Spatio-temporal dynamics of the NO + NH3 reaction on polycrystalline platinum

Article abstract of DOI:10.1016/j.susc.2006.11.005

Spontaneous reaction rate oscillations and spatio-temporal patterns have been observed by mass spectrometry and photoemission electron microscopy (PEEM) during the reduction of NO by NH₃ on polycrystalline platinum at 1 × 10^-4 Torr and temperatures from 460-520 K. The appearance of both oscillations and patterns was found to be strongly dependent on the gas phase composition and the temperature. In addition, the overall dynamics of the catalyst were found to be dominated by the nonlinear behavior of Pt(1 0 0) type grains, while other types of grains did not participate. In contrast to previous studies, a large number of complex multimodal oscillations were observed, particularly as the coupling between the surface and the gas phase was increased. The appearance of these complex oscillations demonstrates the importance of gas phase coupling to understanding catalytic reactions, even in high vacuum systems.

Full text of DOI:10.1016/j.susc.2006.11.005

Surface Science 601 (2007) 772–780
www.elsevier.com/locate/susc
Spatio-temporal dynamics of the NO + NH reaction
3
on polycrystalline platinum
a
a,b
a,
*
Noah McMillan , Christopher Snively , Jochen Lauterbach
a
Department of Chemical Engineering, Center for Catalytic Science and Technology, University of Delaware, Newark, DE 19716, United States
Department of Materials Science and Engineering, Center for Catalytic Science and Technology, University of Delaware, Newark, DE 19716, United States
b
Received 20 July 2006; accepted for publication 2 November 2006
Available online 27 November 2006
Abstract
Spontaneous reaction rate oscillations and spatio-temporal patterns have been observed by mass spectrometry and photoemission
ꢀ4
electron microscopy (PEEM) during the reduction of NO by NH on polycrystalline platinum at 1 · 10 Torr and temperatures from
3
4
60–520 K. The appearance of both oscillations and patterns was found to be strongly dependent on the gas phase composition and the
temperature. In addition, the overall dynamics of the catalyst were found to be dominated by the nonlinear behavior of Pt(100) type
grains, while other types of grains did not participate. In contrast to previous studies, a large number of complex multimodal oscillations
were observed, particularly as the coupling between the surface and the gas phase was increased. The appearance of these complex oscil-
lations demonstrates the importance of gas phase coupling to understanding catalytic reactions, even in high vacuum systems.
ꢀ
2006 Elsevier B.V. All rights reserved.
Keywords: Electron microscopy; Polycrystalline surfaces; Platinum; Surface chemical reaction
1
. Introduction
6NO + 4NH
3
! 5N
2
+ 6H
2
O
ð1Þ
The second pathway is the partial reduction of NO to N O,
a strong greenhouse gas:
2
Selective catalytic reduction (SCR) is widely used for the
reduction of NO in the presence of O from stationary
x
2
sources, particularly in power generating facilities and
nitric oxide plants [1]. Mobile SCR systems for transporta-
tion applications are being introduced in Europe in order
to meet more stringent air pollution control regulations,
however this technology is not yet available in the United
States [2,3]. Although mechanistically simpler than SCR,
8
NO + 2NH
3
! 5N
2
O + 3H
2
O
ð2Þ
When carried out over platinum catalysts, this reaction sys-
tem gives rise to nonlinear behavior, such as reaction rate
oscillations and spatio-temporal pattern formation. Partic-
ularly intriguing is the fact that, in contrast to most oscil-
lating reactions, nonlinear behavior is seen not only in
the catalyst activity, but also in the selectivity [5,6].
the reduction of NO by NH has been studied frequently
3
as a model reaction for SCR. It has been found to exhibit
considerable dynamic complexity and has been used to
gain insight into the industrial process.
Self-sustained reaction rate oscillations were first re-
ported for the reaction of NO and NH over a Pt wire at
3
The NO + NH reaction follows two predominant path-
ways [4]. The first pathway is the complete reduction of NO
to N2:
pressures between 0.2 and 1.0 Torr and temperatures be-
tween 473 and 1473 K [6]. Oscillations were also reported
at atmospheric pressures and temperatures between 603
and 673 K on a polycrystalline platinum foil [7]. Studies
of the same reaction over Pt(100) single crystals at ultra
high vacuum (UHV) conditions have shown similar behav-
3
*
Corresponding author. Tel.: +1 302 831 6327; fax: +1 302 831 3009.
