Flame-made nanocrystalline ceria/zirconia: Structural properties and dynamic oxygen exchange capacity

Article abstract of DOI:10.1016/S0021-9517(03)00235-5

The current new generation of three-way catalysts uses ceria/zirconia as key components for its high dynamic oxygen exchange capacity. Different production methods lead to different molecular mixing ceria and zirconia and varying contents of defects and cracks in the material. Ceria/zirconia nanocrystals were prepared by flame-spray synthesis. Application of carboxylic acid-derived solvents allowed the continuous production of homogeneous mixed oxides with narrow particle-distribution. The high preparation temperature favored the formation of a highly crystalline solid solution of the constituents. Product homogeneity and phase stability were directly related to a homogeneous delivery of both constituent precursors from the spray droplet into the flame. Stable solid solutions were obtained for all compositions of ceria/zirconia. Both in terms of reduction behavior and dynamic oxygen exchange capacity, the flame-made samples showed similar properties as highly sintered ceria/zirconia. With respect to specific surface area, the flame-made nanocrystals were much more stable toward temperature impact than corresponding precipitated materials.

Full text of DOI:10.1016/S0021-9517(03)00235-5

Journal of Catalysis 220 (2003) 35–43
www.elsevier.com/locate/jcat
Flame-made nanocrystalline ceria/zirconia: structural properties and
dynamic oxygen exchange capacity
a,b
a
b
Wendelin J. Stark, Marek Maciejewski, Lutz Mädler,
b
a,∗
Sotiris E. Pratsinis, and Alfons Baiker
a
Institute for Chemical and Bioengineering, ETH Hönggerberg, CH-8093 Zurich, Switzerland
b
Particle Technology Laboratory, ETH Zentrum, CH-8092 Zurich, Switzerland
Received 2 January 2003; revised 20 May 2003; accepted 20 May 2003
Abstract
Ceria/zirconia nanocrystals were prepared by flame-spray synthesis. Application of carboxylic acid-derived solvents allowed the continu-
ous production of homogeneous mixed oxides with narrow particle-size distribution. The high preparation temperature favored the formation
of a highly crystalline solid solution of the constituents. Structural and textural properties of the ceria-zirconia nanoparticles were charac-
terized by high-resolution electron microscopy, nitrogen adsorption, and X-ray diffraction. The redox properties of the nanoparticles were
investigated by temperature-programmed reduction and by dynamic oxygen exchange capacity (OEC) measurements using H , CO, and
2
propene as reductants. For CO and H a similar reduction was found, whereas propene afforded a considerably higher degree of reduction.
2
2
◦
While the flame-made powders had very stable specific surface area (up to 80 m /g) even after severe calcination (2 h at 900 C in air) their
OEC was lower than in corresponding ceria-zirconia prepared by coprecipitation. This difference is traced to the high crystallinity and low
defect concentration of the flame-made material.

2003 Elsevier Inc. All rights reserved.
Keywords: Ceria/zirconia mixed oxide; Flame-spray synthesis; Nanoparticles; Structural stability; Oxygen exchange capacity; Pulse thermal analysis
1
. Introduction
Coprecipitation of ceria/zirconia can lead to mixed oxide
powders with extremely high specific surface area [3,4]. Un-
fortunately, the temperature stability of as-prepared oxides
is characterized by a severe loss of surface area at elevated
temperature. Preparation at high temperature may produce
an oxide with increased stability. This has prompted several
authors to prepare ceria and ceria containing oxides by flame
methods which led to a series of patents [5–10]. These meth-
ods either use expensive precursors, such as alkoxides [8,
The current new generation of three-way catalysts (TWC)
uses ceria/zirconia as key components for its high dynamic
oxygen exchange capacity [1]. In the treatment of noxious
gases from car exhaust, the ceria switches between its two
major oxidation states Ce(III) and Ce(IV) [2]. The temper-
ature in a typical automotive exhaust rapidly changes and
very high temperature causes severe thermal stress. There-
fore, stability is a major issue in TWC research as reflected
in a wealth of scientific papers. It is well established that
formation of ceria-zirconia solid solution greatly enhances
reducibility of the ceria [2]. Different production methods,
however, lead to different molecular mixing of ceria and
zirconia and varying contents of defects and cracks in the
material. Maximum stability is found for intensively mixed
powders forming a stable solid solution of the constituents.
9
], demanding atomization devices [7,11] and tend to pro-
duce inhomogeneous powders containing large lumps [10].
Aruna et al. investigated the ceria/zirconia synthesis by com-
busting mixtures of redox compounds (urea) and oxidiz-
ing metal precursors (nitrates) [12]. This high temperature
preparation yielded a high surface area product with excel-
lent phase mixing. It is a batch process, however, and ex-
plosion hazard causes great difficulties. Laine et al. prepared
ceria/zirconia by flame spray synthesis but the specific sur-
face area of the product powder stayed low, typically at 10
2
*
to 15 m /g [13,14]. Sutorik et al. [15] very recently pro-
Corresponding author.
