Ce/Zr composition
near the unity [6]. A point to stress is the apparent absence
of well-resolved bands from the Zr-rich phase formed upon
phase segregation as a consequence of the aging treatment
at T > 1273 K. In this respect, for tꢀ/t phases, apart from
other vibration modes, a Zr–O stretch mode should be eas-
ily detected at around 235 cm−1 for a Zr-rich (Ce/Zr ca.
0.33) phase or below ca. 260 cm−1, the corresponding value
of the tetragonal ZrO2 phase [25]. The absence of such fea-
tures suggests a significant structural disorder in the anion
sublattice of the Zr-rich phase, in accordance with EXAFS
results reported for such phases [26].
The model systems obtained from the aging treatment of
the CZA support (PdCZA-m1273, PdCZA-m1373; Fig. 3)
do not show a significantly perturbed Raman shape with re-
spect to the support alone. A difference is the presence of a
band at ca. 650 cm−1 ascribable to the B2g Raman mode of
PdO [27], which, on the other hand, is absent in the fresh
PdCZA catalyst [28], but discernible in PdCZ aged speci-
mens (Fig. 3). A general decrease in Raman intensity can
be also noticed in the model systems and could be mostly
associated to the colour change in samples. The presence
of a better crystallized PdO phase is detected in the aged
PdCZA-a1273 and PdCZA-a1373 samples; the numerous
peaks associated with this well-developed noble metal ox-
idic phase are readily ascribable by comparison with the
spectra of the PdA-a1273 system (Fig. 3). It must be noted
that possible attribution of these bands to some of the tran-
sitional alumina phases observed by XRD (Fig. 2) is dis-
carded on the basis of comparison with the spectrum of
PdCZ-a1273 (Fig. 3). On the other hand, the presence of
the noble metal in the PdCZA-a series of catalysts affects
to some extent the CZ component. A loss of Raman peak
intensity of CZ-associated bands is detected in the PdCZA-
a1273 spectrum (see, for instance, the peak at ca. 475 cm−1
in Fig. 3), which suggests that a certain loss of anion-lattice
ordering is produced in that phase upon aging at 1273 K.
The intensity of such features (particularly evident for the
ca. 475 cm−1 band) is, however, significantly recovered af-
ter the phase segregation induced by the aging treatment at
1373 K (Fig. 3). Preliminary studies in a similar series of
ceria/alumina supported specimens indicate the exclusivity
of such interaction to the ceria–zirconia-containing alumina-
supported materials, although similar effects may be present
in Pd/Ceria systems, as formation of a Pd–Ce solid solution
has been reported [29].
Fig. 3. Raman spectra of support and Pd-containing materials. Peaks re-
lated to Ce-containing components are marked with a cross; other peaks are
associated with the PdO phase.
better developed in the PdCZ-a1373 sample, according to
the presence of all characteristic reflections of its P42/nmc
symmetry (marked in Fig. 2).
Peaks attributable to Pd-containing phases are only de-
tected for aged PdCZA samples, independent of the thermal
treatment (Fig. 2). PdO peaks are shown as vertical dashed
lines in Fig. 2 and their absence is noted in the PdCZ series,
indicating some limitation in the growth of the noble metal
with respect to the PdCZA series. In fact, TEM/EDS analysis
(not shown) gives evidence that large Pd particle sizes are as-
sociated with sample regions depleted in ceria–zirconia and
are thus characteristic of PdA-like regions.
In contrast to the cation-dominated information yielded
by XRD, Raman spectra of fluorite-related phases are dom-
inated by lattice oxygen vibrations [23] and are obviously
sensitive to crystalline symmetry. Raman spectra of the
CZA support after the different treatments (Fig. 3) give evi-
dence of the presence of a pseudo-cubic, tetragonal tꢀꢀ Ce–Zr
phase, which accounts for the ca. 470 and 618 cm−1 phonon
modes [6,14,22], irrespective of the treatment temperature.
An additional peak at 251 cm−1 has been also detected in
supported ceria specimens [24], and could be ascribed ei-
3.3. In situ DRIFTS
DRIFTS spectra obtained under reaction conditions for
the fresh PdCZA (Fig. 4A), model PdCZA-m1273 (Fig. 4B),
and PdCZA-m1373 (Fig. 4C) systems and the aged PdCZA-
a1373 catalyst (Fig. 4D) show a general evolution of COg (as

154
A. Iglesias-Juez et al. / Journal of Catalysis 221 (2004) 148–161
Fig. 4. DRIFTS spectra of (A) PdCZA, (B) PdCZA-m1273, (C) PdCZA-m1373, and (D) PdCZA-a1373 samples in the presence of nitrogen (a) and a flow of
1% CO, 0.45% O , 0.1% NO, N balance at (b) 303, (c) 318, (d) 333, (e) 348, (f) 363, (g) 378, (h) 393, (i) 408, (j) 423, (k) 438, (l) 453, (m) 468, (n) 483,
2
2
(o) 498, (p) 513, (q) 528, (r) 543, (s) 558, and (t) 573 K.
a barely discernible doublet in the region 2050–2250 cm−1
)
Analysis of the PdCZA sample has been provided else-
where [21]. In brief, the presence of carbonyls adsorbed
on Pd(II) sites (2140–2160 cm−1), on Pd(I) (ca. 2110–
2120 cm−1), and atop (2096–2050 cm−1) and bridge (1975–
1945 cm−1) Pd(0) sites is observed at low temperatures
and CO2,g, which roughly correlates with catalytic activity
results (Fig. 1). NOg is typically not detected in the em-
ployed conditions due to its lower absorption coefficient and
concentration in the feed [21].