E-mail address: lauterba@udel.edu (J. Lauterbach).
ꢀ
6
ior in which oscillations were observed near 1 · 10 Torr
0
039-6028/$ - see front matter ꢀ 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.susc.2006.11.005

N. McMillan et al. / Surface Science 601 (2007) 772–780
773
and temperatures from 420 to 485 K [5,8]. Spatio-temporal
patterns such as spirals and targets have also been observed
on Pt(100) single crystals by photoemission electron
microscopy (PEEM) at these conditions [9].
Although the mechanism accounting for this behavior is
not fully understood, the adsorbate-induced reconstruction
mechanism, previously used successfully to explain non-
linear phenomena during CO oxidation on Pt(110) and
Pt(100) [10,11], has also been proposed as a model for
pump. This differential pumping arrangement allowed the
PEEM to be operated with pressures in the main chamber
ꢀ
4
as high as ꢁ5 · 10 Torr. As has been described previ-
ously [17–19], PEEM allows the imaging of surfaces based
on local differences in work function. Photoelectrons are
excited from the sample surface by illumination with a
UV light source with peak intensity between 200 and
240 nm. The photoelectrons are focused through a series
of electrostatic lenses and projected onto a phosphorous
screen. The image is captured by a CCD camera and
recorded directly to a DVD recorder. The overall reaction
rate within the vacuum chamber was monitored by a
quadrupole mass spectrometer (Hiden HAL-201) that
was attached to the back of the PEEM and was also differ-
entially pumped.
NO reduction by NH on Pt(100) [12]. The Pt(100) surface
3
exhibits a surface reconstruction that can be lifted by NO
[
13,14], but not by NH [15]. At low coverages of NO,
3
the bulk-terminated 1 · 1 surface becomes unstable and
reconstructs to the hex phase. The overall production rate
of N is low on the hex reconstructed surface because the
2
dissociation of NO and NH is slow on the hex surface
The sample used for these experiments was a polished
polycrystalline platinum foil 1 cm on each side and
0.5 mm thick. The sample was mounted vertically in front
of the PEEM on a manipulator with an x–y–z translation
stage and a feedthrough, rotatable in the horizontal plane.
The sample was mounted by means of two 0.5 mm-thick
Ta wires spot-welded to the back of the sample, allowing
the sample to be resistively heated. A type-K thermocouple
was also spot-welded to the sample in order to monitor the
temperature. Before performing reaction experiments, the
sample was cleaned by repeated cycles of annealing at
3
compared to the 1 · 1 surface [15,16]. Accumulation of
NO on the hex surface lifts the reconstruction, and the
reaction rate on the 1 · 1 surface increases. If a sufficient
amount of NO is removed by reaction with NH , the
3
1
· 1 surface again becomes unstable and reverts to the
reconstructed surface, completing the cycle. Repetition of
this cycle accounts for the observed reaction rate oscilla-
tions in this system, and spatio-temporal patterns arise
due to coupling across the surface due to surface diffusion.
Because oxygen is always present in combustion ex-
haust, it can also be added to the reactant mixture in order
to approach more realistic conditions. In fact, the ability to
reduce NO in the presence of oxygen is a key feature of
SCR [1]. Again, this process occurs primarily by two reac-
tions, producing N and N O, respectively:
ꢀ
6
1100 K, oxidizing at 950 K under 1 · 10 Torr of O2,
and Ar-ion sputtering with 5 lA of sputtering current
ꢀ
4
and 1 · 10 Torr of Ar.
Reactant gases were metered by mass flow controllers
(Brooks 5850E) and mixed before being introduced into
the chamber through a leak valve. Pressure in the vacuum
chamber was measured by a Pirani gauge and was con-
trolled by manual adjustment of the leak valve. The gases
used were NO (Praxair, 99.8%), NH3 (BOC), and O2
2
2
4
NO þ 4NH þ O ! 4N þ 6H O
ð3Þ
ð4Þ
3
3
2
2
2
4
NO þ 4NH
þ 3O
2
! 4N
2
O þ 6H
2
O
This paper presents new examples of pattern formation
and complex oscillations for the reaction of NO and NH
(BOC, 99.9999%). The gas inlet system was pumped con-
3
tinuously by a mechanical pump, which maintained the
ꢀ
4
on a polycrystalline platinum foil at 1 · 10 Torr, with
and without O . Oscillations in the products N , H O,
ꢀ1
pressure near 1 · 10 Torr.
