Oscillating Behavior in N2O Decomposition over Rh Supported on Zirconia-Based Catalysts: 2. Analysis of the Reaction Mechanism

Article abstract of DOI:10.1006/jcat.2000.2919

The mechanism responsible for the oscillating behavior in N₂O decomposition of rhodium particles supported on zirconia-based oxides in the presence of oxygen and water is studied using transient reactivity and thermogravimetric tests. The behavior over different oxide supports is also reported. The results demonstrate that (i) the atomic oxygen produced in the decomposition of N₂O remains chemisorbed on the Rh surface, leading to its progressive accumulation, (ii) the activity of the catalyst increases progressively with increasing coverage of Rh with oxygen, and (iii) above a threshold concentration of oxygen on the Rh surface, a rapid reconstruction of Rh particles occurs with release of the chemisorbed oxygen and the start of a new oscillation cycle, but only in the presence of water in the feed. It is also shown that the presence of water in the feed, inducing the desorption of chemisorbed oxygen from the Rh surface, stabilizes a more reduced working state of the catalyst surface.

Full text of DOI:10.1006/jcat.2000.2919

Journal of Catalysis 194, 130–139 (2000)
doi:10.1006/jcat.2000.2919, available online at http://www.idealibrary.com on
Oscillating Behavior in N O Decomposition over Rh Supported
2
on Zirconia-Based Catalysts
2. Analysis of the Reaction Mechanism
1
G. Centi, L. Dall’Olio, and S. Perathoner
Department of Industrial Chemistry and Materials Engineering, University of Messina, Messina, Italy
Received February 21, 2000; revised May 3, 2000; accepted May 4, 2000
Much less frequently observed is the case of oscillations
The mechanism responsible for the oscillating behavior in N2O in the absence of a combination of a reductant and an
decomposition of rhodium particles supported on zirconia-based oxidant and in the absence of possible thermally induced
oxides in the presence of oxygen and water is studied using transient
reactivity and thermogravimetric tests. The behavior over different
oxide supports is also reported. The results demonstrate that (i)
the atomic oxygen produced in the decomposition of N2O remains
chemisorbed on the Rh surface, leading to its progressive accumu-
tion state in the zeolite, with N2O having the double role of
lation, (ii) the activity of the catalyst increases progressively with
increasing coverage of Rh with oxygen, and (iii) above a threshold
concentration of oxygen on the Rh surface, a rapid reconstruction
of Rh particles occurs with release of the chemisorbed oxygen and
phenomena (8). The decomposition of N2O on Cu/ZSM-5
catalysts (11–16) has recently been reported. The phe-
nomenon was not fully understood and demonstrated. It
was attributed to a periodic change in the copper oxida-
reductant and oxidant (forming nitrate and N2 species, re-
spectively). The oscillations are determined by a different
reaction rate of N2O with the oxidized and reduced cop-
the start of a new oscillation cycle, but only in the presence of water per sites (15, 16) or by the presence of two different states
in the feed. It is also shown that the presence of water in the feed, of catalytic activity related to different coverages with ad-
inducing the desorption of chemisorbed oxygen from the Rh sur- species (atomic oxygen produced from N O dissociation
2
face, stabilizes a more reduced working state of the catalyst surface.
�
c 2000 Academic Press
Key Words: N2O; Rh; oscillating behavior; zirconia; water;
mechanism.
and nitrate) (13). In the first case, the mechanism can again
be related to the case of oscillations due to redox reac-
tions, but with the important difference that the reduction
is associated with the spontaneous reduction of the copper
sites, which is known occur at the reaction temperatures
and oxygen partial pressures of the experiments of N2O on
Cu/ZSM-5 catalysts (17). In the second hypothesis for the
reaction mechanism, the driving force for the oscillations
may instead be related to the phenomena of adsorbate-
induced changes in the surface reactivity.
In part 1 of this work (18) we reported that oscillations in
N2O decomposition can also be observed in a completely
different catalytic system compared to Cu/zeolite catalysts:
namely, on Rh supported on Zr-Nd mixed oxides (indicated
hereafter as Rh-ZrNdOx). A characteristic feature of the
Rh-ZrNdOx catalyst, in contrast to the case of Cu/ZSM-5,
is that the oscillations in the decomposition of N2O are ob-
served only when water is fed together with N2O and O2.
Other featuresofthe oscillations, such astheir characteristic
shape and dependence on the reaction conditions, indicate
that the mechanism responsible for the oscillations should
be different from, although with reasonable points of anal-
ogy to, those proposed for Cu/ZSM-5 catalysts. A better
understanding of the reaction mechanism responsible for
INTRODUCTION
Steady-state multiplicity and a stable oscillating change
between two active states have often been observed in het-
erogeneous catalysts and have been interpreted according
to various reaction mechanisms, such as an adsorption-
induced surface reconstruction (1), a subsurface reservoir
of a reactant (2), and a periodic formation of an oxide or
a carbon layer (3, 4). Such mechanisms responsible for the
observed stable oscillations in the catalytic activity require,
however, the presence of an oxidant and a reductant, such
as in the oxidation of CO and hydrocarbons (1–8) or in
the reduction of N2O by CO over supported Pt catalysts
(
9, 10).