A. Iglesias-Juez et al. / Journal of Catalysis 221 (2004) 148–161
155
[21,28,30–32]. Carbonyls adsorbed onto threefold Pd(0)
sites are also visible as a red shoulder of the bridge carbonyl
contribution extending to ca. 1880 cm−1. After contact with
the reactive mixture at room temperature, differences be-
tween the fresh PdCZA sample and PdCZA-m systems
mainly concern the easier detection of Pd(0) surface species
for the fresh catalyst and the larger intensity, among car-
bonyls characteristic of oxidized Pd species, of those ad-
sorbed onto Pd(I) sites for PdCZA-m catalysts. On the other
hand, the PdCZA-a1373 sample (Fig. 4D) does not show
features ascribable to adsorbed species, in agreement with
previous studies for the sample aged at 1273 K [32]. This
is consistent with a significant sintering of the noble metal
phase.
When the temperature is raised, transformation of Pd(II)
and Pd(I) surface sites to Pd(0) is observed; except for the
aged PdCZA-a1273 catalyst, carbonyls adsorbed onto ox-
idized Pd species are detected up to about 423 K. When
increasing the temperature of reaction, the carbonyls ad-
sorbed on atop and bridge zerovalent surface sites display
the characteristic red shift induced by changes in adsorbed-
CO dipole coupling by the effect of decreasing CO cov-
erage as the CO oxidation reaction progresses [28,30–32].
Differences between the fresh and model PdCZA systems
(PdCZA-m series) are also detected during the light-off ramp
and concern the behavior of Pd(0) centers with temperature.
Considering carbonyls adsorbed on Pd(0) atop sites, a con-
tinuous decrease in frequency is detected for fresh PdCZA
while the model catalysts appear to have two contributions,
one at 2096–2094 cm−1, invariant up to 423 K, and an-
other at lower wavenumber, which apparently begins to shift
downwards at lower temperature. Concerning bridge Pd(0)
sites, it can be noted that these are populated by CO from
room temperature (RT) and display invariant frequency up to
ca. 423 K in the PdCZA case; contrarily, practical absence
of bridge carbonyls intensity is evident in both model sys-
tems at low temperature, beginning to appear at T above ca.
400 K.
species, present at temperatures up to ca. 600 K (Figs. 4B,
4C), favors the second assignment [35,37]. CN− group could
be obtained from NCO decomposition on the metal sur-
face [35,36] or, less likely, may indicate C–O bond disso-
ciation, either directly or through the Boudouart reaction
(2CO → C + CO2). Interesting here is the existence of a
visible correlation between the onset of NCO detection and
the intensity changes with temperature above described for
bridge-bonded Pd(0) carbonyls; however, details of the cor-
relation vary with sample and will be further analyzed in
Section 4.
To examine the state of the metal surfaces on com-
pletion of the reaction runs and in the absence of ther-
mally driven adsorption effects, the reactants were flushed
at 623 K and the sample was cooled under N2 to room
temperature and subsequently exposed to 3% CO/N2 be-
fore the cell was flushed again with N2. Spectra recorded
during this treatment are displayed in Fig. 5. All spectra, ex-
cept that corresponding to PdCZ-a1273, show contributions
from the already assigned bands of on-top (2105, 2095–
2085, 2044 cm−1) and bridge-bonded (1992–1979 cm−1
)
carbonyls on Pd(0) sites; contributions from CO molecules
adsorbed on Pd(II) centers (2147–2144 cm−1) can be also
observed in the fresh PdCZA and PdCZA-m1273 systems
(Fig. 5). A first point to note is that the atop to bridge inten-
sity ratio differs from the fresh to model PdCZA systems; the
larger ratio of the PdCZA sample clearly indicates a more
dispersed Pd phase and, consequently, a lower average par-
ticle size [21,28,30]. On the other hand, the aging procedure
induces a general, strong decrease of the carbonyl-related in-
tensity, consistent with previous reports [32]; however, some
differences are clearly visible among our samples. The in-
fluence of the support nature on the aging effects exerted on
the noble phase is described after the treatment at 1273 K.
The PdCZA-a1273 spectrum (Fig. 5) shows a strong resem-
blance to the spectrum corresponding to PdA-a1273, while
adsorbed species are absent in the spectrum of PdCZ-a1273
(Fig. 5). A more subtle effect of the aging treatment in the
In the fresh PdCZA sample, two bands at ca. 2240 and
2211 cm−1 are detected at temperatures close to 400 K.
The first is assigned to isocyanate (NCO) species adsorbed
onto aluminum cations [21,32,33], while the second may
be described as NCO bending species also adsorbed onto
the support. An alternative assignment to NO2+-related
species [34] is possible, although such adspecies are unlikely
to form under stoichiometric conditions. NCO species are
formed by N–O bond dissociation and reaction with CO at
metal surfaces followed by spillover to the support. The large
intensity of NCO when adsorbed on alumina makes it a use-
ful (though approximate) fingerprint for monitoring the NO
dissociation process. In the model systems (Figs. 4B, 4C),
a band at 2168 cm−1 is detected instead of at 2211 cm−1
.
The band intensity runs parallel to the NCO one. This con-
tribution can be assigned to ionic (NCO)− species absorbed
on metallic surfaces [35,36] or CN− species probably ad-
sorbed onto alumina [37]. The thermal stability of such ad-
Fig. 5. DRIFTS spectra following experiments shown in Fig. 4 after expo-
sure to CO and subsequent flushing with N .
