Structural evolution of Pt/ceria-zirconia TWC catalysts during the oxidation of carbon monoxide

Article abstract of DOI:10.1016/j.jssc.2003.11.001

The structural evolution of two Pt/ceria-zirconia catalysts, characterized by different amounts of supported Pt, was monitored by in situ X-ray diffraction during the anaerobic oxidation of CO at different temperatures. In a first phase, oxygen coming from the surface layers of the ceria-zirconia mixed oxide is consumed and no structural variation of the support is observed. After this induction time, bulk reduction of Pt/ceria-zirconia takes place as a step-like process, while the CO₂ production continues at a nearly constant rate. This behavior is totally different from that of the metal-free support in similar reaction conditions, that show a gradual bulk reduction. In repeated oxidation-reduction cycles, it was observed that the induction time in Pt/ceria-zirconia is a function of the thermal history, of the amount of supported Pt and of the structural evolution of the samples.

Full text of DOI:10.1016/j.jssc.2003.11.001

ARTICLE IN PRESS
Journal of Solid State Chemistry 177 (2004) 1268–1275
Structural evolution of Pt/ceria–zirconia TWC catalysts
during the oxidation of carbon monoxide
A. Martorana,a,b,ꢀ G. Deganello,a,b A. Longo,b A. Prestianni,a L. Liotta,b
A. Macaluso,a G. Pantaleo,b A. Balerna,c and S. Mobilioc,d
a
`
Dipartimento di Chimica Inorganica e Analitica ‘‘Stanislao Cannizzaro’’, Universita di Palermo, viale delle Scienze, 1-90128 Palermo, Italy
b ISMN-CNR, Sezione di Palermo, via Ugo La Malfa 153, 90146 Palermo, Italy
c INFN-LNF, via Enrico Fermi 44, 00044 Frascati, Italy
d Dipartimento di Fisica ‘‘E. Amaldi’’, Universita` di Roma Tre, via della Vasca Navale 84, 00146 Roma, Italy
Received 18 July 2003; received in revised form 27 October 2003; accepted 5 November 2003
Abstract
The structural evolution of two Pt/ceria–zirconia catalysts, characterized by different amounts of supported Pt, was monitored by
in situ X-ray diffraction during the anaerobic oxidation of CO at different temperatures. In a first phase, oxygen coming from the
surface layers of the ceria–zirconia mixed oxide is consumed and no structural variation of the support is observed. After this
induction time, bulk reduction of Pt/ceria–zirconia takes place as a step-like process, while the CO2 production continues at a nearly
constant rate. This behavior is totally different from that of the metal-free support in similar reaction conditions, that show a
gradual bulk reduction. In repeated oxidation–reduction cycles, it was observed that the induction time in Pt/ceria–zirconia is a
function of the thermal history, of the amount of supported Pt and of the structural evolution of the samples.
r 2003 Elsevier Inc. All rights reserved.
Keywords: In situ XRD; synchrotron radiation; TWC; Ceria–zirconia; CO oxidation
1. Introduction
pollutants. Platinum promotes the NO conversion [3],
as well as the oxidation of CO and hydrocarbons [4].
Most of the recent research on TWCs is devoted to
the development of new systems that should comply
with the zero emission demand of the forthcoming
international rules. With this aim, it is important to
improve the catalytic performance at the relatively low
temperatures (473–573 K) of cold-engine regime that
produces the maximum environmental pollution. A
remarkable property of Pt/ceria and Pt/ceria–zirconia
catalysts is their enhanced activity with respect to ceria
alone in the cold-engine regime [5–6], although the
relation activity-structural modification, in different
reaction environments, is not yet clear. Therefore, a
thorough structural characterization of TWC systems
during various treatments could be very important for
the development of better catalysts.
Three-way catalysts (TWC) are constituted of a metal
nanophase supported on modified CeO2. Due to the
peculiarity of ceria that acts as an oxygen buffer, the
TWCs are employed for the exhaust treatment in
gasoline-fuel vehicles to achieve simultaneous control
of CO, hydrocarbons and NO emissions [1]. TWCs
constituted of platinum supported on ceria–zirconia
solid solution show enhanced oxygen-storage properties
with respect to ceria alone [2] and are extensively
investigated. The oxygen-storage capacity of ceria is
based on the redox process of the Ce(IV)/Ce(III) couple,
involving oxygen take up from lean fuel gas and oxygen
delivery during fuel-rich conditions [1]. This property
allows to control the reaction environment that can be
kept within a narrow composition range in order to
ensure the effective catalytic removal of all the
The different ionic radii of Ce(IV) (0.97 A) and
Ce(III) (1.14 A) produce, in reductive or oxidizing
environment, a variation of the lattice constant of
ceria-based promoters that can be monitored by in situ
X-ray diffraction (XRD) [7]. The use of in situ XRD for
ꢀCorresponding author. Fax: +1-39-091-6809399.