E-mail address: baiker@tech.chem.ethz.ch (A. Baiker).
duced ceria/zirconia mixed oxides by flame spray synthesis
0021-9517/$ – see front matter  2003 Elsevier Inc. All rights reserved.
doi:10.1016/S0021-9517(03)00235-5

36
W.J. Stark et al. / Journal of Catalysis 220 (2003) 35–43
showing formation of solid solutions, but obtained rather low
specific surface area while no tests on the oxygen exchange
capacity were undertaken. It was shown recently that com-
plex mixed oxides with high specific surface areas could be
prepared at even higher temperatures by using continuous
flame and flame-spray pyrolysis [16].
Here, we report the use of continuous flame-spray pyroly-
sis for production of high specific surface area ceria/zirconia
from low-cost precursors [17,18] and compared them to
samples prepared by coprecipitation [19]. Since volatile or-
ganic components (VOC) are a major component in auto-
motive exhaust gas, the role of propene is investigated as a
reductant in OEC measurements beside CO and hydrogen.
2.2. Characterization
2.2.1. Specific surface area and isotherms
The specific surface areas of the collected powders were
analyzed by nitrogen adsorption at 77 K using the BET
method (Micrometrics Gemini 2360). The results were
cross-checked by recording a full adsorption isotherm (Mi-
crometrics ASAP 2010 Multigas system).
2.2.2. High-resolution electron microscopy
The HRTEM investigations were performed on a CM30ST
microscope (Philips, LaB6 cathode, 300 kV, point resolution
2 Å). Particles were deposited onto a carbon foil supported
on a copper grid.
2
.2.3. X-ray diffraction
XRD diagrams were measured on a Bruker D 8 Ad-
2
. Experimental
◦
vance diffractometer (Cu-Kα radiation) from 20 to 75 at
◦
◦
2
.1. Preparation of mixed oxides
a step size of 0.02 and a scan speed of 0.24 /min. Crys-
tallite characteristics and average crystal size were obtained
from the XRD diagram using the Topas 2.0 software (Bruker
AXS, 2000) by the fundamental parameter approach (Ri-
etveld method).
Ceria/zirconia mixed oxide powders were produced by
flame-spray pyrolysis in a laboratory scale setup [17].
Cerium (III) acetate hydrate (Aldrich, > 99.7%) and zir-
conium tetraacetylacetonate (Aldrich, 99%) were mixed
according to the product composition and dissolved in a
lauric/acetic acid mixture (1/1 by weight), heated to full dis-
solution resulting in a total metal concentration of 0.15 M.
In a preliminary study [18], the lauric/acetic acid mixture
proofed to be the most suitable carrier liquid for the pro-
duction of high surface area ceria/zirconia. Solutions were
delivered into the flame at 5 ml/min by a syringe pump
2
.2.4. Thermal analysis (TA)
Thermoanalytical experiments were carried out both
isothermally and nonisothermally (heating rates 3, 5, or
0 K/min) on a Netzsch STA 409 thermoanalyzer equipped
1
with a gas injector (PulseTA box, Netzsch) which allows for
the injection of a defined sample gas volume into the carrier
gas stream [21]. If not stated otherwise, the amount of in-
jected gases (hydrogen, oxygen, propene, carbon monoxide)
was 1.0 ml. The total reduction of the samples was carried
out in the atmosphere of 20 vol% H2, balance Ar with a rate
of 5 K/min in the range 300–1075 C. Gases evolved during
reaction and/or injected into the system were monitored on-
line with a Balzers QMG 420 quadrupole mass spectrometer
(
9
Inotec R 232) and atomized by 5 L/min oxygen (Pan Gas,
9.8%) dispersion gas. The apparatus [17] consisted of a
◦
central spray delivery (two-phase nozzle), a premixed, cir-
cular support flame (diameter 6 mm, slit width 10 µm), and
a circular sheet gas delivery (ring of sinter metal, inner di-
ameter 11 mm, outer diameter 18 mm, 5 L/min oxygen).
A mixture of methane (1.5 L/min, Pan Gas, 99%) and oxy-
gen (2 L/min) was fed to the inner slit and formed the
premixed flame. All gas flow rates were controlled by cal-
ibrated mass flow controllers (Bronkhorst EL-Flow F201).
Powders were denoted as CexZr1−xO2, where x (0 < x < 1)
is the amount of ceria in the samples. Depending on com-
position, a nominal production rate of 3.3 to 4.6 g/h was
achieved.
◦
connected to the thermoanalyzer by a heated (ca. 200 C)
capillary.