2

156
A. Iglesias-Juez et al. / Journal of Catalysis 221 (2004) 148–161
PdCZA system is the presence of two atop species (2105,
2044 cm−1), which may indicate a more heterogeneous dis-
tribution of noble metal particle size.
Table 3
Principal component analysis of Pd K-edge XANES results
Factor
Eigenvalue
%SL
R(r)
Variance
A. PdCZA
3.4. In situ XANES
1
2
3
4
5
6
7
8
458.49
0.00
0.00
1334.23
194.34
1.45
99.930
0.070
0.001
0.001
0.31594
0.00149
0.00092
0.00040
0.00029
0.00021
0.00018
10.24
12.38
25.48
29.50
34.36
35.02
Factor analysis results at the Pd K-edge (Table 3) for Pd-
CZA, PdCZA-m1273, PdCZA-m1373, and PdCZA-a1273
obtained during a light-off run under CO+NO+O2 are con-
sistent with the existence of two different chemical species
in all cases. The number of species is clearly indicated by
sharp variation in the %SL and R(r) values displayed in Ta-
ble 3. In Fig. 6, the fraction of these chemical species as a
function of reaction temperature is depicted. This plot shows
that Pd is initially present as a single chemical species which
evolves into a single final state. Small differences in the on-
set of the initial species disappearance for fresh and model
systems can be noted. As can be expected, the aged PdCZA-
a1273 evolves much later. Differences between the fresh and
model PdCZA catalysts are somewhat maximized at inter-
mediate stages of the reaction run; the temperature of max-
imum rate of initial-state (chemical species) disappearance
is ca. 510, 550, 570, and 640 K, respectively for PdCZA,
PdCZA-m1273, PdCZA-m1373, and PdCZA-a1273. For the
fresh and model systems, only one species is detected above
ca. 625 K, while this is true above 700 K in the PdCZA-
a1273 case. It may thus become apparent that, according to
XANES, differences between the fresh and model PdCZA
catalysts are absent at the onset and final reduction tempera-
tures, while the aged specimens reduction starts about 75 K
later.
When compared with the reference spectrum of PdO
(Fig. 7A), the shape of the XANES spectrum of the sam-
ples in their oxidized state shows only slight differences,
which may be a consequence of the limited particle size
of the oxidic Pd-containing entities and/or some interaction
with the support. This suggests a similar oxidation state,
Pd(II), local geometry (D2h), and oxygen first-shell coor-
dination distance for all the samples displayed in Fig. 7A.
Similarly, the XANES spectra of the final species detected
in the reaction (Fig. 7B) do not show significant differences
between the fresh and model samples and can be readily
identified in all cases as Pd particles in a zerovalent oxida-
tion state. A higher intensity of the 5sp and 4f continuum
resonances (CRs) and a different energy position of the for-
mer CR are, however, detected for the aged PdCZA-a1273
specimen. As these CR energy positions have a geometrical
contribution with an inverse dependence on the coordination
distance [17], it appears that the XANES technique is able to
evidence differences in particle size only between the aged
(PdCZA-a) series and the remaining systems. Differences
between the fresh and model PdCZA particle size are thus
not visible in the XANES spectra. On the other hand, the
similar XANES shapes detected for oxidized and reduced
Pd-containing species of fresh and model PdCZA systems
indicates a reasonable similarity in electronic properties.
2.06
1.23
1.22
1.01
1.10
B. PdCZA-m1273
1
2
3
4
5
6
7
8
677.54
1.2837
0.00
0.00
7.19
490.2
361.4
1.21
3.42
1.19
1.46
0.95
0.97
99.810
0.189
0.001
0.03283
0.02503
0.00067
0.00051
0.00032
0.00030
4.57
23.75
26.59
35.77
34.93
C. PdCZ-m1373
1
2
3
4
5
6
7
8
459.32
0.00
0.00
5.99
475.32
267.41
1.93
2.89
1.18
1.14
1.03
0.99
99.808
0.191
0.001
0.87895
0.00298
0.00137
0.00042
0.00031
0.00023
0.00019
9.85
29.90
33.75
37.81
40.26
D. PdCZA-a1273
1
2
3
4
5
6
7
8
383.46
0.00
0.00
8.79
22.53
28.37
40.02
40.13
46.49
494.32
162.43
2.61
1.43
1.57
0.91
1.07
1.06
99.819
0.179
0.001
0.68891
0.00372
0.00123
0.00072
0.00038
0.00033
0.00023
−3
Variances lower than 10 are not reported.
Fig. 6. Concentration profiles during the temperature-programmed reaction
run for Pd-containing species. See text for details.

A. Iglesias-Juez et al. / Journal of Catalysis 221 (2004) 148–161
157
Fig. 8. EPR spectra following oxygen adsorption at RT (and subsequent RT
outgassing) on the samples outgassed at 773 K. (a) CZA; (b) CZA-m1273;
(c) CZA-m1373. Computer simulations are overlapped as dotted lines on
the experimental spectra.
mentally achieved by performing a treatment consisting of
outgassing of the sample at 773 K followed by O2 adsorption
at RT and subsequent RT outgassing [6,39]. Fig. 8 shows
the results obtained after this treatment for the fresh and
aged supports. The spectra are basically constituted by the
overlapping of two kinds of signals: first, an orthorhombic
signal OZ at gz = 2.033–2.032, gy = 2.009, and gx = 2.002,
which corresponds to superoxide species chemisorbed on
zirconium cations (O2−–Zr4+ species) [6,19,40]. Second,
quasiaxial signals type OC with g = 2.040–2.031 and
g = 2.011–2.010, which correspond to superoxide species
⊥
Fig. 7. XANES spectra for PCA predicted chemical species. (A) Pd initial,
oxidized species; (B) Pd final, reduced species. Reference compound, PdO
(4A), as dotted line.