E-mail address: cric2@unipa.it (A. Martorana).
0022-4596/$ - see front matter r 2003 Elsevier Inc. All rights reserved.
doi:10.1016/j.jssc.2003.11.001

ARTICLE IN PRESS
A. Martorana et al. / Journal of Solid State Chemistry 177 (2004) 1268–1275
1269
the structural analysis of TWC systems was for the first
time exploited by Ozawa et al. [8] and recently by
Rodriguez et al. [9]. This paper reports on the structural
modifications of two Pt/ceria–zirconia catalysts, 0.5 and
1 Pt wt%, respectively, in different conditions of flowing
gas mixtures and of isothermal treatment temperatures;
this behavior is compared with that of a metal-free
ceria–zirconia sample.
diameters ranging from 80 to 120 mm, were obtained by
meshing. A gas flux heater (Hot-air Blower, Cyberstar,
Grenoble, France) with remote temperature control
allowed the tuning of the sample temperature in the RT–
1223 K range.
The XRD experiments were performed at fixed
wavelength (l ¼ 0:73286 A). The data were elaborated
by the FIT2D program [14], allowing to extract the
XRD patterns from the IP. The structural analyses were
carried out by Rietveld refinement using the GSAS
package [15]. A a-alumina diffraction standard (NIST
Standard Reference Material 676) was used for testing
the effective temperature experienced by the sample
during the heating runs and to determine the sample-to-
detector distance. Furthermore, the investigated samples
were mechanically mixed in a 1:1 ratio with the
chemically inert a-alumina standard. This procedure
allowed to exactly determine the correct 2y scale in the
FIT2D extraction route, by fulfilling the requirement
that the reference lattice constant of a-alumina is
retrieved. The structural parameters and peak shape of
a-alumina, as determined in the calibration runs, were
kept fixed in the Rietveld analyses on the samples.
Pseudo-Voigt peak profiles were used.
2. Experimental
2.1. Sample preparation
The synthesis of ceria–zirconia mixed oxide with
nominal composition Ce0.6Zr0.4O2 was accomplished by
using a sol–gel procedure that provides pure, reprodu-
cible, homogeneous and high thermally resistant materi-
als [10]. Calculated amount of Zr(OCH2CH2CH3)4
(70% in n-propanol, Aldrich) was added to CeðNOÞ3 Á
6H2O (99.999%, Aldrich) dissolved in absolute ethanol.
In order to promote the hydrolysis step, the reaction
temperature was raised from 298 to 338 K. The resulting
sol was stirred overnight and gelation occurred in about
12 h. Then the wet gel, after drying under vacuum, was
calcinated in air at 923 K for 8 h. The resulting solid was
subjected to two reduction/oxidation cycles, up to
1050 K in H2 (5%/Ar) and up to 500 K in O2 (5%/
He), in order to stabilize the phases composition [11,12].
A surface area of 28 m2/g was determined by the BET
method. Two platinum catalysts, 0.5 and 1 wt%,
respectively, from now on PtCZ5 and PtCZ1, were
prepared by impregnation of the ceria–zirconia promo-
ter with a solution of Pt(acac)2 in toluene at 343 K. After
drying under vacuum, the solid was calcined at 673 K
for 5 h in order to decompose the organic ligand.
A first set of in situ treatments, denoted with ‘‘A’’,
were carried out on a fresh PtCZ1 sample at the
temperatures of 473, 553, 593 and 773 K. At these
temperatures, three successive isothermal gas treatments
were performed:
(a) Initially, the sample was fluxed for 30 min with a
10 vol% O2/He mixture (5 mL/min) to ensure a
good and complete oxidation of the sample. In this
step no XRD pattern was collected;
(b) Helium was then fluxed (5 mL/min) for 30 min to
ensure a cleaned reaction atmosphere. Only in the
last 10 min the X-ray beam shutter was opened and
the time-resolved XRD patterns were collected;
(c) Finally, after He switching off, the reductive
mixture CO/He (0.1 vol% CO, 5 mL/min) was sent
for 30 min, keeping the XRD data collection.
Taking into account 10 min of pattern recording
in He flux (point b above), the overall XRD
measurements lasted therefore 40 min.