2
2
2
and N O were observed simultaneously with several types
2
3. Results and discussion
of spatio-temporal patterns. The relationship between pat-
tern formation and reaction rate oscillations is discussed.
Finally, unlike previous reports of oscillations on Pt(100)
single crystals, we report the appearance of multimode
oscillations in this system which suggests the importance
of gas phase coupling between different Pt(100) grains on
the polycrystalline surface.
Fig. 1 shows the reaction rate hysteresis curve for the
reaction of NO and NH on a polycrystalline platinum foil
3
ꢀ4
at 3 · 10 Torr and NO/NH = 0.5. This curve is qualita-
3
tively similar to results previously reported for this reaction
on Pt(100) [8]. The shape of the curve can be explained as
follows. At temperatures below 420 K, the surface is pre-
dominantly covered by NO and the reaction rate is low.
Increasing the temperature above 470 K causes a rapid in-
crease in the reaction rate due to the autocatalytic freeing
of surface sites upon the dissociation and reaction of NO
on the 1 · 1 surface. As the temperature is increased be-
yond 570 K, the reaction rate decreases due to thermal
desorption of surface species, and the surface reconstructs
to the hex phase due to the low coverage of NO. When
the temperature is lowered, the reaction rate increases
due to increasing coverage, however, the rate remains
2
. Experimental
All experiments were performed in a stainless steel UHV
ꢀ
9
chamber with a base pressure below 1 · 10 Torr. The
chamber was continuously pumped by a turbomolecular
pump backed by a mechanical pump and was equipped
with an ion gun for sample cleaning. The PEEM (Staib
Instruments) was mounted directly onto the chamber and
was differentially pumped by a second turbomolecular

774
N. McMillan et al. / Surface Science 601 (2007) 772–780
as indicated in Fig. 1. In the current study, both pattern
formation and reaction rate oscillations were observed over
a wide range of temperature (460–520 K) and gas phase
composition (NO/NH = 0.2–0.8).
3
Fig. 2 shows the parameter space over which spatio-
temporal patterns were observed at a total pressure of
ꢀ4
1
· 10 Torr. Several different types of self-sustained
patterns were observed in different areas of the pressure–
composition parameter space. These patterns were qualita-
tively classified as three types: island growth, turbulence,
and wavetrains. These three types occurred simultaneously
with global reaction rate oscillations and this could be seen
visually as a uniform darkening of the surface periodically
interrupted the pattern formation. This uniform darkening,
followed by the resumption of pattern formation was
observed for each of the pattern types.
An example of island growth is shown in Fig. 3a. The
darker, high work function areas indicate a high coverage
*
*
of NO and O , while the lighter, low work function areas
Fig. 1. Hysteresis curve for NO + NH
constant gas phase composition (NO/NH
3
on platinum foil, measured at
= 0.5) and constant pressure of
ꢂ
are predominantly covered by NH and the products of
ammonia decomposition [9]. Initially, several NH islands
nucleate on the dark NO-covered surface. The bright is-
3
3
ꢀ4
1
· 10 Torr. The rectangle indicates the range of temperatures where
3
oscillations and pattern formation are observed.
lands grow until the entire surface is covered by NH . This
3
lower than that on the heating branch. Between 520 and
transformation is quickly followed by a uniform darkening
of the entire surface, indicating the adsorption and dissoci-
ation of NO. The entire cycle takes about 30 s. At high
4
70 K, the reaction rate abruptly increases resulting in a
crossover with the heating branch. A similar crossover
was previously attributed to the hex to 1 · 1 phase transi-
tion on the Pt(100) single crystal surface, which is induced
by adsorption of NO to the hex surface [8]. Cooling below
temperatures (>490 K), a small number of NH islands
3
nucleate and these grow to sizes of 50–100 lm during the
oscillatory period. At lower temperatures, a much larger
4
70 K returns the surface to a low reactive state in which
number of NH islands nucleate, but they only grow to
3
the surface is covered by NO.
In nonlinear systems hysteresis in the reaction rate is of-
ten accompanied by oscillations and pattern formation
10–20 lm in diameter, before being extinguished by NO
adsorption. This type of island growth gives the surface a
speckled appearance.