3. Results
3
.1. Structural properties
Spraying corresponding mixtures of cerium and zirco-
Reference samples were prepared by coprecipitation ac-
cording to Leitenburg et al. [19]: In brief, corresponding
solutions of cerium nitrate (Fluka, > 99.5%) and zirconium
oxynitrate (Fluka, > 99%) in distilled water at a total con-
centration of 0.2 M were dropped into concentrated aqueous
solution of ammonia (24 mass%) under stirring. The solu-
nium containing precursor solutions into a methane/oxygen
flame continuously produced ceria/zirconia nanoparticles.
Fig. 1 (left side) depicts transmission electron micrographs
of as-prepared ceria/zirconia Ce0.5Zr0.5O . The product con-
2
sists of well-crystalline, sharp-edged nanoparticles of 4–
10 nm size [18]. Fig. 1 (right side) further shows the same
◦
tions were filtered after 2 h and washed with distilled water
powder after sintering at 900 C for 2 h in air. Regularly
◦
6
times. The yellowish to white slurry was dried at 200 C
shaped particles of high crystallinity are obtained as the
high-resolution images (bottom) reveal. They are similar to
for 2 h.

W.J. Stark et al. / Journal of Catalysis 220 (2003) 35–43
37
◦
O (left), and corresponding samples after sintering at 900 C, 2 h in air (right).
2
Fig. 1. High-resolution electron micrographs of as-prepared Ce Zr
0
.5 0.5
The crystallites at high magnification are shown on the bottom: Flame-made ceria/zirconia forms highly crystalline nanoparticles of narrow particle-size
distribution. Sintering affords well-defined, regular crystallites of about 10 nm diameter. Surfaces are well-structured and flat; no defects or twinned crystals
appear.
2
specific surface area up to 170 m /g were obtained that
decreased with higher zirconia content. The correspond-
ing particle sizes were similar since the density of zirconia
3
3
(
5.9 g/cm ) is lower than that of ceria (7.1 g/cm ). After sin-
◦
tering in air at 900 C for 2 h the pure constituents lost much
specific surface area, whereas mixed oxides were more sta-
2
ble and conserved up to 82 m /g for x = 0.3.
Fig. 3A shows the XRD patterns of ceria/zirconia
CexZr1−xO2 ranging from pure ceria (x = 1, bottom) to
pure zirconia (x = 0). The broad signals from the cubic ce-
ria phase slowly shift to higher theta values. Considerable
transition occurs leading to widening of the peaks. At high
zirconia content (x = 0.2 and 0.1) the mixed oxides show
some indication of phase instability and weak reflections of
pure zirconia appear on top of the reflections of the solid
solution phase.
Fig. 2. Specific surface area of mixed oxide samples as prepared and after
calcination ranging from pure ceria (left) to pure zirconia (right). While
as-prepared particles are of similar size since the density of zirconia is
higher than that of ceria, a considerable loss of specific surface area is found
for the pure oxides upon sintering. Mixed oxides are rather stable and con-
Fig. 3B gives the XRD results for calcined ceria/zirconia
900 C, 2 h, air). The reflections generally narrow due to
2
◦
serve up to 80 m /g surface area after 2 h at 900 C in air.
◦
(
larger crystallite size. Samples of high ceria content (x >
0.7) exhibit a ceria-like XRD pattern with shifted position
of reflections. At intermediate composition (0.4 < x < 0.8)
broad reflections appear and a transition from ceria-like to
zirconia-like patterns can be observed. At high zirconia con-
tent (x = 0.2 and 0.1) the diffractogram resembles the pat-
earlier studies of flame-spray-made pure ceria nanoparti-
cles [17].
Fig. 2 shows the specific surface area of as-prepared
◦
CexZr1−xO2 (squares) and calcined powders (circles, 900 C,
2
h in air) as a function of the Ce content x. Powders with

38
W.J. Stark et al. / Journal of Catalysis 220 (2003) 35–43
(A)
(
B)
Fig. 5. (A) Adsorption isotherms of flame-made (squares, Type IV accord-
ing to IUPAC) and precipitated Ce Zr (Type I). (B) Corresponding
pore-size distributions as calculated from the isotherms. The precipitated
material contains a large number of mesopores. The flame-made material
shows almost no micropores which is consistent with the observation of
dense, rather regularly shaped particles (Fig. 1).
O
0.5 0.5 2
Fig. 3. X-ray diffraction pattern of as-prepared ceria/zirconia ranging from
pure ceria (x = 1) to pure zirconia (x = 0). (A) As prepared, a continuous
shift in peak reflections corroborates the mixing of ceria and zirconia. (B)
◦
After calcination in air at 900 C. Powders containing up to 80% zirconia
are stable against phase transformation. At 90% zirconia a small peak split
indicates the onset of a new phase. Pure zirconia undergoes a transformation
to the tetragonal and monoclinic phase.
ides. The flame-made oxides from this study exhibit high
specific surface area in a wide range of compositions. For
comparison a commercial catalyst from Rhodia is shown af-
◦
ter calcination for 6 h at 900 C. Leitenburg et al. [19] and
Bozo et al. [20] precipitated ceria/zirconia from the corre-
2
sponding nitrates and achieved up to 26 m /g after calci-
◦
nation at 900 C for 2 h in air. Note that flame-spray-made
powders of previous studies report specific surface area of
2
about 10 m /g [14,16] even without calcination.
