A vibrational spectroscopic investigation of the Co+O2 reaction on Pt{110}

Article abstract of DOI:10.1063/1.1483069

Carbon monoxide and oxygen coverage of a platinum surface under steady state reaction conditions and during kinetic oscillations and pattern formation was studied using real-time infrared reflection-absorption spectroscopy, together with photoelectron emission experiments. Results indicated that CO was evenly distributed over the surface at all coverages with and without coadsorbed atomic oxygen. Changes on the surface were directly related to carbon dioxide production rate under both steady state and oscillatory reaction conditions.

Full text of DOI:10.1063/1.1483069

A vibrational spectroscopic investigation of the CO+O2 reaction on Pt{110}
J. H. Miners, S. Cerasari, V. Efstathiou, M. Kim, and D. P. Woodruff
Citation: J. Chem. Phys. 117, 885 (2002); doi: 10.1063/1.1483069
View online: http://dx.doi.org/10.1063/1.1483069
View Table of Contents: http://jcp.aip.org/resource/1/JCPSA6/v117/i2
Published by the AIP Publishing LLC.
Additional information on J. Chem. Phys.
Journal Homepage: http://jcp.aip.org/
Journal Information: http://jcp.aip.org/about/about_the_journal
Top downloads: http://jcp.aip.org/features/most_downloaded
Information for Authors: http://jcp.aip.org/authors
Downloaded 19 Aug 2013 to 128.118.88.48. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

JOURNAL OF CHEMICAL PHYSICS
VOLUME 117, NUMBER 2
8 JULY 2002
A vibrational spectroscopic investigation of the CO¿O2 reaction
on Ptˆ110‰
J. H. Miners, S. Cerasari, V. Efstathiou, M. Kim, and D. P. Woodruffa)
Fritz–Haber–Institut der Max–Plank–Gesellschaft, Faradayweg 4-6 D-14195 Berlin, Germany
͑Received 24 January 2002; accepted 12 April 2002͒
The CO coverage of a Pt͕110͖ surface in both the high and low reaction rate branches of the bistable
CO oxidation reaction has been determined by Infrared Reflection-Absorption Spectroscopy
͑IRAS͒, first performing extensive calibration experiments on the various factors determining the
absorbance and frequency associated with the C–O vibrational stretching mode. The same two
states of the surface are shown to be present under steady-state low and high reaction rates and when
the surface is undergoing pattern formation and homogeneous reaction rate oscillations. Using the
CO coverages determined by IRAS, the intensities observed in a series of photoelectron emission
microscopy images have been used to elucidate the oxygen coverage in both coadsorption states.
The low reaction rate branch is found to be associated with a high CO coverage (0.5Ϯ0.1 ML) and
very low O coverage (0.03Ϯ0.01 ML) consistent with the (1ϫ1) unreconstructed phase. In the
high rate branch the surface has a low CO coverage (0.05Ϯ0.03 ML) and O coverages in the range
0.3–0.7 ML ͓(1ϫ2) reconstructed phase͔. No evidence for bridged CO, oxide, or subsurface
oxygen, variously proposed to play a role in the reaction rate bistability, was found under the
conditions measured. These findings are consistent with the site blocking and reconstruction model.
Coadsorption experiments of CO and oxygen under nonreactive conditions, performed as part of the
IRAS calibration process, demonstrate that CO and O can occupy a mixed adlayer and identify two
different chemical environments for CO adsorption. © 2002 American Institute of Physics.