[
20]. As in previous studies on Pt(100) single crystals [9],
A second type of pattern is turbulence, as shown in
Fig. 3b. This type consists of a series of alternating bright
and dark waves with no apparent long-range structure.
reaction rate oscillations and pattern formation were ob-
served on polycrystalline platinum in the hysteretic region,
ꢀ4
Fig. 2. Phase diagram for the NO + NH system on polycrystalline platinum at 1 · 10 Torr.
3

N. McMillan et al. / Surface Science 601 (2007) 772–780
775
Fig. 3. Several examples of pattern formation. Each image shows patterning on a single grain which is outlined by white lines. (a) Island growth; 512.2 K;
ꢀ
4
ꢀ4
ꢀ4
1
.6 · 10 Torr; NO/NH
3
= 0.6. (b) Turbulence; 461.3 K; 1.6 · 10 Torr; NO/NH
3
3
= 0.6. (c) Bistability; 436.7 K; 1.6 · 10 Torr; NO/NH = 0.6.
The turbulent patterns are interrupted periodically by a
brightening and then uniform darkening of the entire grain,
due to the adsorption of NO. After the surface becomes
completely NO-covered, the turbulent patterns appear
again. Within the turbulence, spatio-temporal structures
similar to target patterns and spirals can be observed, how-
ever these structures are usually unstable and short-lived.
In addition, the turbulent patterns give rise to wavetrains
at certain conditions. As in the case of the turbulent pat-
terns, the periods of wavetrain propagation are regularly
interrupted by a uniform brightening and darkening of
the entire grain.
ticipate in the observed nonlinear behavior on polycrystal-
line platinum. This is supported by comparing our
polycrystalline sample to the results of previous single crys-
tal studies that show there is a large (ꢁ1 eV) work function
difference between the NH -covered and NO-covered sur-
3
faces [9]. Although we cannot report absolute work func-
tions directly from PEEM images, the difference in
brightness between two surfaces is proportional to the dif-
ference in work function. In Fig. 4, only two of the grains
visible on the polycrystalline foil exhibit a large change in
brightness between NO- and NH -covered conditions,
3
and these are exactly the grains that participate in the non-
linear behavior in subsequent experiments. In addition,
brightness measurements on our polycrystalline sample
under CO- and O-covered conditions were compared to
literature values of the work functions on various single
crystal surfaces under these conditions. Again, the grains
that exhibit nonlinear behavior during the NO + NH3
reaction are consistent with the (100) orientation. The
conclusion that only the (100) grains participate in pattern
formation is further supported by the fact that neither
At temperatures below 460 K, the system exhibits bista-
bility, an example of which is shown in Fig. 3c. In the bista-
ble region, both the NO and NH covered surfaces are
3
observed, but neither self-sustained pattern formation nor
oscillations occur.
The pattern formation described above does not involve
every grain on the surface. Instead, only about 20–30% of
the grains participate. We propose that only the Pt(100)
grains, or possibly some higher index stepped grains, par-

776
N. McMillan et al. / Surface Science 601 (2007) 772–780
3
Fig. 4. PEEM images of an area of polycrystalline platinum under (a) NO- and (b) NH -covered conditions. The grains with the largest change in
brightness were determined to be Pt(100) grains. Image diameter = 500 lm.
pattern formation nor oscillations during the NO + NH3
reaction have been previously reported on platinum single
crystals other than Pt(100) [1].
oscillations were observed simultaneously in every case that
pattern formation was observed. Fig. 6 shows the relation-
ship between the two phenomena. Pattern formation is rep-
resented by the PEEM images and average brightness of
the patterned grain; the global reaction rate is shown by
the mass spectrometry signals of N and H O. The surface
Although only certain grains exhibit pattern formation,
the active grains are synchronized across the entire poly-
crystalline surface. Fig. 5 shows simultaneous average
brightness measurements for several grains exhibiting pat-
tern formation during reaction conditions. The brightness
profiles show synchronization of the grains in terms of their
overall period of oscillation which occurs due to coupling
through the gas phase. Despite the synchronization, the
different profiles are not identical. For example, Grains 2
and 3 lag Grain 1 by 2–3 s in the transition from the min-
imum brightness to maximum. These differences between
grains are attributed to slight differences in rates of adsorp-
tion or reaction due to differences in steps or defect concen-
trations from grain to grain.