1
To whom correspondence should be addressed at Dipartimento di
Chimica Industriale ed Ingegneria dei Materiali, Universit a` di Messina,
Salita Sperone 31, 98166 Messina, Italy. Fax: +39-090-391518. E-mail:
centi@unime.it.
130
0021-9517/00 $35.00
Copyright �c 2000 by Academic Press
All rights of reproduction in any form reserved.

MECHANISM OF OSCILLATION IN N2O DECOMPOSITION
131
the oscillations in N2O decomposition on Rh-ZrNdOx cata- in the sticking coefficient of N2O and thus to an increase
lysts should not only provide better insight into the surface in the rate constant of its decomposition, which contrasts
catalytic chemistry of N2O decomposition on this catalyst, with the parallel effect of a decrease in the number of free
but also offer keys for a better understanding (i) in gen- Rh sites for N2O decomposition that leads to a decrease in
eral of the surface phenomena responsible for steady-state the reaction rate. The combination of these two effects and
multiplicity and in situ dynamic modification of the surface the in situ reconstruction of Rh particles above a thresh-
reactivity and (ii) specifically of the oscillations in N2O de- old of chemisorbed oxygen may explain the characteristic
composition on Cu/ZSM-5 catalysts.
shape of the oscillations. Although this reaction mechanism
On the other hand, the direct low-temperature decom- is a reasonable interpretation of catalytic and kinetic data,
position of N2O is a reaction of relevant industrial inter- further data in support of the hypothesis are necessary.
est, as discussed in part 1 (18) and also recently reviewed
The objective of the work reported here was to ana-
(
19, 20). In fact, N2O is a powerful greenhouse gas and lyze the reaction mechanism in N2O decomposition on Rh-
its emissions should be reduced to counteract atmosphere ZrNdOx catalysts using thermogravimetric tests, tests of
warming (20). N2O is produced in relevant amounts, al- transient reactivity as a function of catalyst pretreatment,
though in low concentrations, in several combustion or and tests in which the evolution of oxygen from the catalyst
chemical processes. In particular, it is produced in the com- is monitored during in situ oscillating N2O decomposition,
bustion of industrial waste containing nitrogen and in the to further check the proposed mechanism.
chemical processes using nitric acid as oxidant (20), in ad-
dition to the case of adipic acid production in which, how-
ever, in contrast to the previous cases, N2O is quite high
EXPERIMENTAL
(
30% or above). While commercial catalysts exist for the
Preparation of Catalysts
latter case, Rh supported on zirconia-based catalysts are
new interesting catalysts when N2O concentration is low in
The preparation of ZrNdOx catalyst is reported in detail
the emissions, as for the former cited cases. In terms of total in part 1 (18). The zirconia–neodymia mixed oxide sup-
volume of N2O emitted, the emissions where N2O is present port, with formal composition Zr0.92Nd0.08ᮀ O
(where
x
4� x
in low concentrations (below 1% ) are more relevant than x = 0.04), was prepared using the sol-gel technique. Final
�
the case of adipic acid (20). It should also be mentioned calcination in air was carried out at 550 C for 4 h, giving a
2
that due to the rapidly growing number of applications of total surface area of 84 m /g using the BET method. The
wet oxidation processes of N-containing waste (for exam- rhodium was added by incipient wet impregnation using an
ple, wet oxidation of active sludges), which produce large aqueous solution of Rh(NO ) � H O in order to have a final
3
3
2
amounts of N2O, the market for low-temperature N2O de- Rh loading of 1 wt% . After impregnation, the samples were
�
composition catalysts is estimated to grow considerably in dried, calcined at 500 C, and then activated by a reducing
the near future (20).
treatment. This procedure of reductive activation involved
�
A further reason for interest in obtaining a better under- an initial treatment at 500 C in helium flow up to complete
standing of the oscillation behavior in the decomposition elimination of the adsorbed water and then a treatment in
�
of N2O on Rh-ZrNdOx is related to the possibility of ob- a flow of 50% H in He at 500 C for 2 h followed by cooling
2
�
taining information about the surface processes determin- to 200 C in the same type of flow, and finally mild reoxida-
ing the surface reactivity of Rh supported on zirconia-type tion in a flow of 0.05% N O + 6% O in helium for 3 h. The
2
2
catalysts, which represent an important class of catalysts in, Rh-ZrNdO catalyst was used for more than about 250 h in
x
for example, automotive applications (21). Therefore, there N O catalytic decomposition tests after which the oscilla-
2
are various reasons that stimulate interest in studying the tions in the catalytic decomposition of N O became regular.
2
mechanism of oscillations in the decomposition of N2O on For shorter times on stream, oscillations were observed, but
Rh-ZrNdOx.
were less regular.
In part 1 (18) the analysis of the characteristic features
Rh supported on zirconia–lantania and on zirconia–
of the oscillations and their dependence on reaction con- alumina, samples that will hereafter be referred to as Rh-
ditions suggested that water is not directly involved in the ZrLaO and Rh-ZrAlO , respectively, were prepared with
x
x
mechanism responsible for the oscillations, but is respon- the same method used for the Rh-ZrNdO catalyst. The
x
sible mainly for favoring the accumulation of the oxygen loading of Rh, the content of lantania or alumina, and
produced by the dissociation of N2O on the Rh surface the procedure of calcination and activation were also the
and for weakening metal–support interactions, making sur- same as for Rh-ZrNdO . Rh supported on alumina, here-
x
face in situ reconstruction of Rh particles possible when the after referred to as Rh-Al O , instead was prepared using
2
3
2
concentration of chemisorbed oxygen on the Rh surface is a commercial � -Al O (Rhone-Poulenc, RP531; 106 m /g),
2
3
above a certain limit. The accumulation of chemisorbed while the loading of Rh, the procedure of addition, and
atomic oxygen over the Rh surface leads to an increase the method of calcination and activation (including time of

1
32
CENTI, DALL’OLIO, AND PERATHONER
pretreatment) were the same as for the Rh-ZrNdOx
RESULTS
catalyst.