2.2. In situ XRD experiments and data analysis
The experimental setup for time-resolved XRD
experiments at the GILDA beamline of ESRF [13]
consists of an imaging-plate (IP) detector (Fuji, 200 Â
400 mm2) that translates horizontally behind a vertical
adjustable Ta slit. A fresh IPsurface is therefore
continuously exposed to the diffracted X-rays during
sample treatments, allowing to collect time-resolved
X-ray patterns as narrow vertical strips of the Debye
rings on the detector. To carry out the in situ structural
investigations on the TWCs [7], the sample was loaded
in a quartz capillary, 1 mm internal diameter, supported
by an aluminum frame fitted in a goniometer head. The
capillary was open on both sides allowing the reaction
gases to be input at one end and the products to be
conveyed to a quadrupole mass spectrometer (VG 650
London, England) at the other end. To avoid gas flow
blocking, the coarser powder fraction, with particles of
After the first treatment loop (a–c) at 473 K, the
temperature was raised to 553 K and the loop was
repeated. This procedure was iterated also at 593 and
773 K, so that the same sample was subjected to four
consecutive redox cycles.
A second set of six redox cycles and XRD measure-
ments, denoted with ‘‘B’’, was performed on a fresh
PtCZ1 sample. All the cycles were carried out at 773 K,
except for the fifth, that was made at 593 K. Each redox
cycle was accomplished in the previously described (a–c)
sequence, with the only difference that the CO/He flux

ARTICLE IN PRESS
A. Martorana et al. / Journal of Solid State Chemistry 177 (2004) 1268–1275
1270
(step c) lasted 50 min and therefore that the length of
each XRD data collection was 60 min. According to this
procedure, in the last experiment a fresh PtCZ5 catalyst
was subjected to two redox cycles at 773 K.
Finally, a fresh sample of metal-free ceria–zirconia
promoter was subjected to a redox cycle at 773 K. To
improve the velocity of the structural evolution, the
same gas mixtures of treatments A and B were fluxed at
20 mL/min. The treatment consisted of: flux of O2/He
for 30 min; flux of He for 30 min and recording of the
time-resolved XRD patterns during the last 5 min; flux
of CO/He, keeping the XRD recording, for 55 min.
3. Results
An example of the structural analysis carried out on
the recorded XRD patterns is provided by the first cycle
of sample PtCZ5 in CO/He flux at 773 K. The
corresponding time-resolved XRD image is reported in
Fig. 1a and in the enlargement of Fig. 1b. At a first
glance, broken and continuos lines can be observed:
while the continuous lines are relative to the a-alumina
diffraction standard, that does not undergo any
structural modification except for thermal expansion,
the lines belonging to the ceria–zirconia phase present a
discontinuity originated from the reduction of Ce(IV)
(ionic radius 0.97 A) to Ce(III) (ionic radius 1.12 A). The
larger ionic radius of Ce(III) involves an increase of the
lattice constant and the consequent shift of the diffrac-
tion peaks towards lower 2y values. It is clear, by
inspection of Figs. 1a and b, that the reduction of
cerium takes place in a step-like mode.
After extraction of the XRD patterns using FIT2D,
the Rietveld analysis is carried out allowing for the
presence of a-alumina (space group R-3ch) and of ceria–
zirconia (fluorite-like, space group Fm-3m); lines char-
acteristic of Pt phases, either oxide or metal, were
not observed in any of the recorded patterns. In Figs. 2a
and b, the calculated and residual patterns are reported,
showing the fitting run relative to the strip recorded
after 17 min (7 min after the CO/He switching on). At
this stage, bulk cerium is still in the (IV) oxidation state.
Conversely, the fitting shown in Figs. 3a and b concerns
the strip recorded at the 54th minute of experiment,
when part of bulk Ce(IV) was reduced to Ce(III). Both
the reduced and oxidized patterns were suitably
analyzed with the a-alumina phase and one ceria–
zirconia component with a well-defined lattice constant.
The most evident feature of Fig. 1 is clearly the abrupt
rise of the cubic lattice constant a during the reduction
of the catalyst in CO/He flow. In Figs. 4a and b, the
fitting of a representative XRD pattern recorded during
the structural transition (at the 34th minute of the XRD
experiment) is drawn. The fitting has been carried out
with two cubic ceria–zirconia components: phase 1,
Fig. 1. (a) IPrelative to the reduction of the sample PtCZ5 at 773 ꢁC.
The broken lines are relative to the sudden variation of lattice constant
of the ceria–zirconia mixed oxide; the alumina diffraction standard
(continuous lines) does not undergo structural modification. (b).