2
2
of the oscillating Pt(100) grain is brightest when the reac-
tion rate is at a minimum. By the surface reconstruction
model, this state corresponds to an NH -covered surface
3
in the hex phase. In about 10 s, the surface darkens, indi-
cating the adsorption of NO and the lifting of the recon-
struction. This accompanies an increase in the reaction
rate due to the increased rate of NO and NH decomposi-
3
tion on the 1 · 1 surface. NH islands nucleate and grow on
3
the 1 · 1 surface which slowly reconstructs to the hex
phase. Within 25 s, the reaction rate again reaches a mini-
mum and the brightness of the grain is at a maximum, indi-
cating that the grain is once more predominantly NH3
covered and much of the grain has reconstructed to the
hex phase. This sequence is compatible with the adsor-
bate-induced reconstruction mechanism utilized in the
LFI model [12].
The changes in local work function observed by PEEM
correspond to changes in the global reaction rate as mea-
sured by mass spectrometry. In contrast to previous obser-
vations on Pt(100) at lower pressures [9], reaction rate
Many of the oscillations observed in these experiments,
particularly those in the area of the parameter space (indi-
cated in Fig. 2) where turbulent patterns are observed,
exhibited complex multi-peaked behavior. Similar oscilla-
tions have been reported previously, but the mechanism
for such oscillations has not been fully explained [5]. As
an example of this behavior, consider a series of experi-
ments where the sample is cooled from 494 to 472 K at con-
stant gas phase composition (NO/NH = 0.5) and constant
3
ꢀ
4
pressure (1.6 · 10 Torr). The nitrogen (mass = 28 U)
mass spectrometry signals are shown in Fig. 7. At 494 K,
the oscillation is single-peaked with a period of 28 s. This
asymmetric, single-peaked oscillation is typical of the oscil-
lations that are observed in the island growth region of the
parameter space. Cooling the sample to 492 K, the large
oscillation continues, now with a period of 30.5 s. In addi-
tion, a second peak can be easily distinguished with nearly
half the amplitude of the large oscillation and a period of
1
5.2 s. In comparison with Fig. 2, the appearance of the
Fig. 5. Average brightness of several simultaneously patterning-forming
grains; T = 492 K; P = 1.6 · 10 Torr, NO/NH = 0.63.
3
ꢀ4
higher frequency oscillation coincides with the appearance

N. McMillan et al. / Surface Science 601 (2007) 772–780
777
ꢀ4
Fig. 6. The correspondence between oscillations and spatio-temporal pattern formation. Data was collected at 1.6 · 10 Torr, NO/NH
The width of each image is 200 lm.
3
= 0.5 and 477 K.
three-peaked oscillation at 489 K, where the period of the
higher frequency oscillation is 10.2 s. Cooling further de-
stroys the complex oscillations as well as the highly-
ordered spatio-temporal patterns. At 477 K, the oscillation
is again single peaked with a long, sloping shoulder on the
trailing edge. At 472 K, the oscillation is still single-peaked,
similar in appearance to the oscillations observed at 494 K,
but now the amplitude of the oscillations is half that
observed at the higher temperature. The amplitude of the
oscillations continues to decrease as the sample is cooled
into the bistable region. This example demonstrates that
increasing amplitude and complexity of the reaction rate
oscillations accompany increasing complexity in the simul-
taneous spatio-temporal patterns.
Even more complex oscillations are observed when the
coupling between the surface and the gas phase is increased
by partially closing a gate valve in front of the main turbo-
pump, thereby reducing the pumping speed and increasing
the gas phase residence time. Based on the volume of the
reactor and the pumping speed of the main turbopump,
it is estimated that the residence time of the chamber with
the gate valve fully open is ꢁ0.1 s. Partially closing the gate
valve obviously increases the residence time, however it was
Fig. 7. Development of multipeak oscillations at low pressure. The curves
are mass spectrometry signals for nitrogen. Data was collected at 1.6 ·
ꢀ4
10
3
Torr, NO/NH = 0.5 and (a) 494 K, (b) 492 K, (c) 489 K, (d) 477 K,
(e) 472 K.
of more complex spatio-temporal patterns such as target
patterns and wavetrains. Further cooling gives rise to a

778
N. McMillan et al. / Surface Science 601 (2007) 772–780
not possible to quantitatively determine the magnitude of
this change. The total pressure inside the reactor can still
be adjusted freely by opening or closing the leak valve.