Influence of the Catalyst Composition on the Occurrence
of Oscillations
Thermogravimetric Tests
Thermogravimetric tests were made using a Perkin–
The role of catalyst composition in determining the oc-
Elmer TG-2C apparatus modified to allow tests with a con- currence of the oscillations in N2O decomposition was
tinuous stream passing through the head and a side stream, analyzed by studying the catalytic behavior of zirconia-
just above the furnace, to add the reactive components. A based samples prepared and activated as Rh-ZrNdOx cata-
standard inert Pt pan was used to load approximately 13 mg lysts, but containing a different lanthanide ion instead of
of sample (Rh-ZrNdOx) using the same pellet dimensions neodymium (Rh-ZrLaOx) or a different trivalent ion such
as those used for the catalytic tests (18).
as aluminum (Rh-ZrAlOx). The use of a zirconia-based
Tests were made under isothermal conditions, after treat- mixed oxide instead of pure zirconia makes it possible to
�
ment at 480 C in a pure helium flow to remove adsorbed improve the surface area and to shift the activity to lower
species. After cooling to a selected temperature in the He reaction temperatures, although due to the low content of
flow and stabilization of the weight, a feed containing N2O the second element (10 wt% ) the surface characteristics of
in He (0.6 L/h) and passed through a saturator contain- the mixed oxide remain similar to those of zirconia (22). In
ing distilled water was added from the side entrance of the addition, the behavior of Rh supported on commercial alu-
thermobalance to the main flow of pure He (3 L/h) passing mina, but prepared and activated as Rh-ZrNdOx, was also
through the head of the thermobalance, in order to have analyzed. The results are reported in Fig. 1 (Fig. 1a, Rh-
a feed concentration over the thermobalance pan of about ZrLaOx; Fig. 1b, Rh-ZrAlOx; Fig. 1c, Rh-Al2O3) for typical
0
.05% N2O + 1% H2O in helium. The saturator contain- reaction conditions in which Rh-ZrNdOx shows regular os-
ing water could be bypassed to check the behavior in the cillations in N2O decomposition (18).
absence of water in the feed.
The two catalysts based on Rh supported on zirconia-
type oxides show the presence of defined oscillations in
�
Catalytic Tests
the temperature range 300–400 C, although they are less
regular than those observed in the case of the Rh-ZrNdOx
catalyst. No oscillations were noted using the Rh supported
on alumina catalyst in a similar range of temperature and
activity (Fig. 1c), nor were they observed on a range of
other types of Rh-supported catalysts (23). This shows that
the oscillations are a specific characteristic of Rh supported
on zirconia-based oxides, while the role of the second ele-
ment in the zirconia-based mixed oxide (neodymium, for
example) is related mainly to a secondary change in the
textural/surface properties that determines the presence
of more or less regular oscillations in N2O decomposi-
tion.
Catalytic tests were carried out with the same apparatus
and using the same modalities as described in part 1 (18).
A quartz fixed-bed reactor loaded with 0.5 g of catalyst
(
particles with diameters in the range of 0.1–0.3 mm) and
equipped with an on-line mass-quadrupole system for the
continuous analysis of the feed and of reaction products
�
1
was used. The space velocity was 27,500 h
.
The same apparatus was used to check the transient
behavior of the catalyst as a function of the catalyst
�
pretreatment, made at 480 C for 20 min using different
feeds as described above. After the pretreatment, the feed
was switched to pure He flow and the reactor temperature
�
was cooled to 320 C. After stabilization of the reactor
temperature, the feed was switched to a mixture containing Oxygen Desorption from the Catalyst Surface
0
.05% N2O, 6% O2, and 2% H2O in He (6 L/h) or 0.05%
during Oscillations
N2O and 6% O2 in He (6 L/h) and the catalytic behavior
was monitored using the mass quadrupole apparatus.
Blank tests with an empty reactor were made to determine
the time of the system response, which was subtracted
from the time of the system response when the catalyst
The mechanism proposed in part 1 (18) to interpret cata-
lytic data in N2O decomposition on Rh-ZrNdOx was based
on a process of surface accumulation of atomic oxygen,
produced by N2O dissociation on the Rh surface and a
rapid release in one step of the chemisorbed oxygen when
its concentration was above a threshold limit. In order
to check this hypothesis, an experiment was carried out
in which the conversion of N2O and the oxygen and ni-
(
Rh-ZrNdOx; 0.4 g) was present.
Surface Area
The surface area was determined with the BET method trogen partial pressure in the feed (Fig. 2) were analyzed
N2 adsorption at 77 K) using a Carlo Erba Sorpty 1750 simultaneously.