Enlargement relative to the (311) ceria–zirconia line, showing in detail
the different behavior of ceria–zirconia and of alumina. The gray scales
are linear with diffracted intensity. Columns and rows are in units of
pixels. The abscissas range (200, 1800) corresponds linearly to the time
interval (6 min: 20 s, 56 min: 45 s); the ordinate scale is not linear in 2y;
due to the geometry of the experimental setup. For reference, the
strong lines at 1600 and B2360 are at B13.5ꢁ 2y and B26.3ꢁ 2y;
respectively.
having a larger amount of Ce(III) ions and, therefore, a
larger lattice constant and phase 2, still oxidized in the
initial stages of the structural transition. The suitability
of introducing two ceria–zirconia components is parti-
cularly evident from inspection of Fig. 4b, relative to the
high 2y angles, where the peaks of the two components
are well resolved. It cannot be excluded (in fact, fitting
runs carried out with three phases gave better agreement
between calculated and observed patterns, although
with higher correlations between fitting parameters) that
even more phases, corresponding to different degrees of
reduction, are present. In Fig. 5 the lattice constants of
phase 1 and phase 2 are drawn; the phase fraction 1 is

ARTICLE IN PRESS
A. Martorana et al. / Journal of Solid State Chemistry 177 (2004) 1268–1275
1271
Fig. 4. Experimental data (J) and fitted intensity (—) relative to the
PtCZ5 sample during the phase of fast structural variation: (a) in the
11–31ꢁ 2y; (b) in the 31–45ꢁ 2y angular interval. Upper tickmarks:
reduced ceria–zirconia; central tickmarks: still-oxidized ceria–zirconia;
lower tickmarks: alumina.
Fig. 2. Experimental data (J) and fitted intensity (—) relative to the
still oxidized PtCZ5 sample: (a) in the 11–31ꢁ 2y; (b) the 31–45ꢁ 2y
angular interval. Upper tickmarks: ceria–zirconia; lower tickmarks:
alumina.
Fig. 5. Lattice constants of the two cubic ceria–zirconia phases
detected in the time-resolved XRD pattern of PtCZ5 during structural
rearrangement. Phase 1 (B) is promptly reduced; phase 2 (&) is still
oxidized in the 32–37 min time interval and undergoes reduction in the
last minutes of the structural transition, while its amount progressively
vanishes as the phase fraction 1 (J, right ordinate scale) tends to 1.
indicates that the structural transition takes place at
slightly different times for different subvolumes of the
sample, while the overall amount of unreacted phase
decreases. After an initially fast reduction, each sub-
volume continues reducing at a slower rate; in the last
minutes of the process, from about 37 to 40 min, the
reduction of the remaining unreacted fraction is
observed. Fig. 6 reports the value of lattice constant of
PtCZ5 during the whole period of the in situ XRD
Fig. 3. Experimental data (J) and fitted intensity (—) relative to the
reduced PtCZ5 sample: (a) in the 11–31ꢁ 2y; (b) in the 31–45ꢁ 2y angular
interval. Upper tickmarks: ceria–zirconia; lower tickmarks: alumina.
also reported, referred to the rightmost ordinate scale.
The presence of a still unreacted phase, clearly
detectable from 32 min to about 37 min in Fig. 5,

ARTICLE IN PRESS
A. Martorana et al. / Journal of Solid State Chemistry 177 (2004) 1268–1275
1272
For the fully oxidized PtCZ5 sample, the lattice constant
of 5.298 A, corrected for thermal expansion, was
determined. This value fully corresponds to the input
preparation amount y ¼ 0:4: The composition of the
sample after the reduction treatment at 773 K is
Ce(IV)0.32Ce(III)0.28Zr0:40O1:86
:
The structural analysis described for the sample
PtCZ5 was carried out for all the in situ XRD data,
mainly concerning the PtCZ1 sample, exhibiting a
reduction behavior very similar to that of PtCZ5. The
catalyst PtCZ1 was subjected to two different redox
treatments, described in the Experimental Section as
procedures A and B. According to A, a first set of redox
treatments was carried out in sequence at the tempera-
tures of 473, 553, 593 and 773 K. The variation of the
lattice constant during the reduction with CO/He
0.1 vol% is reported in Fig. 7, showing that the rise is
steeper and that the height of the jump is larger at
increasing temperatures. Furthermore, an induction
time is present at the different temperatures. The
behavior of the lattice constant during the B treatment
is reported in Fig. 8. It can be observed that the
induction time of the subsequent cycles is abbreviated
after each subsequent reduction treatment, and that the
fifth cycle, although carried out at a lower temperature
(593 K), fits well in this trend. On the other hand, the
jump of lattice constant is accomplished in nearly the
same time for all cycles, no matter of the induction time
and of the lower temperature of the fifth cycle. The
overall amount of Ce(III) in the fifth cycle is slightly less
than at 773 K. Before and after the jump, the cell
constant keeps nearly constant, at least on the time scale
of the XRD experiment.
Fig. 6. Average lattice constant of PtCZ5 (&, left ordinate scale) and
FWHM of the (220) peak (J, right ordinate scale) during the
reduction in CO/He. The errors are of less than 0.01% for lattice
constant and of about 5% for FWHM. The CO/He flux is switched on
at 10 min, while the structural variation begins at about 35 min, with
an induction time of about 25 min.
experiment; the a value during the structural transition
is calculated as the mean of the two a values reported in
Fig. 5 and weighted with the respective phase amounts.