Fig. 5A compares the nitrogen adsorption/desorption
isotherms of Ce0.5Zr0.5O2 prepared by precipitation (cir-
cles) and by flame-spray synthesis (squares). According to
IUPAC, the precipitated sample is characterized by a Type I
isotherm with prominent hysteresis. The flame-made sample
exhibits a Type IV isotherm with a weak hysteresis. Fig. 5B
depicts the corresponding differential pore size distribution
of precipitated and flame-made ceria/zirconia. While the
precipitated sample contains mainly pores of 4–10 nm in
diameter and few pores at higher diameters, the flame-made
sample shows few small pores probably originating from
necks between particles and the largest contribution from
interparticle pores 20–100 nm in diameter.
Fig. 4. Specific surface area of different ceria/zirconia prepared by precip-
itation of the corresponding nitrates (diamonds [20]), commercial products
from Rhodia (down triangles [22]), and flame-made ceria/zirconia from Su-
torik and Baliat (up triangles [16]), and this work (squares).
tern of tetragonal zirconia. Pure zirconia (x = 0, top line)
exhibits the typical pattern of monoclinic zirconia while re-
taining some tetragonal phase and obviously consists of a
mixture of the two modifications.
Fig. 4 further compares the specific surface area of flame-
spray-made ceria, ceria/zirconia, and precipitated mixed ox-

W.J. Stark et al. / Journal of Catalysis 220 (2003) 35–43
39
Fig. 6. Reduction behavior of flame-made ceria/zirconia. On top, the reduc-
tion rate of 50-mg samples of increasing temperature is shown. The degree
of reduction calculated for the ceria in the sample is indicated at the bot-
tom. Pure ceria is characterized by a prominent peak at ca. 700 C. Adding
Fig. 7. Schematic presentation of measuring principle for determining the
oxygen exchange capacity by pulse thermal analysis. Samples (50 mg) were
reduced at 700 C by pulses of reductant (H2, CO, and propene) and reox-
◦
◦
zirconia results in a much lower reduction temperature.
idized by pulses of O . Gases injected and evolved during redox processes
2
were monitored by mass spectrometry.
3
.2. Redox properties
Fig. 6 depicts reduction rate (top) and reduction pro-
files (bottom) for pure ceria (CeO2), 40% ceria/zirconia
Ce0.6Zr0.4O2), and 70% ceria/zirconia (Ce0.3Zr0.7O2) mea-
(
sured under 20 vol% hydrogen in argon on a thermobalance.
The reduction rate is obtained by differentiating the TG
◦
curve. Pure ceria shows a single reduction peak at 700 C.
The degree of ceria reduction calculated from the mass loss
(
TG) is presented as a function of temperature. Zirconia
◦
is not reduced in the range 200–1100 C. The increase of
zirconia content promotes the reduction of ceria, the sam-
ple Ce0.3Zr0.7O2 was almost fully reduced to Ce(III) above
◦
1
000 C.
Fig. 8. Mass changes of ceria/zirconia samples (50.0 mg) resulting from
◦
Fig. 7 illustrates the principle of dynamic oxygen storage
pulses of the reducing agent (H , CO, 2 ml, propene, 1 ml) at 700 C. After
2
reduction the samples were fully reoxidized by 1 ml oxygen pulses.
capacity measurements performed on the TG-MS system us-
ing the pulse TA technique [21]. Reduction is very fast for
all reducing agents (hydrogen, CO, and propene) and steady
state is attained very fast. After oxidation, the sample mass
reaches exactly the same level as prior to the experiment
corroborating full reversibility of the redox process. The ap-
pearance of a strong hydrogen signal (m/z = 2) in the MS
trace indicates that excess hydrogen was provided in the
pulse. Incomplete conversion could be excluded by applying
a second hydrogen pulse that did not cause further reduc-
tion. While hydrogen and CO result in a comparable degree
of reduction, propene removes much more oxygen and two
O2 pulses are needed for reoxidation.
(open symbols). Using H or CO as reductant, the pre-
cipitated materials showed superior OEC compared to the
2
flame-made ones. This tendency was less clear when using
propene. At higher zirconia loading (x < 0.2), the differ-
ences between the results obtained with the different reduc-
tion agents disappeared, mainly due to high conversion in
both cases. The maximum reducibility was found for the
samples containing around 50% Ce (when reduced by CO
or H ) and about 60–75% Ce (reduction by propene).