(
instrument. The samples were pretreated by evacuation at
To have enough sensitivity in the detection of the oxy-
00 C. The surface areas of the sample before and after the gen released from the catalyst, this experiment was possi-
catalytic tests were found to be nearly equivalent, taking ble only when oxygen was absent from the feed to the re-
into consideration the experimental error. actor. As shown in part 1 (18), although the oxygen partial
�
3

MECHANISM OF OSCILLATION IN N2O DECOMPOSITION
133
pressure in the feed modifies the frequency and amplitude
The formation of N2 parallels the conversion of N2O,
of the oscillations, it does not alter the basic features of while the formation of O2 shows a different trend. Dur-
the phenomenon and thus the experiment without oxygen ing the first part of the cycle, when the conversion of N2O
in the feed is also representative of what happens in the and the formation of N2 rapidly rise, the oxygen does not
presence of oxygen in the feed.
desorb. Oxygen starts to desorb only later, when the conver-
sion of N2O approaches a constant behavior. Furthermore,
a clear peak in the desorption of oxygen is noted coincident
with the rapid drop in N2O conversion and N2 formation.
This can be better seen in the expansion of one cycle of the
oscillations (Fig. 2b).
It may be noted also that during the sequence of cycles an
increase in the background oxygen formation can be seen
(
Fig. 2a). This suggests that two different processes of oxy-
gen desorption occur. The first, responsible for the peak in
O2 coincident with the drop in activity in N2O decomposi-
tion, is due to the simultaneous release of oxygen and the
dynamic reconstruction of the catalyst. The second desorp-
tion of oxygen, responsible for O2 formation during the final
part of the cycle and for the increase in the background O2
formation, occurs via a different mechanism, tentatively via
surface migration from the metal to the support interface,
in agreement with previous observations (22).
The experiment reported in Fig. 2 shows two important
aspects related to the mechanism of N2O decomposition on
the Rh-ZrNdOx catalyst:
�
During the first step of increase in the conversion of
N2O, no desorption of the oxygen produced during the dis-
sociation occurs (Fig. 2b); at the final part of the cycle,
oxygen starts to desorb via a different mechanism involving
probably the migration from the metal to the support.
�
Simultaneously with the rapid drop in the activity in
N2O conversion, there is a peak in the desorption of oxygen
(
Fig. 2b).
This experiment confirms that (i) the oxygen produced
during N2O dissociation remains stuck to the catalyst sur-
face, and reasonably on the Rh surface, although indica-
tions are that it also tend to migrate from the metal to
the support, (ii) the progressive accumulation of oxygen
on the catalyst leads to an increase in the reaction rate, and
(
iii) the rapid decrease in activityisrelated to reconstruction
of the catalyst, which leads to release of the chemisorbed
oxygen.
Weight Changes during in Situ Experiments
The process of surface accumulation of the oxygen pro-
duced by N2O dissociation and one-step desorption with a
parallel change in the surface reactivity can also be mon-
itored in parallel thermobalance experiments. The results
for two reaction temperatures are summarized in Fig. 3.
Tests were made in the absence of oxygen in the feed, in or-
der to better analyze the weight change due to accumulation
on the catalyst of the oxygen produced by N2O dissociation,
but in the presence of water in the feed necessary for the
occurrence of the oscillations (18).
FIG. 1. Conversion of N2O as a function of time for different reaction
temperatures for Rh-ZrLaOx (a), Rh-ZrAlOx (b), and Rh-Al2O3 (c) cata-
lysts. Feed: 0.05% N2O, 2% H2O, 6% O2 in helium.

1
34
CENTI, DALL’OLIO, AND PERATHONER
FIG. 2. Conversion of N2O and oxygen or nitrogen normalized partial pressure (O2 or N2 partial pressure in quadrupole detector divided by the
total pressure in the quadrupole chamber; the value is proportional to the effective O2 or N2 partial pressure in the feed) as a function of time for
�
Rh-ZrNdOx. Feed: 0.05% N2O, 0% O2, 2% H2O, remainder He; reaction temperature was 341 C. In Fig. 2b an expansion of a single cycle is reported.
�
Although the oscillations appear to be more symmetri- at 360 C it is about 0.03 wt% , corresponding to a mean
cal with respect to the shape of oscillations in N2O decom- maximum change in the conversion of approximately 60
position activity, reasonably due to an attenuation of the and 10–15% , respectively (18). The frequency of cycles
�
3
time response in thermobalance experiments (the weight in the thermobalance experiments is about 5 � 10 Hz,
response was mediated to eliminate the effect of vibrations, which also agrees with that observed in the catalytic tests
due to the small weight change being monitored—on the or- (18).
�
5
der of 1 � 10 g), there is good agreement between change
The amount of Rh on the catalyst during these exper-
�
6
in weight and change in activity in terms of frequency of iments is approximately 1.3 � 10 mol. A weight change
the oscillations and the effect of the reaction temperature of 0.1 wt% corresponds to an accumulation of about
�
� 7
on the amplitude of the oscillations. At 340 C the mean 8.4 � 10 molofatomicoxygen. Thisshowsthat the process
maximum weight change is approximately 0.1 wt% , while of surface reconstruction with release of the stored atomic

MECHANISM OF OSCILLATION IN N2O DECOMPOSITION
135
FIG. 3. Thermogravimetric tests of the weight change of Rh-ZrNdOx
at different reaction temperatures. Feed: 0.05% N2O, 0% O2, 1% H2O,
remainder He.