The values of the lattice constants reported in Figs. 5
and 6 are corrected for thermal expansion, assuming
that the thermal expansion coefficient is not dependent
on the degree of reduction. Before and after the jump,
the ceria–zirconia lattice constant does not vary
appreciably. This behavior is typical also of the peak
width, as can be seen in Fig. 6 (right scale), showing the
step-wise decrease of the ceria–zirconia (220) full width
at half maximum (FWHM). By reoxidation, the peak
width is restored to the initial value; it is likely therefore
that the FWHM parameter is associated to an increase
of the effective domain size upon reduction, and not to
the progressive sintering of the ceria–zirconia particles.
As described in the Experimental Section, from t ¼ 0 to
10 min the sample is flowed with helium; then, the
mixture is switched to CO/He for the successive 50 min.
Therefore, the structural rearrangement takes place, for
the sample PtCZ5, after an induction time of about
20 min in the exploited reaction conditions of tempera-
ture and CO/He gas flow.
The comparison between PtCZ1 and PtCZ5 during
the redox cycles at 773 K allowed to see that the a jump
is slightly steeper for PtCZ5 and that its induction time
The chemical formula of the ceria–zirconia phase,
allowing for the partial reduction of cerium, can be
written as Ce(IV)1ꢂxꢂyCe(III)xZryO2ꢂx/2. The stoichio-
metric coefficients can be calculated under the assump-
tion of linear dependence of the lattice constant on the
dopant amount, according to the Vegard’s law. This
hypothesis found several confirmations in literature on
doped ceria [16–18], so that it is likely valid also for the
determination of the dopant amount of Zr(IV) and
Ce(III). The empirical formula proposed by Kim [18]
was used to evaluate as first the amount of zirconium
in the fully oxidized sample and then, assuming that
the Zr content is independent on the gas flow
treatments, the stoichiometric coefficient of Ce(III) [7].
Fig. 7. Variation of the lattice constant of a fresh PtCZ1 sample
during successive redox cycles at increasing temperatures. The errors
are less than 0.01%. The CO/He flux is switched on at 10 min; the
induction time is about the same for the cycles at 553, 593 K while it is
longer for the redox cycle at 773 K. The redox cycle at 473 K was
repeated on a fresh sample and for a longer period, to get more data
about the structural evolution of PtCZ1 at this temperature.

ARTICLE IN PRESS
A. Martorana et al. / Journal of Solid State Chemistry 177 (2004) 1268–1275
1273
Fig. 10. Lattice constant of unsupported ceria–zirconia (J, left
ordinate scale) and FWHM of the (220) peak (&, right ordinate
scale) during the reduction in CO/He at 773 K. The CO/He flux is
switched on at 5 min. The progressive structural rearrangement in
reducing environment is evident.
Fig. 8. Variation of the lattice constant of a fresh PtCZ1 sample
during successive redox cycles at 773 K (but for the fifth cycle, at
593 K). The errors are less than 0.01%. The CO/He flux is switched on
at 10 min; the induction time is larger than 50 min for the first cycle and
decreases for the successive cycles. The redox cycle at 593 K fits well
this trend.
The reduction behavior of the unsupported ceria–
zirconia promoter in CO/He flux is illustrated in Fig. 10,
reporting the cell constant and the FWHM of the (220)
peak as a function of treatment time. The comparison
with Fig. 6 clearly demonstrates that the structural
rearrangement in the metal-free promoter takes place in
a gradual way.
4. Discussion
The main results of this study can be summarized as
follows: (i) the time-resolved XRD patterns show, as a
consequence of the reduction processes, a narrowing of
the diffraction peaks of ceria–zirconia and their shift
towards smaller 2y angles (Fig. 6); (ii) these phenomena,
that take place with an induction time of variable length
with respect to the switching on of the CO flux, are
carried out in a step-like fashion and are completely
reversible through reoxidation of the samples (Figs. 7
and 8); (iii) the production of CO2 begins immediately
after the CO/He flux is sent (Fig. 9), at all the treatment
temperatures, and lasts at a nearly constant rate until
the catalyst can provide oxygen for the anaerobic
oxidation of CO; (iv) the reduction process of unsup-
ported ceria–zirconia (Fig. 10) is completely different
from that of Pt/ceria–zirconia (Fig. 6).