Under these conditions of greater coupling through the
(when N is a maximum) and a state where selectivity to to-
2
tal reduction is at a minimum (when N O is a maximum).
2
The oscillations in selectivity are qualitatively consistent
with the surface reconstruction model for the reaction
already discussed. Previous studies have suggested that
gas phase, the overall period of oscillations (t ) is signifi-
1
cantly longer than before, typically around 90 s. Multiple
higher frequency oscillations can occur within this period,
the formation of N O is favored by a shortage of vacancies
2
on the surface [1,21]. On the high reactive (1 · 1) surface,
with the time between small peaks (t ) varying from 6 to
dissociation of NO and the reaction to form N proceeds
2
2
1
6
4 s. As summarized in Table 1, oscillations containing 3,
, 7, 10, 11, and 14 high frequency peaks have been
easily. The fast desorption of N creates additional vacant
sites that further accelerate the dissociation of NO. When
the surface reconstructs to the hex phase, dissociation of
2
observed within a single period at various conditions of
temperature and pressure. In addition to the temperature
and pressure, two periods (t and t ), corresponding to
the periods of the lowest frequency and highest frequency
excursions, are reported for each oscillation.
A typical example of the complex oscillations is shown
in Fig. 8, along with the Fourier spectrum of the nitrogen
mass spectrometry signal. The largest peak in the Fourier
spectrum (b) corresponds to an oscillation with a period
of 6.6 s. Another peak in the Fourier spectrum (a) repre-
sents a period of 94.3 s, and corresponds exactly with the
frequency of the large amplitude excursions. The 13 subse-
quent peaks in the Fourier spectrum occur at integer mul-
NO no longer occurs and the reaction to form N O is
2
favored. It has been proposed previously that intentional
operation of certain reactions in an oscillatory regime
could be used to improve selectivity beyond that achievable
by steady state operation [22,23], however, we have found
no evidence of an overall improvement in selectivity during
1
2
oscillatory behavior in the NO + NH system.
3
The multi-peaked oscillations similar to those shown in
Fig. 8 have also been observed to evolve into chaotic oscil-
lations. The Fourier spectrum of the N signal during one
2
example of chaotic oscillations is shown in Fig. 9, where
several frequencies are distinguishable. The highest fre-
quency peaks appear in a band centered near 0.15 Hz, cor-
responding to a period near 6.9 s. The width of this band
reflects the chaotic character of the high frequency oscilla-
tions. Unlike this high frequency band, the lowest fre-
quency peak in the spectrum (0.013 Hz) is very sharp and
corresponds to an oscillation with a period of 75 s. The
sharpness of this peak and its two harmonics suggest that
the corresponding oscillation is not related to period dou-
blings of the higher frequency processes and instead arises
from some independent mechanism that has not been pre-
viously explained.
tiples of (a). Oscillations in the N O signal are also evident
in Fig. 8, which are exactly out of phase with the oscilla-
tions in N and H O, in agreement with previous reports
2
2
2
[
5,6]. In effect, therefore, the system oscillates between a
state where selectivity to total reduction is at a maximum
Table 1
Summary of observations of complex oscillations
Number of peaks
s
1
(s)
s
2
(s) T (K) NO/NH
3
P (Torr)
ꢀ
ꢀ
ꢀ
ꢀ
ꢀ
ꢀ
4
4
4
4
4
4
6
7
79.4
87.7
82.0
73.0
94.3
94.3
13.2
11.2
8.4
6.8
6.5
501.5 0.5
506.6 0.5
495.9 0.5
499.4 0.5
500.5 0.5
491.6 0.5
3.50 · 10
3.60 · 10
3.40 · 10
3.50 · 10
3.60 · 10
3.60 · 10
The effect of oxygen on the reaction dynamics provides
additional insight into the nature of the multimode oscilla-
tions. Initially, multimode oscillations are observed during
1
1
1
1
0
1
4
4
6.6
the reaction of NO and NH as shown in Fig. 10. The
3
ꢀ
4
Fig. 8. An example of a 14-peaked complex oscillation. T = 481 K, NO/NH
3
= 0.5, P = 3.6 · 10 Torr. The Fourier spectrum of the mass 28 signal
shows a peak corresponding to the low frequency, large amplitude oscillations (a) as well as 13 higher frequency peaks.