�
FIG. 4. Thermogravimetric tests of the weight change at 340 C of Rh-
ZrNdOx after water vapor was added to the feed. Other conditions were
as described in the legend to Fig. 3.
oxygen produced by N2O dissociation occurs when the
amount of accumulated monoatomic oxygen corresponds
to approximately 60% of all Rh atoms.
Role of Catalyst Pretreatment
To obtain information on the influence of catalyst pre-
The determination of dispersion and specific surface area treatment on the occurrence and induction time of oscilla-
of Rh particles using conventional methods does not give tions in N O decomposition, tests were made in which the
2
�
directly usable data, because prereduction of the catalyst is catalyst was preheated at higher temperature (480 C) using
necessary and thus the method gives information on a dif- different feed compositions, cooling the reactor in an inert
�
ferent situation than that present during the process of N2O flow to a reaction temperature of 320 C, and sending, after
decomposition occurringin the presence ofgaseousO2 (18). stabilization of the temperature, a feed containing 0.05%
2
Values of approximately 40 m /g Rh, depending on the tem- N O, 6% O , and 2% H O in helium. The results of the
2
2
2
perature of prereduction, that indicate good dispersion of transient activity in N O decomposition after the introduc-
2
Rh on the zirconia–neodymia support were obtained. As- tion of the latter feed are shown in Fig. 5 for three different
suming thus that nearly 60% of the Rh atoms are available
on the surface or subsurface for storing the monoatomic
oxygen produced by N2O dissociation, thermobalance ex-
periments suggest that the process of surface reconstruction
with release of the stored oxygen occurs when the surface of
the Rh particles is nearly completely covered by the oxygen
produced in the dissociation process.
The role of the presence of water in the feed was also
analyzed in thermogravimetric experiments. The results at
�
3
40 C for the Rh-ZrNdOx catalyst are shown in Fig. 4. Af-
ter about 10 min of feed composed of only 0.05% N2O in
He, water vapor is added to the feed. A rapid decrease in
the weight is observed, after which oscillations in the weight
start to become visible. It is worthwhile to note that after
addition of water to the feed, a weight increase may be ex-
pected due to adsorption of water. The evident decrease in
weight thus indicates that desorption of chemisorbed or lat-
tice oxygen of the catalyst may instead occur. Although sur-
prising, this is consistent with previous spectroscopic stud-
ies on Cu-ZrO2 catalysts (24) that clearly showed that faster
reduction and reoxidation of the copper ion were possible
only on a hydroxylated surface.
FIG. 5. Effect of the pretreatment on transient reactivity of Rh-
�
ZrNdOx at 320 C in a feed of 0.05% N2O, 6% O2, and 2% H2O in He.
�
Pretreatment (480 C, 20 min): flow of (1) 6% O2 in He, (2) 6% O2 + 4%
H2O in He, and (3) pure helium. After the pretreatment, the reactor was
�
cooled to 320 C in a pure He flow.

1
36
CENTI, DALL’OLIO, AND PERATHONER
types of pretreatment at high temperature: (i) 20 min in
a feed of 6% O2 in helium, (ii) 20 min in a feed of 6%
O2 + 4% H2O in helium, and (iii) 20 min in a feed of pure
helium.
In all cases, after an induction time ranging from 1 to
4
min, oscillations in the decomposition of N2O start to be-
come evident, although the frequency and regularity of the
oscillations depend on the type of pretreatment. Further-
more, not only the induction time, but also the modality of
approaching the oscillating regime, depends on the type of
pretreatment. In both pretreatmentsin the absence ofwater
in the feed, both in the presence and in the absence of oxy-
gen, the initial activity is 100% and after an induction time
abruptly decreases to a lower value after which oscillations
become evident. The induction time is approximately three
times higher for the pretreatment in the presence of oxygen
FIG. 6. Effect of the pretreatment on transient reactivity of Rh-
�
than in pure helium, contrary to that expected if the higher ZrNdOx at 320 C in a feed of 0.05% N O and 6% O in He. Pretreatment
2
2
�
initial activity is related only to catalyst reduction during at 480 C (20 min) in a flow of pure He for all samples; the reactor was then
�
cooled to 320 C in a pure He flow and the catalyst was further treated for
the higher temperature catalyst pretreatment.
1
0 min in a flow of (1) pure He, (2) 2% H2O in He, and (3) 20% H2O in
He. Before admission of the reaction feed (0.05% N2O and 6% O2 in He),
water vapor was removed from the lines by sending a pure He feed for
In the case of the pretreatment in the presence of water
and oxygen in the feed, the initial activity is instead low and
progressively increases to a maximum value after approx- 1 min.
imately 4 min, after which there is a rapid decrease in the
N2O conversion and then oscillations become evident. This
result is further indication that the presence of water in the
�
In the case of the pretreatment at 320 C in the flow of
feed during pretreatment considerably influences the cata- water and helium, the initial catalyst activity is nearly zero
lytic reactivity and the working state of the catalyst surface. and only slowly increases to a constant conversion value
To obtain further information on the role of water dur- that coincides with that observed in the case of the pre-
ing pretreatment in changing catalyst reactivity, a differ- treatment with a helium-only flow. The transient change
ent type of experiment was carried out (Fig. 6). In these in catalyst activity is in agreement with that observed in
�
tests, the same pretreatment at high temperature (480 C) Fig. 5 for the higher temperature pretreatment in the pres-
in pure helium flow (20 min) was made, but after the reactor ence of water in the feed, although in this case the time to
�
was cooled to the reaction temperature (320 C), the cata- reach stationary activity is longer and the abrupt decrease
lyst was further pretreated (10 min) using a feed contain- in conversion after which the oscillating activity starts is
ing a different concentration of water in helium. After the absent.