The origin of the different reduction mechanisms of
the two materials, metal-free and Pt-supported ceria–
zirconia, must evidently be ascribed to the presence of
the metal phase. It is well established by experimental
methods [19] and theoretical analysis [20] that the
surface of the supported metal phase constitutes the
preferred site of CO adsorption and that CO oxidation
is mainly carried out at the metal–support interface with
oxygen coming from the support. The observation of an
induction time, during which the CO2 production goes
on at nearly constant rate, while the support does not
Fig. 9. Comparison between lattice constant (left ordinate scale) and
CO2 production (right scale) for a PtCZ1 sample, showing that the
CO2 production starts with no induction at the CO/He switching on
and lasts at a nearly constant (or slightly increasing) rate. The sudden
structural rearrangement takes place in the reported case with an
induction time of about 20 min. It is possible to observe that the CO2
yield is not affected at all by the structural rearrangement taking place
step-wise at about 30 min.
is shortened in the second redox cycle of PtCZ5, like in
the case of PtCZ1. The induction time for the respective
fresh samples is definitely shorter for PtCZ5.
During the XRD measurement, the production of
CO2 was monitored by means of mass quadrupole
spectroscopy. This part of the experiment allowed to
establish that during the flow of CO/He the production
of CO2 is nearly constant, as it is illustrated by Fig. 9,
showing also, for comparison, the simultaneous varia-
tion of the lattice constant. It is clear that the reduction
of Ce(IV) and the CO oxidation proceed in different
ways, and in particular that there is not a direct
dependence of the degree of bulk reduction of the
promoter from the amount of CO2 detected by mass
spectroscopy.

ARTICLE IN PRESS
A. Martorana et al. / Journal of Solid State Chemistry 177 (2004) 1268–1275
1274
show evident bulk modification, can be interpreted
within this scheme of reaction. According to Martin and
Duprez [21], the coefficient for oxygen surface diffusion
of ceria is several orders of magnitude larger than the
corresponding bulk diffusion parameter, while Giorda-
no et al. [22] found that the step of surface reduction is
preliminary to that of bulk reduction in high-surface
(44 m2/g) ceria. On the basis of similar estimates of
oxygen mobility, Holmgren and Andersson [23] pro-
posed that oxygen surface diffusion in a Pt/ceria catalyst
is a fast process with respect to bulk diffusion. Assuming
things in ceria–zirconia not so different from ceria, it
seems sound to conclude that the in situ XRD
experiments on Pt/ceria–zirconia provided evidence that
the first stage of promoter reduction by CO involves the
consumption of surface oxygen at the metal–support
interface and, considering the nearly constant CO2 yield,
that CO2 desorption is rate-limiting.
After the induction period, bulk reduction of ceria–
zirconia takes place through diffusion of oxygen
vacancies into the bulk; this process depends on the
temperature and, in particular at 773 K, is very fast on
the time scale of CO2 production. Also, in this phase no
sharp variation in the rate of CO2 can be observed,
allowing to conclude that, like during surface reduction,
CO2 desorption continues to be the rate-determining
step. The hypothesis that bulk reduction of ceria is a fast
process was already made by Martin and Duprez [21]
and by Giordano et al. [22]. In particular, the latter
authors pointed out that the bulk diffusion length of
oxygen at 700 K is about one order of magnitude larger
than the typical crystallite size for high-surface ceria,
concluding that the oxygen diffusion is not the limiting
step for ceria reduction. Holmgren et al. [24] estimated
that at 623 K oxygen takes about 20 h to completely
diffuse through a 120 A pure ceria crystallite, but
pointed out that this time could be reduced by a factor
of 1000 at 673 K, if the oxide was oxygen defective. The
different behavior of the metal-free support must be
related to the mechanism of CO oxidation, that in this
case can be carried out at any site on the exposed surface
of ceria–zirconia and involves the gradual consumption
of oxygen coming both from the surface and the bulk of
the mixed oxide. On the contrary, in Pt-supported ceria–
zirconia the CO oxidation proceeds mainly at the metal–
support interface and initially involves the more mobile
surface oxygens [25]. Although the matter is still object
of debate [26,27], a possible explanation of this behavior
could be found in the hypothesis proposed for the first
time by Frost [28] that the supported metal particles
improve the concentration of surface oxygen vacancies
of the carrier through a transfer of electronic charge on
the metal. So, the higher concentration of oxygen
vacancies enhances the surface oxygen mobility towards
the interface with the metal and is therefore at the origin
of the induction period for bulk reduction.
The other evident structural modification associated
with bulk reduction produces the narrowing of the
diffraction lines (Fig. 6); also this occurrence is totally
reversible under reoxidation and it cannot therefore be
ascribed to sintering effects. It could be due to the larger
ionic radius of Ce(III) ions, that can partially compen-
sate for the large strain introduced in the structure by
the small Zr(IV) ions [29].