2
Fig. 9 depicts high-resolution electron micrographs of
precipitated Ce0.5Zr0.5O . The morphology may be char-
2
Fig. 8 gives the dynamic oxygen exchange capacity
acterized as a heavily aggregated, rather dense conglomer-
(
OEC) of ceria/zirconia for 11 samples ranging from pure
ate of small crystallites (left). Sintering the precipitated ce-
ria/zirconia increased its crystal size and further compacted
the structure. Compared to the flame-made nanocrystals, the
ceria (left, x = 1) to pure zirconia (right, x = 0) for flame-
made oxides (filled symbols) and 5 precipitated materials

40
W.J. Stark et al. / Journal of Catalysis 220 (2003) 35–43
◦
Fig. 9. High-resolution electron microscopy images of precipitated Ce Zr O as prepared and after sintering (900 C, 2 h in air). The fresh material consists
0
.5 0.5 2
of ill-defined, partially amorphous ceria/zirconia. Sintering results in large, heavily agglomerated and twinned crystals with a rough surface (see inset). A clear
contrast is seen to the flame-made materials where a smooth surface and well-defined crystallites are observed (Fig. 1).
Fig. 11. Comparison of the dynamic oxygen exchange capacity of flame-
made and precipitated CeO and Ce Zr
O at increasing temperatures.
2
2
0.5 0.5
At higher temperatures, the precipitated and flame-made materials become
similar since defects are annealed in the precipitated sample due to the heat
impact. In unstabilized ceria, the pretreatment step has healed out the de-
fects rendering the precipitated similar to the flame-made ceria.
presented in Fig. 8, the reduction is favored by the addition
of Zr.
Fig. 11 compares the dynamic oxygen exchange capac-
ity at increasing temperature of flame-made and precipitated
CeO2 and Ce0.5Zr0.5O2. Samples were sintered at 700 C,
Fig. 10. Mass change of the 50-mg sample Ce Zr O due to 4-ml pulses
0.6 0.4 2
of hydrogen followed by 2-ml pulses of oxygen at different temperatures.
The insert explains the construction of the dependence ꢀm−T used for the
calculation. Due to full reoxidation by the oxygen pulses, the composition
of the sample is always the same prior to the hydrogen pulses.
◦
1
6 h in air prior to the measurements. The conventionally
prepared Ce0.5Zr0.5O2 shows a higher degree of dynamic re-
duction/oxidation than the flame-made samples. At higher
temperature, however, the behavior of both materials is sim-
ilar. In contrast, corresponding ceria samples show very sim-
ilar behavior over the whole range of temperature.
surface of the precipitated material is rather rough and irreg-
ular.
Fig. 10 shows the TG trace for several consecutive
reduction-oxidation pulses for Ce0.6Zr0.4O2 when increas-
ing the temperature (top) from 300 to 1150 C. The resulting
4
. Discussion
◦
degree of reduction as a function of temperature for samples
ranging from pure ceria to 10% ceria/zirconia is shown in the
lower part of the figure. Ceria is more reduced in zirconia-
rich samples at all temperatures. As in the case of the results
4.1. Flame-spray synthesis
Fig. 1 shows homogeneous nanostructured ceria/zirconia.
No large lumps of ceria could be detected, corroborating

W.J. Stark et al. / Journal of Catalysis 220 (2003) 35–43
41
good dispersion of the metal precursor and subsequent pre-
cipitation in the flame. For pure ceria, high combustion heat
of the carrier liquid is crucial for obtaining a homogeneous
material [17]. A droplet entering the flame may continu-
ously release precursor solution or an enrichment of the
metal may take place if the carrier liquid (fuel) leaves the
droplet first. In the latter case, a nonuniform metal con-
centration in the flame with locally high amounts of metal
oxide may favor the formation of larger particles and broad
particle-size distribution. Furthermore, if two or more metal
precursors are present in the droplet, element segregation
from different evaporation and decomposition rates may re-
sult in inhomogeneous products. Here, it was shown that by
using an acetic/lauric acid carrier liquid instead of isooc-
tane/acetic acid/2-butanol [17] elemental segregation and
resulting phase instability could be avoided [18]. The for-
mation of a single-phased mixed oxide with particles of few
nanometers in size corroborates the excellent degree of mix-
ing of ceria and zirconia in the flame. The continuous shift of
pertinent reflections in XRD patterns (Fig. 3) confirms that
this holds true for compositions up to 80% zirconia.
presented a low-cost route to ceria/zirconia. They used an ul-
trasonic mister to aerolize ethanol-based precursor solutions.