water vapor was removed from the line, a feed containing
.05% N2O and 6% O2 in helium (i.e., without water vapor) in the case of pretreatment in the presence of water in the
The initial low-level activity and slow increase in activity
0
was introduced and the transient catalytic activity in N2O feed cannot be attributed simply to a competitive effect of
decomposition was monitored. Due to the absence of water chemisorbed water, because (i) as the concentration of wa-
�
in the feed, oscillations in N2O decomposition were not ob- ter in the feed in the pretreatment at 320 C (Fig. 6) increases
served in this case, but the data obtained give information from 2 to 20% , the induction time to reach the stationary
on the effect of water in changing the initial state of the activity becomes shorter, and (ii) the same type of trend is
catalyst toward N2O decomposition.
also observed in the experiments of Fig. 5 (for the higher
In the case of pretreatment in helium only, the initial ac- temperature pretreatment in the presence of water in the
tivity is 100% and only slowly declines to a conversion of feed), for which water is present in the reaction feed dur-
N2O of about 70% . In contrast to the case of Fig. 5, when ing N2O decomposition. The role of water must therefore
water was present in the reaction feed, the conversion is be related to a change in the working state of the catalyst
initially 100% for about 8 min (against 1 min in the case of during the catalytic reaction.
Fig. 5) and then takes around 25 min to reach a stationary
value (against a few seconds in the case of Fig. 5 and further-
more for a more drastic decrease in the conversion). These
data clearly show that the presence of water in the feed,
DISCUSSION
In part 1 (18) the following working model to explain
together with N2O and O2, determines a complete change the oscillations in N2O decomposition on the Rh-ZrNdOx
in the dynamic reactivity characteristics of the catalyst. catalyst was suggested:

MECHANISM OF OSCILLATION IN N2O DECOMPOSITION
137
(
i) N2O decomposes over the free sites of the Rh surface, desorb. The shape of the oxygen curve does not parallel that
but the atomic oxygen produced in decomposition remains of N2 (Fig. 2), indicating that although both are formed via
chemisorbed on the Rh surface, leading to its progressive N2O dissociation, the mechanisms of desorption of nitrogen
accumulation.
ii) The sticking coefficient for N2O increases progres- oxygen is also noted in the series of cycles (Fig. 2), also
sively with the increase in the coverage of Rh with oxygen indicating a process of oxygen desorption that occurs on a
thus leading to an increase in the rate constant), but at the time scale different from the process of dissociation of N2O.
and oxygen are different. An increase in the background
(
(
same time the number of available free sites on the Rh sur- As shown elsewhere (18, 22), the metal–support interface
face decreases, leading to a decrease in the reaction rate; and especially the presence of oxygen vacancies at the in-
the cumulative effect of these two contrasting phenomena terface play a relevant role in determining the reaction rate
explains the characteristic shape of the oscillations during in Rh on zirconia-based catalysts. The phenomenon was
phase 1 (increase in conversion).
interpreted on the basis of the hypothesis that the atomic
(
iii) Above a threshold concentration of oxygen on the oxygen produced from N2O dissociation migrates from the
Rh surface, rapid reconstruction of Rh particles occurs with Rh surface and accumulates at the perimeter area between
release of the chemisorbed oxygen and the start of a new Rh particles and the support, from which they then des-
oscillation cycle.
orb. The same mechanism is reasonably responsible for the
desorption of oxygen in the final part of the cycle and for
The data obtained in the present study give further evi- the increase in the background oxygen formation. Data in
dence in support of this hypothesized reaction mechanism Figs. 2 and 3 thus provide good support for the first step in
responsible for the oscillations in N2O decomposition on the proposed reaction mechanism (dissociation of N2O at
the Rh-ZrNdOx catalyst.
the Rh surface and progressive accumulation on it of the
The first question in the proposed mechanism regards atomic oxygen produced from this reaction), although the
the hypothesis that the atomic oxygen produced by N2O results in Fig. 2 indicate that part of the chemisorbed oxy-
decomposition remains chemisorbed on the rhodium sur- gen may desorb via a mechanism that probably involves the
face and progressively accumulates. The data in Fig. 2 show surface migration to the support as the first step.
that there is no desorption of O2 in the first part of the os-
The second question in the proposed reaction mechanism
cillations, when the conversion of N2O increases, although regards the hypothesis that the sticking coefficient for N2O
they show that in part oxygen may desorb via a different increases progressively with an increase in coverage of the
mechanism, probably involving surface migration from the Rh with oxygen, leading to a progressive increase in the
metal to the support and then its desorption. The tests conversion of N2O. The data in Fig. 2 regarding the rising
reported in Fig. 2 correspond to tests with a feed mix- part of the oscillations demonstrate this hypothesis well.