The cycles at 473, 553 and 593 K in Fig. 7 show about
the same induction time (although the initial rate of a
variation at 473 K is so slow that it is hard to precisely
define the beginning of bulk reduction), while the delay
time at 773 K is enlarged (Fig. 7). This larger induction
time with respect to the lower treatment temperatures is
found also in the series of reduction–oxidation cycles
performed at 773 K (Fig. 8) and could be related to a
more effective oxidation. In the successive redox cycles
of treatment B (Fig. 8), the induction time progressively
decreases from more than 50 min to about 17 min and
the fifth cycle at 593 K fits this trend. The data of the in
situ XRD experiments suggest that the induction times
are function of several parameters such as the amount of
supported metal, the treatment temperature and the
history of the sample. It is clear that further investiga-
tions are necessary to get a fully satisfactory explanation
of the metal–support interaction; in particular, X-ray
absorption experiments on the Pt L-edges are planned,
also to overcome the fact that XRD cannot see the Pt
particles. On the basis of the existing evidences, the
sources of variable induction times could be, besides the
effectiveness of the oxidation step, a progressive
modification of the metal–support interaction and/or
the enhanced mobility of bulk oxygens.
The modification of the Pt–ceria interaction by
reduction treatments has been investigated by several
authors in the frame of a strong metal–support
interaction. Mullins and Zhang [30] demonstrated
that reduction pretreatments modify the strength of
CO adsorption on a Pt/ceria catalyst or, if the
pretreatment is carried out at temperatures higher
than 800 K, that a drastic drop in the adsorption
capacity of CO is produced. Bitter et al. [31] used
X-ray absorption spectroscopy to demonstrate that
reduction pretreatments at 773 K induce a modification
of the electronic structure of the supported Pt particles,
while only at higher reduction temperatures decoration
of Pt by ceria can be observed. Zhou et al. [32]
demonstrated by X-ray photoelectron spectroscopy that
Pt deposited on a reduced ceria film diffuse in the oxide
giving rise to the formation of Pt–Ce bonds. Holmgren
et al. [5] found that a mild reduction pretreatment at 473
and 573 K improves the low-temperature catalytic
activity of a Pt/ceria catalyst and that this treatment is
more effective with a moderate rather than with low Pt
dispersion. All these results demonstrate that the
interaction of the metal phase with a ceria-based

ARTICLE IN PRESS
A. Martorana et al. / Journal of Solid State Chemistry 177 (2004) 1268–1275
1275
support is a function of the state of reduction of the
catalytic system.
constant CO2 production rate during the whole process
of ceria–zirconia reduction.
In the course of the anaerobic oxidation of CO
performed during the time-resolved XRD experiments
at 773 K, it is likely that the metal–support interaction is
modified, leading either to an increase of Pt dispersion
originated by diffusion of Pt in ceria–zirconia [32] and/
or to electronic effects [31]. The degree of metal–support
contact affects the catalytic properties [5]. In particular,
the longer induction time of fresh PtCZ1 with respect to
fresh PtCZ5 could be attributed to a less effective metal–
support contact due to lower dispersion of the metal in
PtCZ1. It can be also hypothesized that the in situ
reoxidation treatments are not sufficient to restore
exactly the situation of the metal–support interface of
the fresh catalyst, so that the materials keep memory of
the previous treatments in the iterated redox cycles and,
as a consequence, show a shortening of the induction
times.
References
[1] A. Trovarelli, Catalysis by ceria and related materials, in: G.J.
Hutchings (Series Editor), Catalytic Science Series, Vol. 2,
Imperial College Press, 2002.
[2] J. Kaspar, P. Fornasiero, M. Graziani, Catal. Today 50 (1999)
285–298.
[3] R. Burch, J.P. Breen, F.C. Meunier, Appl. Catal. B 39 (2002)
283–303.
[4] Y.J. Mergler, A. van Aalst, J. van Delft, B.E. Nieuwenhuys, Appl.
Catal. B 10 (1996) 245–261.
[5] A. Holmgren, F. Azarnoush, E. Fridell, Appl. Catal. B. 22 (1999)
49–61.
[6] G. Centi, P. Fornasiero, M. Graziani, J. Kaspar, F. Vazzana,
Top. Catal. 16–17 (2001) 173–180.
[7] A. Martorana, G. Deganello, A. Longo, F. Deganello, L. Liotta,
A. Macaluso, G. Pantaleo, A. Balerna, C. Meneghini, S. Mobilio,
J. Synchrotron Radiat. 10 (2003) 177–182.
As mentioned above, the shortening of the induction
times could also depend on an improved mobility of
bulk oxygens. Mamontov et al. [33] demonstrated by
wide-angle neutron scattering that ceria–zirconia pre-
sents a notable amount of Frenkel defects, with oxygens
that are found in octahedral interstices rather than in the
crystallographic tetrahedral sites. These oxygens can
hardly be forced in the tetrahedral sites even after
prolonged treatment in air at high temperature and are
more available for exchange with the reaction environ-
ment [33]. It is possible therefore that after each
reoxidation step an increasing amount of oxygens do
not return in the regular tetrahedral site, remaining in a
higher-energy state in the more roomy octahedral sites
and producing a net increment of mobile anions in the
successive reduction–oxidation cycles. The behavior of
metal-free ceria–zirconia seems to support this hypoth-
esis (A. Martorana et al., in preparation).