Metal precursors were kept in solution as separately pro-
duced derivatives of tris-ethanolamin. The process resulted
in rather large, unagglomerated ceria/zirconia with a spe-
2
cific surface area of 10 m /g but at larger production rates
than reported here. Besides providing a one-step preparation,
the carboxylic acid-derived solvent (acetic/lauric acid) used
in this study has a higher combustion heat, allowing higher
temperature and faster gas flow in the flame-spray reactor.
Consequently, shorter residence time limits particle growth,
thus favoring smaller particles. With carboxylic acid’s po-
tential for violent droplet explosions, it may favor the homo-
geneous distribution of the precursors in the flame [18].
4
.3. Reduction behavior
Even though Hickey et al. [25] recently showed that
temperature-programmed reduction (TPR) provides a rough
estimate of oxygen storage capacity, at best, the reduction
profiles give some indication on the type and abundance of
different ceria species in a sample. In Fig. 6, pure flame-
made ceria showed a prominent reduction peak at around
4
.2. Structural properties
◦
7
00 C. According to Giordano et al. [26], this ceria may be
classified as low-surface area ceria (LSA), based on its re-
duction behavior. Their LSA ceria had a specific surface area
A comparison of precipitated [20], surfactant-assisted
precipitated [3,4], commercial [22], and flame-made ce-
ria/zirconia from Sutorik and Baliat [15] proves the excellent
sintering stability of the flame-made nanoparticles. Thermal
stability is associated with a high degree of mixing of the
constituents forming the solid solution, an open morphology,
and annealed crystallites with few defects [18]. The phase
diagram of ceria/zirconia indicates a mixed solid solution
as the stable form of intermediate ceria/zirconia mixtures
2
of 3 m /g, whereas the flame-made ceria with a very similar
2
reduction behavior exhibits 39 m /g. The high surface area
ceria (HSA) reported by Giordano et al. [26] had a specific
2
surface area of 44 m /g. Its reduction behavior was char-
◦
acterized by an additional, large reduction signal at 500 C.
Interestingly, the flame process affords a high surface area
ceria with otherwise similar characteristics as classical LSA
ceria.
For mixed ceria/zirconia, Fornasiero et al. [27] reduced
precipitated Ce0.5Zr0.5O2 with hydrogen and found maxi-
◦
for temperatures up to ca. 1000 C [23,24]. Calcination of
◦
flame-made powders at 900 C (Fig. 3B) did not result in
significant phase separation up to 80% zirconia and there-
fore corroborated a good ceria distribution. The phase sta-
bility of ceria/zirconia is not straightforward and depends
on the preparation method. In the case of precipitated ce-
ria/zirconia, Bozo et al. [20] found considerable instability in
◦
mum hydrogen uptake at 400 and 630 C. Similarly, Hickey
◦
et al. [28] reported maximum reduction at 520 to 570 C for
precipitated Ce0.6Zr0.4O2 of only 4 m /g surface area. Com-
mercial Rhodia Ce0.5Zr0.5O2 is reduced at around 500 C
29] and the amount of Ce(III) formed at 700 C increased
2
◦
◦
[
◦
the solid solution for powders calcined at 900 C. The high
from 20 (CeO2) to 77% (Ce0.5Zr0.5O2) [29], depending on
the composition. The dependence of reduction progress of
the flame-made materials increased from 27 (CeO2) to 44%
degree of crystallinity of the flame-made mixed oxides is ev-
ident from Fig. 1 where lattice fringes are discernible in the
high-resolution TEM images. The isotherms in Fig. 5A re-
flect this morphology which results in a Type IV isotherm.
The low pore content observed stems from interparticle
necks. The resulting pore-size distribution in Fig. 5B de-
picts a significant contribution from mesopores 20–100 nm
in diameter. From the HRTEM (Fig. 1) these can be asso-
ciated with the voids between the crystallites. The rather
regular shape of the calcined powders results in a material
where well-structured crystalline surfaces (Fig. 1) do not fa-
cilitate sintering. The well-defined crystallites are of similar
size, few smaller crystallites are prone to sintering and lead
to a drop in specific surface area. Laine et al. [6,13] used
flame-spray pyrolysis to make a series of oxide powders and
(
[
Ce0.6Zr0.4O2). Similarly, the LSA ceria of Giordano et al.
26] was reduced by 20% at 860 C, whereas the flame-made
◦
pure ceria showed 47% reduction at the same temperature.
Consequently, the overall reduction behavior of flame-made
ceria/zirconia is similar to conventionally prepared catalysts.
In the case of pure ceria, however, the high-surface flame-
made oxide can be better reduced showing to some extent
the role of surface area in reduction.