chemisorbed on cerium cations (O2−–Ce4+ species) [6,19,
20,38,39]. The main differences between the samples con-
cern changes in the relative contributions of type OC and
OZ signals. Thus, the results in Fig. 8 clearly show relative
contributions of OC signals increasing with the aging tem-
perature; according to computer simulations, the OZ/OC
ratio of integrated intensities is 1.9, 0.9, and 0.6 for CZA,
CZA-m1273, and CZA-m1373, respectively. This indicates
that cerium surface enrichment is gradually produced with
aging in the 3D-CZ particles, in line with previous studies of
the aging process of unsupported Ce–Zr mixed oxide sam-
ples [41,42]. Another subtle difference concerns the appear-
ance of a type OC signal feature at g = 2.040 for the aged
samples (Figs. 8b and c). According to a previous work [39],
this is characteristic of the formation of relatively large 3D-
CZ particles, which interact less strongly with the underlying
alumina, as a consequence of the sintering process.
3.5. EPR analysis
As shown elsewhere [6,19,20,38,39], O2-probe EPR is
a powerful technique for obtaining insight into surface
structural characteristics and redox properties of cerium-
containing samples. A previous report has shown in detail
the characteristics of the superoxide signals obtained upon
O2 adsorption at 77 K on the CZA support outgassed at
773 K in either the fresh (calcined at 773 K) or the aged
(at 1273 or 1373 K) states [6]. From these results, presence
of dispersed CZ entities (in the form of two-dimensional
patches or 2D-CZ) on the surface of the alumina was re-
vealed. No strong change in the characteristics of these 2D-
CZ entities was apparently produced as a consequence of the
aging treatment [6]. Significant changes are, however, pro-
duced on the surface characteristics of bulk-like, aggregated
three-dimensional CZ particles (3D-CZ entities). Insight
into these surface and redox characteristics can be experi-
Redox properties of some of the catalysts (the PdCZA-
m and -a series treated at 1273 K) have been analyzed by
monitoring the evolution of EPR signals formed after oxy-
gen adsorption on samples pretreated under either CO or
CO + NO at 453 K. Different O2−–Ce4+ species, attribut-

158
A. Iglesias-Juez et al. / Journal of Catalysis 221 (2004) 148–161
ble metal, the promoter, and the carrier. In this section we
discuss the main physicochemical effects exerted by ther-
mal degradation. As already mentioned in the Introduction,
previous reports have analyzed these in terms of the no-
ble metal particle sintering and/or its covering (by support
components) as well as the promoter sintering and phase
segregation reaction [8–12], but much less is known about
the specific interaction between the noble metal and the pro-
moter and the corresponding influence upon the catalytic
properties of the systems.
Concerning the metal component, our study displays re-
sults similar to those of previous reports [9,12]; XRD only
detects a PdO phase in the case of the aged PdCZA samples
(PdCZA-a series), while XRD and Raman give evidence of
the larger crystallinity of the oxidic Pd phase in the PdCZA
system with respect to similarly aged PdCZ samples (Figs. 2
and 3). Additionally, the absence (or significant decrease)
of noble metal presence at the surface of the PdCZ material
is evidenced by the DRIFTS study of CO adsorption after
reaction (Fig. 5). The comparative study of CZ- and CZA-
supported catalysts shows therefore that Pd is significantly
covered in the aged PdCZ samples, limiting the growth of
the noble metal (or metal oxide) particle size [11]. On the
other hand, data presented in Section 3.2 give evidence that
Pd sintering in the PdCZA system under thermal degrada-
tion is more acute in zones depleted in promoter presence
and thus may be primarily associated with PdA-like zones.
The thermal aging treatment also sinters the promoter and
the carrier, although previous results on the CZA-support
clearly evidence the beneficial mutual influence between
these two components [6,14]. Comparative study of the
CZA support and PdCZA system subjected to thermal ag-
ing shows the absence of significant differences in terms of
particle size of the CZ component as well as main com-
positional parameters (Ce, Zr-content of the final phases)
before and after the promoter phase segregation (Fig. 2; Ta-
ble 2). Additionally, Pd is not expected to exert an effect in
the surface Ce-enrichment detected in the promoter patches
(Fig. 8). It could be thus concluded that the noble metal
does not influence the evolution of bulk and surface prop-
erties of the promoter upon aging. A clear effect is, however,
visible in the Raman spectra displayed in Fig. 3; at tempera-
tures below occurrence of the phase segregation reaction, the
ceria–zirconia promoter appears to lose anion-lattice order-
ing to some extent, a fact that is attributable to the presence
of the noble metal and may be associated with the stabi-
lization of fully oxidized Pd(II) species forming part of the
ceria–zirconia solid solution [29]. This effect might be also
present in the aged PdCZ samples, but the bulk-averaged Ra-
man technique is not able to detect it, due to the significant
fraction of the promoter that presumably does not remain
in contact with the noble metal. This noble metal–promoter
interaction is lost after the phase segregation reaction, as
shown by the recovery of the Raman intensity of the ca.