ture containing N2O and H2O in helium, but no gaseous Data on the transient catalyst activity after higher temper-
oxygen, except for traces due to impurities. However, the ature pretreatment (Fig. 5) further support this hypothesis.
observation of the absence of desorption of O2 from the After treatment at higher temperature with a flow contain-
catalyst during the rising part of the oscillations may be ing oxygen, and thus probably after oxidation of the surface
extended to the case of the presence of O2 in the feed, be- of the Rh particles occurs, the catalyst activity is 100% and
cause as noted in part 1 (18), oxygen affects the frequency remains complete for a time on stream approximately three
�
and amplitude ofthe oscillations, but not their characteristic times higher (after cooling to 320 C and starting of the feed
shape.
containing N2O, O2, and H2O) than when the same type of
The second aspect of the first question regards the con- pretreatment is made in a flow of pure helium (thus more
cept that oxygen remains lodged on the Rh surface. As reducing than a flow of oxygen in helium).
shown elsewhere (23), Rh largely promotes the activity of
The data in Fig. 5 further show that when the pretreat-
zirconia-based oxides in N2O conversion and thus it is rea- ment is made in the presence of water in the feed, the initial
sonable to indicate that the reaction of N2O decomposition activity is very low. This is further demonstrated by the data
occurs on the Rh surface and that the oxygen produced reported in Fig. 6 that show that the effect cannot be at-
in this reaction remains lodged on the Rh surface. Thermo- tributed to an inhibition effect on N2O chemisorption due
gravimetric data (Fig. 3) are consistent with this hypothesis, to chemisorbed water. Comparison of the transient change
indicating that the weight change during oscillations is con- in catalyst activity in the case of the data of Figs. 5 and 6,
sistent with the creation of an oxygen layer on the Rh par- where the difference in feed composition is related to the
ticles. Data in Fig. 2, although demonstrating that oxygen presence (Fig. 5) and the absence (Fig. 6) of water in the
does not desorb during the first part of the cycle (when there reaction feed, also shows that after an induction time there
isa rapid rise in N2O conversion;see in Fig. 2b the expansion is a considerable decrease in the conversion of N2O after
of the first part of the cycle starting from time 480 s), show which the oscillations start to become evident, when water
that for longer times during the cycle itself oxygen starts to is present in the feed.

1
38
CENTI, DALL’OLIO, AND PERATHONER
It may be concluded that both the absence of oxygen times (more than about 200 h) were necessary to detect
desorption during the rising part of the oscillations (Fig. 2) regular oscillations, when the neodymia component was
and the effect of the high-temperature pretreatment (Fig. 5) present. The characterization of the samples before and af-
demonstrate that the catalyst activity is higher over an “ox- ter this in situ activation procedure, however, did not allow
idized” Rh surface than over a “reduced” Rh surface (i.e., elucidation of the reasons for this induction period, which
after desorption of the chemisorbed oxygen). The data do thus remain an open question.
not provide information on whether there is an effective
Loong et al. (25, 26) reported that the surface acidity
change in the sticking coefficient for N2O or instead a pro- characteristics in lanthanide-modified zirconia mixed ox-
gressive change in the intrinsic reactivity of the Rh surface ides, and in particular the strength of the hydroxyl groups,
in passing from a reduced to an oxidized state. Reason- depend on the lanthanide ion. We also observed (22) that
ably, both concepts are valid, because both N2O adsorp- the surface acidity characteristics in zirconia–alumina can
tion and dissociation can be faster on a Rh surface near a be progressively changed by increasing the relative content
chemisorbed oxygen atom.
of alumina in the mixed oxide. It is thus reasonable to hy-
The third question in the proposed mechanism is related pothesize that the main effect of the second element in the
to the concept that above a certain concentration of oxygen zirconia-based oxide support is that of modifying the sur-
on the Rh surface, rapid reconstruction of Rh particles oc- face acidity characteristics and in particular the strength of
curs with release of the chemisorbed oxygen and the start of the hydroxyl groups. In the presence of a hydroxylated en-
a new oscillation cycle. The data reported in Fig. 2 show that vironment of the zirconia surface the Rh particles show a
parallel to the abrupt decrease in conversion, there is a des- much weaker interaction with the oxide support and thus
orption of oxygen from the catalyst surface. It may thus be can give rise to in situ reconstruction phenomena not possi-
concluded that in the presence of water in the feed catalyst ble when a stronger interaction with the support is present,
reconstruction that leads to the desorption of chemisorbed such as on a dehydroxylated zirconia surface or on other
oxygen may occur. The reduced catalyst (i.e., after desorp- types of oxide supports. This explains the effect of water
tion of oxygen) shows a much lower activity in N2O de- on the dynamic features of the Rh-ZrNdOx catalyst and
composition that the oxidized catalyst. This explains why the role of the second element in the zirconia-based mixed
pretreatment of the catalyst in the presence of water in the oxide support in determining the presence of regular os-
feed leads to a nearly zero initial activity in contrast to the cillations in N2O decomposition. However, more detailed
case of pretreatment in the presence of oxygen or even a studies using advanced characterization techniques, such as
helium-only flow (traces of oxygen as contaminant and due scanning tunneling microscopy, would be necessary to bet-
to leakage were present in the helium). The thermogravi- ter understand the reasons for this phenomenon.