͓DOI: 10.1063/1.1483069͔
I. INTRODUCTION
switching between two reaction branches with different reac-
tion rates.
The COϩO2 reaction on Pt͕110͖ ͑e.g., Refs. 1, 2, and 3͒
is, after the Belousov–Zhabotinsky reaction, one of the most
investigated examples of a chemical system which exhibits
kinetic oscillations ͑e.g., Refs. 4 and 5͒. It is considered to be
an ideal system to study as the reaction is comparatively
simple and the reactants can only occupy well-defined posi-
tions in a two-dimensional array defined by the crystal sur-
face, thus lending itself to Monte Carlo modeling. Moreover,
one has a very good control of the external parameters, being
able to maintain and vary accurately the partial pressure of
the reactants and the pumping rate, and to sustain a homo-
geneous temperature over the entire sample.
The essential mechanism of the CO oxidation reaction
on metal surfaces is one in which oxygen adsorbs dissocia-
tively and the resulting chemisorbed atomic oxygen reacts
with adsorbed CO to form CO2 which desorbs from the sur-
face. In the case of Pt͕110͖, the clean surface reconstructs to
a (1ϫ2) missing row structure, but this reconstruction is
lifted at sufficiently high coverage of CO, but not by atomic
oxygen adsorption alone. It is this phase transformation, and
associated differences in the activity of the two surface
phases, which is believed to be a key factor in the oscillatory
CO oxidation rate which can occur on this surface, involving
A large range of nonspatially resolving real-time tech-
niques have been applied to the study of the reaction on this
surface
including
Kelvin
probe
work
function
measurement,6,7 mass spectrometry,7 low energy electron dif-
fraction ͑LEED͒,7,8 and reflection high energy electron dif-
fraction ͑RHEED͒.9 More recent measurements with the spa-
tially resolving techniques of photoelectron emission
microscopy ͑PEEM͒,10–12 low energy electron microscopy
͑LEEM͒,13,14 and mirror electron microscopy ͑MEM͒ have
shown spatial variations characterized by two distinct states
of the surface, believed to correspond to those giving rise to
the low and high reaction rates; however, these studies
showed that pattern formation may occur even when macro-
scopic oscillations in the reaction rate are observed in the gas
phase. An important consequence of this result is that the
information previously gained from the nonspatially resolv-
ing techniques concerning oscillatory changes in the state of
the surface does not necessarily correspond to the two states
observed on a local scale.
While these spatially resolving techniques have greatly
increased our understanding of the phenomenology of the
spatiotemporal variations involved in this surface reaction,
they do not provide any direct measure of the local surface
coverage of the reactants. In particular, the relationship be-
tween the intensities used to provide the image contrast and
the state of the surface is complex. In the case of PEEM,
intensity changes are largely related to changes in the work
a͒
Author to whom correspondence should be addressed. Also at: Department
of Physics, University of Warwick, Coventry CV4 7AL, United Kingdom.
Electronic mail: d.p.woodruff@warwick.ac.uk
0021-9606/2002/117(2)/885/12/$19.00
885
© 2002 American Institute of Physics
Downloaded 19 Aug 2013 to 128.118.88.48. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