470 cm−1 peak (compare Figs. 3f and 3g). The Raman spec-
tra thus suggest that a minor part of Pd remains stabilized,
−
4+
Fig. 9. Overall intensity of O –Ce signals obtained after O adsorption
2
2
on the samples indicated. (A) O adsorption at 77 K, sample pretreated
under CO at 453 K; (B) O adsorption at 77 K, sample pretreated under
2
CO + NO at 453 K; (C) O adsorption at RT, sample pretreated under CO
2
2
at 453 K.
able on the basis of previous works to species formed on
2D- and/or 3D-CZ entities at either promoter or Pd–CZ in-
terface sites [21,28,43], are observed in the spectra. For the
sake of simplicity and considering its higher relevance for
this study, exclusively the analysis of the evolution of the
overall integrated intensity of these signals as a function of
the treatment is presented (Fig. 9). The results show a gen-
eral decrease of intensity either upon warming to RT after
O2 adsorption at 77 K on the samples reduced under CO at
453 K or by effect of introducing NO in the pretreatment at
453 K. This decrease must be in both cases a consequence of
oxidation of surface cerium cations, previously reduced by
CO, in accordance with previous reports [21,28,43]. Thus,
the decrease observed upon RT warming of the samples
with O2 adsorbed at 77 K indicates a deeper progress in
the electron transfer from the sample surface toward oxygen
(within the generally accepted oxygen adsorption scheme,
e.g., O2 → O2− → O22− → ··· → 2O2−), giving rise to the
formation of diamagnetic species, e.g., peroxide or oxide an-
ions [43]. On the other hand, the decrease produced in the
presence of NO indicates that some cerium sites are reoxi-
dized upon NO interaction [21,44]. It must be noted that the
magnitude of the electron transfer process responsible for
the surface reoxidation can be related to the reduction degree
achieved in each case after the treatment under CO, not only
in global terms but also at a local level [20,44]. On this basis,
the analysis of Fig. 9 indicates the higher general redox ac-
tivity induced by the presence of Pd. Among Pd-containing
samples, it appears that the redox activity follows the order
PdCZA-a1273 < PdCZA < PdCZA-m1273.
4. Discussion
4.1. Physicochemical effects
Thermal aging influences the size/morphological as well
as chemical properties of all the TWC components, the no-

A. Iglesias-Juez et al. / Journal of Catalysis 221 (2004) 148–161
159
probably in a fully oxidized Pd(II) chemical state, in close
contact with the promoter after aging at 1273 K, and that
the phase segregation may allow this to evolve as Pd single
species, likely maintaining a certain proximity with the pro-
moter.
the main factors governing the catalytic activity of TWC
systems are the palladium dispersion (for active sites con-
stituted by particles in contact with the alumina component,
and considering the mentioned structural insensitivity of this
reaction) and the number and specific properties of Pd–CZ
interface sites. In cases where a sufficient number of the lat-
ter sites are present, e.g., enough contact between the Pd
and CZ components in PdCZA, they have been shown to
dominate the CO oxidation catalytic activity at low tem-
peratures [28,32,43]. Another factor to be considered when
examining this reaction in the presence of NO concerns
the catalytic interferences induced by occupation of active
sites on the metallic particles by species derived from NO
chemisorption [21,48–50]. This interaction produces a kind
of self-poisoning effect as a consequence of the oxidation
of the sites over the metallic particles, clearly visible in the
DRIFTS spectra (Fig. 4).
Taking all the above mentioned factors into account, the
lower activity of the PdCZA-a systems (Fig. 1) as well as the
higher difficulty for Pd reduction (Fig. 6) in them is simply
interpreted as a consequence of the significant Pd sinter-
ing produced upon aging (evidenced by XRD and Raman,
Figs. 1 and 2), which induces the concomitant decrease in
the relative number of Pd–CZ interface sites with respect to
the nonaged PdCZA system. On the other hand, the com-
parison between the fresh PdCZA system and those of the
PdCZA-m series, where only the support suffered the aging
process, appears quite interesting. The absence of significant
differences between them or even the highest CO oxida-
tion activity shown by the PdCZA-m1273 system suggests
differential properties for the Pd–CZ interface sites when
PdCZA-m systems are compared with the fresh PdCZA one.
First, it can be noted that analysis of the DRIFTS spectra in
Figs. 4B and 4C confirms that Pd–CZ interface sites, and
not particles in contact with the alumina component, are the
active sites for CO oxidation in the PdCZA-m systems. To
show this, we must note that the observation of an atop car-
bonyl maintaining its frequency at a relatively high value (ca.