metric data in Fig. 4 further support the evidence that in the
The effect of water on the working state of Rh particles
presence of water in the feed a “reduction” of the catalyst on zirconia-based catalysts is not necessarily related to the
occurs with a decrease in its weight, instead of an increase as reaction of N2O decomposition, but is a general feature of
more reasonably expected due to chemisorption of water. these catalysts. Rh on zirconia catalysts are interesting cata-
The presence of water in the feed, therefore, not only lysts in various catalytic reactions (21) and are a key compo-
gives rise to competitive chemisorption with N2O, as usu- nent in commercial three-way catalysts for the treatment of
ally indicated for other catalysts active in N2O decomposi- exhaust gas emissions from vehicles (27). The present data
tion (20), but remarkably changes the redox characteristics on the role of water and N2O in determining the working
of the catalysts, as well as the dynamic processes during state of Rh particles offer a key with which to analyze the
the catalytic reaction, as exemplified by comparison of the data of this type of catalyst also in other kinds of reactions,
different transient behavior shown in Figs. 5 and 6.
as well as an opportunity to change or tune their reactivity.
The present data, however, cannot clarify the details of
the mechanism responsible for the effect of water on the dy-
namicbehavior ofRh particlesand whytheir reconstruction
with desorption of chemisorbed oxygen occurs. The data in
Fig. 1 clarify that these phenomena are a general feature
of Rh supported on the zirconia-based catalyst, although
not of other types of oxide supports. The role of the second
component in the zirconia-type mixed oxide, besides that of
a structural promoter to stabilize thermal resistance against
sintering (22), is that of modifying the surface properties,
thus making the oscillations more regular and detectable.
It should be mentioned that although oscillations were
also observed after short times on stream, relatively long
REFERENCES
1
2
3
. Ertl, G., Adv. Catal. 37, 213 (1990).
. Basset, M. R., and Imbihl, R., J. Chem. Phys. 93, 811 (1990).
. Yeates, R. C., Turner, J. E., Gellman, A. J., and Somorjai, G. A., Surf.
Sci. 149, 175 (1985).
4. Collins, N. A., Sundaresan, S., and Chabal, Y., Surf. Sci. 180, 136
1987).
. Kobayashi, M., Kanno, T., Takeda, H., and Fujisaki, S., Appl. Catal. A
Gen. 151, 207 (1997).
. Tenchev, K. K., Petrov, L. A., Savelieva, G. A., and Sass, A. S., Appl.
Catal. A Gen. 83, 31 (1992).
(
5
6
7. Schr o¨ der, U., J. Catal. 146, 586 (1994).

MECHANISM OF OSCILLATION IN N2O DECOMPOSITION
139
8
. Sch u¨ th, F., Henry, B. E., and Schmidt, L. D., Adv. Catal. 39, 51
1995).
. Sadhankar, R. R., and Lynch, D. T., J. Catal. 149, 278 (1994).
20. Centi, G., Perathoner, S., and Vazzana, F., CHEMTECH 29(12), 48
(1999).
(
9
21. Burch, R., and Loader, P. K., Appl. Catal. A Gen. 143, 317 (1996).
22. Centi, G., Marella, M., Meregalli, L., Perathoner, S., Tomaselli, M.,
and La Torretta, T., in “Advanced Catalysts and Nanostructured Ma-
terials” (W. R. Moser Ed.), Chap. 4, p. 63. Academic Press, San Diego,
1996.
1
1
1
1
1
0. Sadhankar, R. R., Ye, J., and Lynch, D. T., J. Catal. 146, 511 (1994).
1. Lintz, H. G., and Turek, T., Catal. Lett. 30, 313 (1995).
2. Turek, T., Appl. Catal. B Env. 9, 201 (1996).
3. Turek, T., J. Catal. 174, 98 (1998).
4. Ciambelli, P., Garufi, E., Pirone, R., Russo, G., and Santagata, F., Appl.
Catal. B Env. 8, 333 (1996).
23. Centi, G., Dall’Olio, L., and Perathoner, S., Appl. Catal. A Gen. 194–
195, 79 (2000).
1
5. Ciambelli, P., Di Benedetto, A., Garufi, E., Pirone, R., and Russo, G.,
J. Catal. 175, 161 (1998).
24. Morterra, C., Giamello, E., Cerrato, G., Centi, G., and Perathoner, S.,
J. Catal. 179, 111 (1998).
1
6. Ciambelli, P., Di Benedetto, A., Pirone, R., and Russo, G., Chem. Eng.
Sci. 54, 2555 (1999).
25. Loong, C. K., Thiyagarajan, P., Richardson, J. W., Ozawa, M., and
Suzuki, S., J. Catal. 171, 498 (1997).
17. Centi, G., and Perathoner, S., Appl. Catal. A 132, 179 (1995).
18. Centi, G., dall’Olio, L., and Perathoner, S., J. Catal. 192, 224 (2000).
19. Kapteijn, F., Rodriguez-Morasol, J., and Moulijn, J. A., Appl. Catal. B
Env. 9, 25 (1996).
26. Loong, C. K., Richardson, J. W., and Ozawa, M., J. Catal. 157, 636
(1995).
27. Fornasiero, P., Di Monte, R., Rao, G., Kaspar, J., Meriani, S., Trovarelli,
A., and Graziani, M., J. Catal. 151, 168 (1989).