N. McMillan et al. / Surface Science 601 (2007) 772–780
779
that oxygen may facilitate hydrogen abstraction from ad-
sorbed NH [24].
3
While the surface reconstruction model does provide a
mechanism for simple, single-peaked oscillations, it cannot
easily account for the appearance of the multimodal oscil-
lations of the type described above. Multimodal oscilla-
tions can arise due to the simultaneous influence of more
than one oscillatory mechanism. For example, multimodal
oscillations observed during CO oxidation on Pt(110) were
attributed to the surface reconstruction mechanism cou-
pled with strong surface faceting [10]. Other than the
surface reconstruction mechanism, a vacancy model has
been proposed by Nieuwenhuys as a possible oscillatory
mechanism for NO + NH on Pt(100) [5]. In this model,
3
changes in the number of surface vacancies provide the
necessary feedback to produce oscillations. A shortage of
free surface sites is assumed to inhibit the dissociation of
*
Fig. 9. Fourier spectrum of the nitrogen signal during chaotic oscillations
NO . If sites become free due to desorption and NO disso-
ꢀ4
occurring at 500 K, NO/NH = 0.5, and 3.5 · 10 Torr.
3
ciation can occur, the succeeding steps in the reaction
mechanism establish an autocatalytic feedback loop in
which the reaction increases the number of free surface
sites. The greater number of free surface sites favors the
dissociation of NO, which, in turn further increases the
number of free surface sites. This autocatalytic mechanism
eventually depletes the surface of reactants, leading to a de-
crease in reaction rate that allows the surface to be replen-
ished. Coupling the surface reconstruction mechanism with
the vacancy mechanism could give rise to multimodal
oscillations, but such oscillations have not been observed
on Pt(100) single crystal samples. Nieuwenhuys did ob-
serve double and triple-peaked oscillations, but this only
occurred after the addition of oxygen to the system [5].
Another feasible explanation for multimode oscillations
is the coupling of more than one nonlinear oscillator [20].
This is a good description of a polycrystalline platinum
foil, because each oscillating (100) type grain can be con-
sidered a separate oscillator and each of these oscillators
is coupled only through the gas phase. Slight differences
in adsorption probabilities or surface reaction rates from
grain to grain could account for the multimodal behavior.
This could explain the fact that multimode oscillations are
observed on polycrystalline platinum, but not on platinum
single crystals. It is significant that the complexity of mul-
timode oscillations increases greatly as the pumping speed
of the system is reduced, therefore increasing the gas phase
coupling between various grains of the surface.
Fig. 10. The effect of oxygen on the system is demonstrated by a series of
experiments at 500 K and 3.5 · 10 Torr. Curve (a) is the initial six-
ꢀ4
peaked oscillation with NO/NH
phase such that NO/NH /O = 3/6/1. In (c), the O
mulitpeak oscillations return.
3
= 1/2. In (b), O
2
is added to the gas
is turned off and the
3
2
2
oscillations are six-peaked with periods of 82 s and 14 s.
These oscillations are accompanied by wavetrain patterns.
When 10% O is added to the gas phase, the oscillations in
This coupling through the gas phase occurs due to the
finite residence time of molecules in the chamber and the
exchange of molecules between the gas phase and the oscil-
lating surface. The oscillating surface reaction rate causes
temporal changes in the gas phase composition which in
turn affect the rates of adsorption of species onto the cata-
lyst surface. This feedback results in coupling between the
gas phase composition and the reaction rate oscillations of
each grain on the polycrystalline surface. The strength of
this coupling is a function of operating conditions, in
particular the residence time of gas phase molecules in
2
the NO + NH3 reaction continue, but the oscillations
change in character. Although the low frequency oscilla-
tions continue unchanged, the higher frequency oscillations
are almost completely destroyed. Removing oxygen from
the system causes the fine structure oscillations to reappear.