2095 cm−1) for T ꢁ ca. 423 K is a typical fingerprint of the
behavior of Pd particles in contact with the alumina compo-
nent [45,48]. The onset of CO oxidation in such unpromoted
systems coincides with the point at which a red shift is pro-
duced in that band (taking into account, as mentioned before,
that the frequency of that band is determined by dipolar cou-
pling effects and that the reaction is basically limited by
thermally induced desorption of CO). However, the temper-
ature at which the mentioned shift in the carbonyl band is
detected (ca. 423 K) is well above the onset of CO oxida-
tion in the PdCZA-m systems. A careful analysis of the atop
carbonyl region in Figs. 4B and 4C shows the presence of
another component which begins its red shift at lower tem-
peratures and which, in accordance with previous experience
with systems of this kind [21,32,43], must be related to the
palladium particles in contact with the CZ component, which
could then be considered as the most active sites for this re-
action. The fact that DRIFTS observation of features related
4.2. Catalytic effects
In systems of unsupported Pd or Pd on fairly inert sup-
ports (the Pd/Al2O3 can be considered among these lat-
ter), it is well known that the CO oxidation is essentially
a structurally insensitive reaction [45], while NO reduction
is strongly sensitive due to the structural dependence of the
two steps, NO dissociation and N recombination, which may
be involved in the rate-determining step depending on the
noble metal particle size [30,46]. Recent evidence shows,
however, that CO oxidation may display a weak sensitivity
to structure in the case of small Pd particles [47]. In any case,
when both reactants are present in the reactive mixture, the
kinetic regime and the corresponding surface coverage(s) at
the metal particles vary with particle size, in turn affecting
the elimination rates of each specific reactant [48]. Addi-
tionally, in the presence of a Ce-containing promoter, the
size and morphology of the noble metal and the promoter
become critical, since the centers active at low temperature
are located at the metal–promoter interface [21,28,32,43,44,
49,50]. In PdCZ and PdCZA systems it is therefore obvious
that the “structure sensitivity” concerns size and morphology
factors of the metal–promoter interface. It is also inferred
from this that, in addition to geometric/electronic factors
affecting the active metal or metal/promoter interface for
the CZA-supported systems, the metal distribution achieved
over the complex support is of relevance (i.e., whether the
metal shows a preference for interacting with the CZ or the
alumina component). The dynamic nature of the catalytic
system upon interaction with the reactant atmosphere dur-
ing a light-off run requires monitoring the formation of the
potential active phases during the course of the reaction.
In this sense, comparison of catalytic activity (Fig. 1) and
DRIFTS/XANES results (Figs. 4–7) indicates that the light-
off profile roughly parallels the reduction of the noble metal
and its concomitant aggregation, a fact reported in previous
studies [21,47,51]. This fact indicates that the metallic state
(either at the surface or in the bulk) is the most active for any
of the examined reactions, e.g., CO oxidation and NO reduc-
tion. In this respect, the higher difficulty for achieving the
metallic state displayed, according to XANES experiments
(Fig. 6), by the aged systems (-a series) can be considered
as one of the main factors to account for their lower activ-
ity in general terms. Nevertheless, as mentioned above and
according to previous experience on TWC systems [21,28,
32,43–50], the main factors affecting the catalytic activity
are different for each of the reactions and are therefore dis-
cussed separately.
With regard to CO oxidation, as briefly outlined above
and in accordance with previous works [28,32,43,45,46,48],

160
A. Iglesias-Juez et al. / Journal of Catalysis 221 (2004) 148–161
to Pd particles in contact with alumina is more apparent for
the PdCZA-m systems in comparison with the fresh PdCZA
system suggests that a higher proportion of the Pd is in con-
tact with the alumina component for the former catalysts, as
expected when considering the lower general CZ dispersion
in the model systems (Fig. 2 and Table 2). The results thus
indicate a higher CO oxidation catalytic activity of the Pd–
CZ interface sites of the PdCZA-m systems with respect to
the fresh PdCZA sample. This can be related to the changes
observed in the surface characteristics of the CZ component
upon aging; as revealed by EPR (Fig. 8) and reported in pre-
vious studies [52], the promoter surface is enriched in Ce. In
turn, the EPR results of Fig. 9 evidence a strong parallelism
between redox and catalytic activity; therefore, a correlation
can be established between the CO potential to create va-
cancies at the metal–promoter interface and, associated to
this, the reoxidation ability of such vacancies upon interac-
tion with oxygen. The comparison between model and aged
series of catalysts shows that the CZ phase segregation reac-
tion only produces a moderate effect on CO oxidation.
NO reduction is a structure-sensitive reaction which
should be dependent on noble metal size/morphology prop-
erties, particularly influencing NO dissociation and N re-
combination steps [30,46,48]. The analysis of metal particle
size and its effect in the reaction can be extracted from
DRIFTS experiments (Fig. 4). This technique evidences dif-
ferences in palladium particle size already from the starting
point of the reaction; model systems show a larger sur-
face coverage by N-containing adspecies, which yield Pd(I)
surface states after NO dissociation, as well as absence of
bridge-bonded carbonyls until the NO-product(s) disappears
from the metal surface concomitantly to NCO formation.
These aspects indicate a stronger NO interaction with Pd
particles in the PdCZA-m systems in comparison to the
fresh PdCZA catalyst. Additionally, the CO adsorption on
used samples (Fig. 5) gives further evidence of the larger
particle size, according to the higher bridge-to-atop inten-
sity ratio, of model systems (PdCZA-m series). All these
data are in agreement with the fact that model systems have
a larger metal particle size already from the onset of the
reaction (also observed in the Raman spectra, Fig. 3, of
calcined materials) and show higher efficiency than fresh Pd-
CZA in the NO adsorption/dissociation process [48]. All the
“structural” factors, e.g., higher size of the Pd particles and
lower amount of Pd–CZ contacts, are, generally speaking,
in favour of a higher NO reduction activity of the PdCZA-m
systems, in which active sites for this reaction are probably
related to Pd particles in contact with the alumina compo-
nent. Nevertheless, differences between PdCZA-m1273 and
PdCZA-m1373 may suggest a certain role for the metal–
promoter interface. In fact, the EPR study (Fig. 9) shows
that NO dissociation affects the CZ component, at least at
low temperature, and appears somewhat more favored in
the presence of Ce-enriched promoter surfaces. This may
be grounded in the fact that the Ce content of the surface
governs the nature (isolated vs associated) and number of
anion vacancies present in the promoter, which in turn af-
fects the NO activation processes [6,21,38,43,44]. We may
thus suggest that thermal aging affects NO reduction in Pd-
CZA systems by decreasing the number of surface Pd atoms
(sintering of the noble metal) as well as stabilizing Pd(II)
species before the phase segregation reaction, but also may
have a positive effect driven by the Ce enrichment of the
promoter surface in cases where the NO dissociation is in-
volved in the slow rate step [46]. Pd particle size and, to a
limited extent, Ce concentration at the promoter surface ap-
pear as the key physicochemical factors affected by thermal
aging and influencing catalytic activity for NO elimination.