4
.4. Dynamic oxygen exchange capacity
Dynamic oxygen storage capacity was measured using
H2, CO, and propene as reducing agents. The smooth de-

42
W.J. Stark et al. / Journal of Catalysis 220 (2003) 35–43
pendence of the OEC on the ceria content (Fig. 8) reflects
the homogeneity and intimate mixing of the flame-made ox-
ides as indicated by HRTEM (Fig. 1), specific surface area
rials is higher than that of the flame-made materials. Fig. 11
shows that this holds true up to about 1000 C. Then, a con-
◦
siderable degree of sintering starts to remove defects even in
◦
(
Fig. 2), and XRD (Fig. 3). The highest oxygen exchange
the case of stabilized ceria (Ce0.5Zr0.5O2). Above 1000 C,
activity using H2 or CO was found for Ce0.5Zr0.5O2, i.e.,
for intermediate compositions of ceria/zirconia. This obser-
vation correlates well with Boaro et al. [30] who inves-
tigated the CO-OEC of fresh ceria/zirconia (Rhodia) with
the two samples become comparable and the flame-made
material keeps increasing the OEC with temperature while
the profile of the precipitated material becomes flatter. Since
complete reduction would result in a mass loss of 1354 µg in
Fig. 11, it can be concluded that the trend of the Ce0.5Zr0.5O2
profile did not result from approaching full conversion (com-
plete reduction of ceria to Ce(III)).
2
2
9
2–108 m /g, aged samples with 41–68 m /g, and well-
2
aged samples with low specific surface area (10–18 m /g).
While the OEC strongly depended on the aging and spe-
cific surface area, the maximum OEC was found for high
and intermediate surface area samples with 68% ceria. The
low surface area samples showed a shifted maximum OEC
toward 70–80% ceria. Interestingly, our flame-made oxides
behaved as the freshly prepared samples of Boaro et al. [30],
5
. Conclusions
The production of ceria/zirconia nanocrystals by flame-
spray pyrolysis affords mixed oxides of high specific surface
area and improved thermal stability. Control of the ma-
jor process parameters: combustion energy delivery, metal
concentration, and carrier liquid (solvent) composition al-
low production of nanometer-sized ceria/zirconia crystals
over a broad range of conditions. Product homogeneity and
phase stability are directly related to a homogeneous deliv-
ery of both constituent precursors from the spray droplet
into the flame. Stable solid solutions were obtained for all
compositions of ceria/zirconia. Both in terms of reduction
behavior and dynamic oxygen exchange capacity, the flame-
made samples show similar properties as highly sintered
ceria/zirconia. High-resolution electron microscopy uncov-
ered, however, a striking different morphology. Whereas
precipitated ceria/zirconia consists of irregular, heavily ag-
gregated crystallites, the flame-made material is made up
of regularly shaped single crystals of few nanometer in
size. With respect to specific surface area, the flame-made
nanocrystals are much more stable toward temperature im-
pact than corresponding precipitated materials. This may
pave the way to ceria/zirconia and similar ceramic materi-
als with improved long-term stability for high-temperature
catalytic applications.
◦
even after severe sintering for 16 h at 700 C in air. While the
absolute values are difficult to compare since different tech-
niques were used, this qualitative comparison corroborates
the increased stability of the flame-made samples.
Using a stronger reductant, propene, a considerably
higher degree of ceria reduction was observed for all sam-
ples. At high ceria content, the effect was most pronounced
since only low conversions were achieved using H2 or CO.
At low ceria content, the reduction is favored to the point that
even H2 or CO afford higher conversions. Adding a stronger
reductant consequently could not further increase conversion
and similar degrees of reduction were observed.
A direct comparison of flame-made and precipitated sam-
ples in Figs. 8 and 11 reveals the differences in reducibility
for materials prepared by these two methods. These differ-
ences may be discussed in terms of defect content. The flame
method produces oxides with very low defect concentration.
This agrees with the reduction profiles where a LSA-type ce-
ria with high specific surface area was found. To obtain the
LSA-type ceria, Giordano et al. [26] applied a severe ther-
mal treatment to HSA ceria. This sintering heals out defects.
Consequently, the performance of flame-made materials re-
sembles that of a well-sintered material.
The higher performance of the precipitated mixed ox-
ides in Fig. 8 can now be explained with the aid of Fig. 11.
Pure ceria, or, in other words, unstabilized ceria prepared
Acknowledgments
We thank S. Veith for the pore-size analysis and Dr.
F. Krumeich for the high-resolution transmission elec-
tron microscopy investigations. Financial support by ETH
Gesuch Nr. 19/01-1 and the Swiss Commission for Tech-
nology and Innovation, Top Nano 21, Nr. 5978.2 is kindly
acknowledged.
◦
by precipitation considerably sinters when treated at 700 C
for 16 h in air. This treatment removes the favorable defects
found in precipitated materials and no marked difference in
◦
OEC is found in Fig. 11 from 300 to 1150 C. Using propene
(
Fig. 8), the performance of the flame-made material is even
higher since it could conserve more surface after sintering.