886
J. Chem. Phys., Vol. 117, No. 2, 8 July 2002
Miners et al.
function, but this depends in a complex fashion on the local
CO and O coverage and the number of vacant sites. In
LEEM the image intensity relates to the intensity of a par-
ticular LEED diffracted beam at a specific energy which is
influenced not only by whether the surface shows (1ϫ1) or
(1ϫ2) periodicity, but also by the degree of local long-range
order including the density of surface defects and by the
presence of adsorbates. Nevertheless, by measuring LEED
intensity-energy spectra under different conditions of surface
phase and adsorbate coverage, Meißen et al.15,16 were able to
identify those energies of the incident electron beam for
which the LEEM intensity was sensitive primarily to only
one of these parameters, e.g., the presence of the (1ϫ1)
phase. By recording LEEM images of pattern formation at
these voltages, they could show that a chemical wave con-
sisted of a reconstructed part of the surface, and a (1ϫ1)
component which was CO covered.16 Under the conditions
used, they found no evidence of faceting. More recent MEM
and LEEM measurements of global oscillations ͑in which the
whole surface transforms simultaneously and homoge-
neously͒ have indicated that the rate of CO2 production is
lowest when the surface transforms to the (1ϫ1) phase.15
Despite these achievements, however, there is still only indi-
rect information concerning the nature and concentration of
the chemical species on the surface during the oscillation,
and their relation to the rate of reaction. Many models have
been proposed to explain the behavior in this system, but in
the absence of this basic information all of these models are
necessarily somewhat speculative.
Among the processes considered to be responsible for
the bistability and hysteresis of this reaction system are the
lifting of the surface reconstruction,17 CO site blocking of
oxygen dissociation,18,19 changes of the CO bonding coordi-
nation site, defect formation and the faceting of the surface,8
subsurface oxygen formation,20,12 oxide formation,21 and car-
bon formation.22 It should be noted that the latter two pro-
cesses do not occur on single crystal surfaces at pressures
below 10Ϫ4 mbar.22 The influence of reconstruction and site
blocking are currently thought to be the key mechanisms
determining the behavior at low temperatures and pressures.
A clear resolution of the mechanisms creating the novel
properties of this reaction requires experimental determina-
tion of quite a large number of variables. Specifically, one
needs to know the correlations between ͑i͒ the state of the
surface order, notably whether it is (1ϫ1) or (1ϫ2), but
also whether it is faceted or heavily defected; ͑ii͒ the CO
coverage; ͑iii͒ the CO site occupation; ͑iv͒ the oxygen cov-
erage, and ͑v͒ the rate of reaction.
possible to use the PEEM image intensity to deduce the oxy-
gen coverage. By combining the results from these measure-
ments and from a parallel LEEM/LEED investigation con-
ducted in this laboratory,15,16 and relating them via the rate of
CO2 production, we can obtain a full description of this sys-
tem.
II. EXPERIMENTAL DETAILS AND METHODOLOGY
The instrument used in these experiments has been de-
scribed previously.23 Briefly, it consists of a commercial
FTIR spectrometer ͑Biorad FTS-60A/896͒ modified for op-
eration under low vacuum (ϳ1ϫ10Ϫ3 mbar) interfaced to a
UHV chamber equipped with rear-view LEED optics, a mov-
able quadrupole mass spectrometer and facilities for argon
ion sputtering. Spectra were recorded using a liquid nitrogen-
cooled narrow-band MCT detector. The same Pt͕110͖ crystal
and preparation techniques were used as in the earlier LEEM
experiments.15,16 The sample was mounted on a liquid
nitrogen-cooled cold finger and could be resistively heated
such that temperatures between 85 K and 1500 K could be
maintained to a stability of 0.05 K. The temperature was
measured using a K-type thermocouple spot-welded directly
to the top of the crystal and the signal from this thermo-
couple was fed back into the computer-controlled tempera-
ture control unit. The sample gases used in the experiments
are of the highest commercially available purity: 99.997%
for CO, ͑Linde͒, 99.999% for O2 ͑Messer–Griesheim͒. The
partial pressures quoted in each experiment have been cor-
rected for the different ion gauge sensitivities of the respec-
tive gases, i.e., SCO /SN ϭ1.0 and SO /SN2ϭ0.8 were used
2
2
as correction factors. The UHV chamber routinely obtains a
base pressure of 1ϫ10Ϫ10 mbar. The crystal was cleaned
prior to each experiment via oxygen treatment at 800 K fol-
lowed by repeated cycles of argon ion bombardment at 750
K and annealing at 1100 K.
The infrared absorption band of an adsorbed species can
be described in terms of four parameters ͑i͒ the center fre-
quency, ͑ii͒ the integrated absorbance, ͑iii͒ the full width at
half maximum ͑FWHM͒, and ͑iv͒ the peak shape. Each of
these components can provide information on the coverage
and local environment of the adsorbate under investigation
as follows:
͑i͒
Changes in the center frequency of the vibrational
mode of an adsorbed species can occur as a result of
four distinct effects; these are changes in the adsorp-
tion site, variations in the chemical shift, the extent of
the dynamic dipole–dipole coupling, and the degree
of dephasing.24 The chemical shift ͑a change in the
vibrational frequency due to a change in the electronic
interaction of the molecule with its environment͒ may
vary as a result of changes in the local adsorbate en-
vironment ͑including coverage and substrate order͒.
The extent of dynamic dipole–dipole coupling is de-
termined by the local coverage of adsorbate molecules
with similar vibrational frequencies.25,26 These com-
ponents can be separated by performing appropriate
calibration experiments. Specifically, the frequency
shift due to dipole–dipole coupling can be identified
In this paper we present the results of a real-time infra-
red reflection–absorption spectroscopy ͑IRAS͒ investigation
which we show can be used to obtain the local CO coverage
both under steady state reaction conditions and during ki-
netic oscillations and pattern formation. Concurrent mass
spectrometry measurements allow us to relate the CO cover-
age to the rate of CO2 production. The IRAS data are also
sensitive to the bonding configuration of the CO and the
presence of surface oxygen. We have also reproduced these
experiments in a PEEM instrument, and armed with the
knowledge of the CO coverage from the IRAS study, it is
Downloaded 19 Aug 2013 to 128.118.88.48. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