[8] M. Ozawa, M. Kimura, A. Isogai, J. Mater. Sci. 26 (1991)
4818–4822.
[9] J.A. Rodriguez, J.C. Hanson, J.-Y. Kim, G. Liu, A. Iglesias-Juez,
M. Fernandez-Garcia, J. Phys. Chem. B 107 (2003) 3535–3543.
[10] R.D. Gonzalez, T. Lopez, R. Gomez, Catal. Today 35 (1997)
293–317.
[11] L.F. Liotta, A. Macaluso, G. Pantaleo, A. Longo, A. Martorana,
`
G. Deganello, G. Marcı, S. Gialanella, J. Sol–Gel Sci. Technol. 26
(2003) 235–240.
[12] L.F. Liotta, A. Macaluso, A. Longo, G. Pantaleo, A. Martorana,
G. Deganello, Appl. Catal. A 240 (2003) 295–307.
[13] C. Meneghini, G. Artioli, A. Balerna, A. Gualtieri, P. Norby, S.
Mobilio, J. Synchrotron Radiat. 8 (2001) 1162–1166.
[14] A. Hammersley, ESRF Copyright, 1987–2001.
[15] A.C. Larson, R.B. Von Dreele, GSAS, The general structure
analysis system, Los Alamos National Laboratory, 1991–2001.
[16] T. Miki, T. Ogawa, M. Haneda, N. Kakuta, A. Ueno, S. Tateishi,
S. Matsuura, M. Sato, J. Phys. Chem. 94 (1990) 6464–6467.
[17] S.J. Hong, A.V. Virkar, J. Am. Ceram. Soc. 78 (1995) 433–439.
[18] D.J. Kim, J. Am. Ceram. Soc. 72 (1989) 1415–1421.
[19] T. Bunluesin, H. Cordatos, R.J. Gorte, J. Catal. 157 (1995) 222–
226.
[20] V.P. Zhdanov, B. Kasemo, Appl. Surf. Sci. 135 (1998) 297–306.
[21] D. Martin, D. Duprez, J. Phys. Chem. 100 (1996) 9429–9438.
[22] F. Giordano, A. Trovarelli, C. de Leitenburg, M. Giona, J. Catal.
193 (2000) 273–282.
5. Conclusions
The structural modification of Pt/ceria–zirconia TWC
catalysts during the anaerobic oxidation of CO is
studied by XRD in situ. The technique allowed to
establish that a first stage of the support reduction
involves consumption of surface oxygens and that only
after an induction time of variable length, the bulk
reduction takes place as a fast process. The variable
induction time probably depends on the metal–support
interaction, leading to a modified metal–support inter-
face and to fine tuning of the electronic properties of the
catalyst, and on the presence of an increasing amount of
oxygens in the interstitial octahedral sites, more avail-
able to exchanges with the reaction environment. Both
surface and bulk oxygen diffusion are faster than CO2
desorption. This conclusion can account for the nearly
[23] A. Holmgren, B. Andersson, J. Catal. 178 (1998) 14–25.
[24] A. Holmgren, D. Duprez, B. Andersson, J. Catal. 182 (1998)
441–448.
[25] C. Bozo, N. Guilhaume, E. Garbowski, M. Primet, Catal. Today
59 (2000) 33–45.
[26] T. Ioannides, X.E. Verykios, J. Catal. 161 (1996) 560–569.
[27] V.P. Zhdanov, Surf. Sci. 512 (2002) L331–L334.
[28] J.C. Frost, Nature 334 (1988) 577–581.
[29] G. Balducci, J. Kaspar, P. Fornasiero, M. Graziani, M. Saiful
Islam, J.D. Gale, J. Phys. Chem. B 101 (1997) 1750–1753.
[30] D.R. Mullins, K.Z. Zhang, Surf. Sci. 513 (2002) 163–173.
[31] J.H. Bitter, M.A. Cauqui, J.M. Gatica, S. Bernal, D.E. Ramaker,
C.D. Koningsberger, Stud. Surf. Sci. Catal. 130 (2000) 3183–3188.
[32] Y. Zhou, M. Nakashima, J.M. White, J. Phys. Chem. 92 (1988)
812–818.
[33] E. Mamontov, T. Egami, R. Brezny, M. Koranne, S. Tyagi,
J. Phys. Chem. B 104 (2000) 11110–11116.