This observation is further supported by the reduction pro-
files where a higher degree of reduction was found for flame-
made ceria (Fig. 7) using propene. Adding zirconia presum-
References
[1] J. Kaspar, P. Fornasiero, M. Graziani, Catal. Today 50 (1999) 285.
[2] A. Trovarelli, Catal. Rev.-Sci. Eng. 38 (1996) 439.
[3] D. Terribile, A. Trovarelli, J. Llorca, C. Leitenburg, G. Dolcetti, J. Ca-
tal. 178 (1998) 299.
ably suppresses the defect healing during the pretreatment
◦
(
700 C/16 h/air) and defects are partially conserved. De-
2
spite the much higher specific surface area (100 m /g versus
[4] D. Terribile, A. Trovarelli, J. Llorca, C. Leitenburg, G. Dolcetti, Catal.
2
2
4 m /g after the pretreatment for flame-made compared to
Today 43 (1998) 79.
precipitated Ce0.5Zr0.5O2) the OEC of the precipitated mate-
[5] E.R. Novinshi, Eur. patent 0,086,330, 1983.

W.J. Stark et al. / Journal of Catalysis 220 (2003) 35–43
43
[
6] R.M. Laine, C.R. Bickmore, D.R. Treadwell, F. Waldner, US patent
[20] C. Bozo, F. Gaillard, N. Guilhaume, Appl. Catal. A 220 (2001)
69.
5958361, 1995.
[7] P.R. Strutt, B.H. Kear, R.F. Boland, WO patent 97/18341, 1995.
[8] R.M. Laine, C.R. Bickmore, K.F. Waldner, US patent 5614596, 1995.
[9] A. Gutsch, T. Henning, S. Katusic, M. Krämer, G. Michael, G.J. Varga,
Eur. patent 1142830 A1, 2000.
[21] M. Maciejewski, C.A. Müller, R. Tschan, W.D. Emmerich, A. Baiker,
Thermochim. Acta 295 (1997) 167.
[22] http://www.rhodia-ec.com/site-ec-us/catalysis/page-automotive.htm,
2002.
[
10] C.-H. Hung, J.D. Smith, G.P. Fotou, K.C. Koehlert, WO patent
1/36332 A1, 2000.
[23] E. Tani, M. Yoshimura, S. Somiya, J. Am. Ceram. Soc. 66 (1983)
506.
0
[11] T. Yoshioka, K. Dosaka, T. Sato, A. Okuwaki, S. Tanno, T. Miura,
J. Mater. Sci. Lett. 11 (1992) 51.
[24] L. Li, O. Van Der Biest, P.L. Wang, J. Vleugels, W.W. Chen, S.G.
Huang, J. Eur. Ceram. Soc. 21 (2001) 2903.
[
[
12] S.T. Aruna, K.C. Patil, NanoStruct. Mater. 10 (1998) 955.
13] R.M. Laine, T. Hinklin, G. Williams, S.C. Rand, J. Metastable Nano-
cryst. Mat. 8 (2000) 500.
[25] N. Hickey, P. Fornasiero, R. Di Monte, J. Kaspar, M. Graziani, G.
Dolcetti, Catal. Lett. 72 (2001) 45.
[26] F. Giordano, A. Trovarelli, C. Leitenburg, M. Giona, J. Catal. 193
(2000) 273.
[27] P. Fornasiero, T. Montini, M. Graziani, J. Kaspar, A.B. Hungria, A.
Martinez-Arias, J.C. Conesa, Phys. Chem. Chem. Phys. 4 (2002) 149.
[28] N. Hickey, P. Fornasiero, J. Kaspar, J.M. Gatica, S. Bernal, J. Ca-
tal. 200 (2001) 181.
[29] H. Vidal, J. Kaspar, M. Pijolat, G. Colon, S. Bernal, A. Cordon, V.
Perrichon, F. Fally, Appl. Catal. B 30 (2001) 75.
[30] M. Boaro, C. de Leitenburg, G. Dolcetti, A. Trovarelli, J. Catal. 193
(2000) 338.
[
14] R.M. Laine, R. Baranwal, T. Hinklin, D. Treadwell, A. Sutorik, C.
Bickmore, K. Waldner, S.S. Neo, Key. Eng. Mat. 159 (1999) 17.
15] A.C. Sutorik, M.S. Baliat, Mater. Sci. Forum 386–388 (2002) 371.
16] W.J. Stark, S.E. Pratsinis, Powder Technol. 126 (2002) 103.
17] L. Mädler, W.J. Stark, S.E. Pratsinis, J. Mater. Res. 17 (2002) 1356.
18] W.J. Stark, L. Mädler, M. Maciejewski, S.E. Pratsinis, A. Baiker,
Chem. Commun. 588 (2003).
[
[
[
[
[19] C. Leitenburg, A. Trovarelli, J. Llorca, F. Cavani, C. Bini, Appl. Catal.
A 139 (1996) 161.