The fact that the addition of oxygen has a disproportionate
effect on the high frequency oscillations again suggests that
the low and high frequency oscillations arise from different
mechanisms. The specific role of oxygen in the reaction
mechanism remains unclear, however recent work suggests

780
N. McMillan et al. / Surface Science 601 (2007) 772–780
the vaccum chamber. Although the effect of the reaction
rate on the gas phase composition is relatively small at fast
pumping speeds, as the pumping speed is reduced and the
gas phase residence time increases, the surface reaction rate
has a larger effect on the gas phase composition and the
coupling through the gas phase is greater.
References
[1] N.M.H. Janssen, P.D. Cobden, B.E. Nieuwenhuys, J. Phys.: Con-
dens. Matter 9 (1997) 1889.
[
[
[
2] P.L.T. Gabrielsson, Top. Catal. 28 (2004) 177.
3] M. Koebel, M. Elsener, M. Kleemann, Catal. Today 59 (2000) 335.
4] R.M. Heck, R.J. Farrauto, S.T. Gulati, Catalytic Air Pollution
Control, John Wiley and Sons, New York, 2002.
[
5] M.F.H. van Tol, J. Siera, P.D. Cobden, B.E. Nieuwenhuys, Surf. Sci.
4
. Conclusion
2
74 (1992) 63.
[
[
[
6] C.G. Takoudis, L.D. Schmidt, J. Phys. Chem. 87 (1983) 964.
7] T. Katona, G.A. Somorjai, J. Phys. Chem. 96 (1992) 5465.
8] S.J. Lombardo, F. Esch, R. Imbihl, Surf. Sci. Lett. 271 (1992) L367.
The results presented have shown the appearance of
oscillations and, for the first time, spatio-temporal patterns
on polycrystalline platinum during the NO + NH reac-
[9] G. Veser, F. Esch, R. Imbihl, Catalysis Letters 13 (1992) 371.
10] K. Krischer, M. Eiswirth, G. Ertl, J. Chem. Phys. 96 (1992) 9161.
11] T. Lele, J. Lauterbach, D. Ramkrishna, AICHE J. 49 (2003) 2158.
12] S.J. Lombardo, T. Fink, R. Imbihl, J. Chem. Phys. 98 (1993) 5526.
13] T. Fink, J.-P. Dath, M.R. Bassett, R. Imbihl, G. Ertl, Surf. Sci. 245
3
[
[
[
[
tion. The nonlinear behavior on polycrystalline platinum
was found to be due primarily to the effect of (100) type
grains, which are synchronized across the sample surface
due to coupling through the gas phase. The appearance
of a wide variety of multimode oscillations was observed,
particularly with increased surface/gas phase coupling.
While the surface reconstruction model alone is sufficient
to explain simple oscillations in this reaction system, a sin-
gle mechanism cannot account for these complex oscilla-
tions. Our results indicate that coupling through the gas
phase of independently oscillating Pt(100) grains is the
most likely explanation for this complex behavior.
(
1991) 96.
[14] P. Gardner, M. Tushaus, R. Martin, A.M. Bradshaw, Surf. Sci. 240
1990) 112.
(
[
[
15] J.M. Bradley, A. Hopkinson, D.A. King, Surf. Sci. 371 (1997) 255.
16] D.Y. Zemlyanov, M.Y. Smirnov, V.V. Gorodetskii, J.H. Block,
Surf. Sci. 329 (1995) 61.
17] G. Haas, R.U. Franz, H.H. Rotermund, R.M. Tromp, G. Ertl,
Surf. Sci. 352 (1996) 1003.
18] H.H. Rotermund, W. Engel, S. Jakubith, A. von Oertzen, G. Ertl,
Ultramicroscopy 36 (1991) 164.
19] H.H. Rotermund, Surf. Sci. Rep. 29 (1997) 364.
20] I.R. Epstein, J.A. Pojman, An Introduction to Nonlinear Chemical
Dynamics, Oxford University Press, New York, 1998.
21] A.V. Walker, M. Gruyters, D.A. King, Surf. Sci. 384 (1997) L791.
22] E. Gulari, X. Zhou, C. Sze, Catal. Today 25 (1995) 145.
23] P. Silveston, R.R. Hudgins, A. Renken, Catal. Today 25 (1995) 91.
24] I.M. Irurzun, E.E. Mola, R. Imbihl, Chem. Phys. 323 (2006) 295.
[
[
[
[
Acknowledgement
[
[
[
[
Acknowledgment is made to the donors of the American
Chemical Society Petroleum Research Fund for support of
this research.