5. Conclusions
The behavior of a Pd-based ceria–zirconia-promoted
TWC system has been analyzed for the CO oxidation and
NO reduction processes under the influence of thermal aging
effects. As already well known, the aging process aggre-
gates the noble metal and promoter components and induces
ceria–zirconia phase segregation above 1273 K. Encapsula-
tion of the noble metal seems of relatively low importance
in our alumina-supported Pd TWC. The thermal degradation
process also appears to increase the surface Ce concentra-
tion of the CZ component and to influence the redox state of
the active metal fraction in close contact with the promoter;
aging below 1273 K seems to stabilize Pd in a presumed ox-
idized state, forming part of a solid solution with promoter
components. Such interaction decreased after the promoter
phase segregation reaction.
The light-off behavior of the Pd-based fresh and aged
TWC samples under the CO+NO+O2 atmosphere appears
dominated by the reduction of the noble metal component
and its sintering. Comparison with the model PdCZA-m se-
ries gives evidence that promoter aging only plays a limited
role in the general decreasing of catalytic activity induced
by thermal effects through the loss of metal–promoter con-
tacts and, indirectly, by the loss of alumina surface area. This
comparison also shows that the interface between Pd and
ceria–zirconia is critical for CO oxidation and less impor-
tant for NO conversion, although the enrichment in cerium
detected during promoter aging after the phase segregation
phenomenon may have a positive influence in the NO disso-
ciation process.
Acknowledgments
We thank the Comunidad de Madrid for a grant received
(A.I.-J.) with which this work has been carried out. We are
also grateful to the scientific and technical staff at SRS Syn-
chrotron (Station 9.2; Project 38285) for their help during
the XANES experiments, A.B. Hungría for XEDS results,
F. Sanchez for EPR data, J.A. Anderson for the use of the
DRIFTS spectrometer, M.A. Bañares for recording Raman

A. Iglesias-Juez et al. / Journal of Catalysis 221 (2004) 148–161
161
data, and J.C. Conesa for fruitful discussions. Financial help
by CICYT (Project MAT2000-1467) and EU (Large Scale
Facility Program) is also acknowledged.
[25] D.J. Kim, J.W. Jang, H.-L. Lee, J. Am. Ceram. Soc. 80 (1997) 1453.
[26] Y. Nagai, T. Yamamoto, T. Tanaka, S. Yoshida, T. Nonaka, T. Okamo-
to, A. Suda, M. Sugiura, Catal. Today 74 (2002) 1453.
[27] J.R. McBridge, K.C. Kass, W.H. Weber, Phys. Rev. B 44 (1991) 5016,
and references therein.
[28] M. Fernández-García, A. Martínez-Arias, A. Iglesias-Juez, A.B. Hun-
gría, J.A. Anderson, J.C. Conesa, J. Soria, Appl. Catal. B 31 (2001)
39.
References
[29] K.R. Priolkar, B. Parthasaratm, P.R. Sarode, M.S. Hegde, S. Emura, R.
Kumashiro, N.P. Lalla, Chem. Mater. 14 (2002) 2120.
[30] X. Xu, P. Chen, D.W. Goodman, J. Phys. Chem. 98 (1994) 9242.
[31] A. Bensalem, J.C. Muller, D. Tessier, F. Bozon-Verduraz, J. Chem.
Soc. Faraday Trans. 92 (1996) 3233.
[1] E.S.J. Lox, B.H. Engler, in: G. Ertl, H. Knözinger, J. Weitkamp (Eds.),
Environmental Catalysis, Wiley–VCH, 1999, p. 1.
[2] R.J. Farrauto, R.M. Hech, Appl. Catal. A 221 (2001) 443.
[3] A. Trovarelli, Catal. Rev. Sci. Eng. 38 (1996) 97, and references
therein.
[4] R. van Yperen, D. Lindner, L. Mubmann, E.S. Lox, T. Kreuzer, Stud.
Surf. Sci. Catal. 116 (1998) 51.
[5] K. Kaspar, P. Fornasiero, M. Graziani, Catal. Today 50 (1999) 285.
[6] M. Fernández-García, A. Martínez-Arias, A. Iglesias-Juez, A.B. Hun-
gría, J.A. Anderson, J.C. Conesa, J. Soria, Phys. Chem. Chem. Phys. 4
(2002) 2473.
[7] E.E. Peterson, A.T. Bell, Catalysis Deactivation, Dekker, New York,
1987.
[8] L. Martín, J.L. Arranz, O. Prieto, R. Trujillano, M.J. Holgado, M.A.
Galán, V. Rives, Appl. Catal. B 44 (2003) 41.
[32] A. Martínez-Arias, M. Fernández-García, A. Iglesias-Juez, A.B. Hun-
gría, J.A. Anderson, J.C. Conesa, J. Soria, Appl. Catal. B 38 (2002)
151.
[33] J.A. Anderson, C. Márquez-Alvarez, M.J. López-Muñoz, I. Rodrí-
guez-Ramos, A. Guerrero-Ruiz, Appl. Catal. B 14 (1997) 189.
[34] K.I. Hadjiivanov, Catal. Rev. Sci. Eng. 42 (2000) 71.
[35] J. Raskó, F. Solymosi, J. Catal. 71 (1981) 219.
[36] F. Solymosi, T. Bansagi, J. Catal. 156 (1995) 75, and references
therein.
[37] S. Kameota, T. Chafit, Y. Ukisu, T. Miyadera, Catal. Lett. 55 (1998)
211.
[38] A. Martínez-Arias, M. Fernández-García, A.B. Hungría, J.C. Conesa,
J. Phys. Chem. B 107 (2003) 2667.
[9] A. Martínez-Arias, M. Fernández-García, A.B. Hungría, A. Iglesias-
Juez, K. Duncan, R. Smith, J.A. Anderson, J.C. Conesa, J. Soria,
J. Catal. 204 (2001) 238.
[10] L. Salin, C. Potvin, J.-F. Tempere, M. Boudart, G. Djega-Mariadassou,
J.-M. Bart, Ind. Eng. Chem. Res. 37 (1998) 4531.
[11] G.W. Graham, H.-W. Jen, W. Chun, R.W. McCabe, J. Catal. 182
(1999) 228.
[39] A. Martínez-Arias, M. Fernández-García, L.N. Salamanca, J.C. Cone-
sa, J. Soria, J. Phys. Chem. B 104 (2000) 4038.
[40] M. Anpo, M. Che, B. Fubini, E. Garrone, E. Giamello, M.C. Paganini,
Top. Catal. 8 (1999) 189.
[12] J.R. González-Velasco, J.A. Botas, R. Ferret, M.P. González-Marcos,
J-.L. Marc, A. Gutiérrez-Ortiz, Catal. Today 59 (2000) 395.
[13] A. Martínez-Arias, M. Fernández-García, V. Ballesteros, L.N. Sala-
manca, C. Otero, J.C. Conesa, J. Soria, Langmuir 15 (1999) 4796.
[14] M. Fernández-García, A. Martínez-Arias, C. Belver, J.A. Anderson,
J.C. Conesa, J. Soria, J. Catal. 194 (2000) 385.
[41] G.W. Graham, H.-W. Jen, R.W. McCabe, A.M. Straccia, Catal. Lett. 67
(2000) 99.
[42] P. Fornasiero, T. Montini, M. Graziani, J. Kašpar, A.B. Hungría, A.
Martínez-Arias, J.C. Conesa, Phys. Chem. Chem. Phys. 4 (2002) 149.
[43] M. Fernández-García, A. Martínez-Arias, L.N. Salamanca, J.M. Coro-
nado, J.A. Anderson, J.C. Conesa, J. Soria, J. Catal. 187 (1999) 474.
[44] A. Martínez-Arias, J. Soria, J.C. Conesa, X.L. Seoane, R. Arcoya, R.
Cataluña, J. Chem. Soc. Faraday Trans. 91 (1995) 1679.
[45] X. Xu, P. Chen, D.W. Goodman, J. Phys. Chem. 98 (1994) 9242.
[46] D.R. Rainer, M. Koranne, S.M. Vesecky, D.W. Goodman, J. Phys.
Chem. B 102 (1997) 10769.
[47] J. Hoffman, I. Meusel, J. Hartmann, J. Libuda, H.-J. Freund, J. Ca-
tal. 204 (2001) 378.
[48] A. Martínez-Arias, A.B. Hungría, M. Fernández-García, A. Iglesias-
Juez, J.A. Anderson, J.C. Conesa, J. Catal., in press, DOI: S0021-
9517(03)0027-1/FLA.
[49] J.H. Holles, R.J. Davis, Th.M. Murray, J.M. Howe, J. Catal. 195 (2000)
193.
[50] P. Bera, V. Jayaram, G.N. Subbana, M.S. Hegde, J. Catal. 196 (2000)
293.
[51] A.L. Tarasov, L.K. Prezheral’skays, V.A. Shevts, V.B. Kazanskii,
Kinet. Katal. 29 (1989) 1020.
[52] A.I. Kuzlov, D.H. Kim, A. Yererets, P. Andersen, H.H. Kung, M.C.
Kung, J. Catal. 209 (2002) 417.
[15] M. Fernández-García, C. Márquez-Alvarez, G.L. Haller, J. Phys.
Chem. 99 (1995) 12 565.
[16] C. Márquez-Alvarez, I. Rodríguez-Ramos, A. Guerrero-Ruiz, G.L.
Haller, M. Fernández-García, J. Am. Chem. Soc. 119 (1997) 2905.
[17] M. Fernández-García, Catal. Rev. Sci. Eng. 44 (2002) 59.
[18] E.R. Malinowsky, Factor Analysis in Chemistry, Wiley, New York,
1991.
[19] A. Martínez-Arias, M. Fernández-García, C. Belver, J.C. Conesa, Ca-
tal. Lett. 65 (2000) 197.
[20] J. Soria, A. Martínez-Arias, J.C. Conesa, J. Chem. Soc. Faraday Trans.
91 (1995) 1669.
[21] A. Martínez-Arias, M. Fernández-García, A. Iglesias-Juez, A.B. Hun-
gría, J.A. Anderson, J.C. Conesa, J. Soria, Appl. Catal. B 31 (2001)
51.
[22] M. Yashima, H. Arashi, M. Kakihana, J. Am. Ceram. Soc. 77 (1994)
1067.
[23] J.R. Ferraro, K. Nakamoto, Introductory Raman Spectra, Academic
Press, New York, 1994.
[24] J.-H. Jehn, J. Phys. Chem. B 102 (1998) 5816.