887
J. Chem. Phys., Vol. 117, No. 2, 8 July 2002
COϩO2 reaction on Pt͕110͖
by performing isotopic dilution measurements, while
the influence of the dephasing component can be es-
tablished through its temperature dependence.
͑ii͒
The integrated absorbance of an infrared band pro-
vides information on the coverage of the associated
species, but this may not increase linearly with
coverage24 and the absorption cross-section changes
with
molecular
orientation
and
adsorbate
coordination.27 Proper interpretation of measurements
of the absorbance is therefore also dependent on care-
ful calibration experiments.
͑iii͒ The width of an infrared absorption peak can vary
with temperature due to dephasing, and may also con-
tain a component arising from inhomogeneity in the
adlayer; disorder results in a variety of frequencies of
absorbers and hence a peak broadening.
͑iv͒ The relative importance of these two contributions to
the broadening, ‘‘lifetime’’ and inhomogeneity, may
be determined by the peak shape; if lifetime broaden-
ing is dominant one expects a Lorenzian shape,
whereas inhomogeneities lead to a Gaussian line
shape.28
In order to exploit as much as possible of this more
detailed information from the IRAS measurements of the
C–O stretching vibrations from a Pt͕110͖ surface under CO
oxidation reaction conditions, a series of calibration experi-
ments involving CO and oxygen adsorption were first carried
out under static conditions. The results of these measure-
ments are reported here in Sec. III.
FIG. 1. Graph of the main quantitative features of the C–O stretching band
seen in IRAS as a function of CO coverage on Pt͕110͖ at sample tempera-
tures of 200 K ͑light gray͒ and 300 K ͑black squares͒.
While IRAS can provide us with considerable quantita-
tive information on the local and average CO coverage on
the surface during the reaction and provide clear evidence for
the presence and absence of coadsorbed O in the two states,
they do not allow us to quantify the oxygen coverage.
PEEM, on the other hand, is much more sensitive to the
surface concentration of oxygen and thus provides comple-
mentary information, so additional experiments were per-
formed using this technique. The PEEM experimental setup
used here has been described in detail elsewhere,29–33 and
the same sample cleaning procedures were applied to the
crystal as in the IRAS experiments. The intensities seen in a
PEEM image are primarily determined by the variations in
work function which in turn depend on the coverage of CO
and oxygen; these relationships have been previously
calibrated.30–33 If the CO concentration is known indepen-
dently from IRAS, its contribution to the PEEM intensity can
be calculated and the measured PEEM intensity can then be
used to determine the oxygen coverage.
the (1ϫ1) phase occurs and the reconstruction is lifted.36,37
The value of the C–O stretching frequency in vibrational
spectroscopy indicates that CO adsorbs in an atop configu-
ration, and a saturation coverage of 1 ML can be obtained by
slowly cooling a clean Pt͕110͖,38,39 from 700 to 300 K, in an
ambient pressure of 1ϫ10Ϫ7 mbar CO.35,40
Hayden et al. have used IRAS and temperature-
programmed desorption ͑TPD͒ to measure the coverage de-
pendence of the C–O stretching frequency on Pt͕110͖ at a
surface temperature of 300 K.38 By diluting the 12CO in
13CO, they were able to show that all coverage-dependent
changes in the frequency of the mode were due to dipole–
dipole coupling, and that there is no coverage-dependent
chemical shift on this surface. Furthermore, since the surface
changes phase from (1ϫ2) to (1ϫ1) during the absorption
sequence, there is evidently no chemical shift associated with
this surface phase transition. These experiments also showed
that the absorption cross section of the IRAS band associated
with the C–O stretching frequency decreases after a cover-
age of 0.5 ML has been reached. They assigned this to the
tilting of the CO molecule at higher coverages, a conclusion
confirmed by angle-resolved ultraviolet photoelectron spec-
III. IRAS CALIBRATION STUDIES
A. CO adsorption on Ptˆ110‰
The clean Pt͕110͖ surface is reconstructed into the
(1ϫ2) missing-row phase.34 CO adsorption at temperatures
below 250 K occurs without lifting the reconstruction.35
At temperatures greater than 250 K, however, LEED and
STM studies have shown that, on reaching an average cov-
erage of 0.2 ML, CO-induced homogeneous nucleation of
troscopy ͑ARUPS͒,41 XPS and XPD,42 and reproduced by
43
¨
extended Huckel molecular orbital calculations.
Figure 1 shows the results of our first calibration experi-
ment in which we measured the coverage dependence of the
C–O internal stretching vibration at sample temperatures of
200 and 300 K. These results are clearly consistent with
Downloaded 19 Aug 2013 to 128.118.88.48. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

888
J. Chem. Phys., Vol. 117, No. 2, 8 July 2002
Miners et al.
those of Hayden et al., albeit at higher spectral resolution.
The coverages were determined by TPD, using the saturation
1 ML CO phase formed as described above as our calibration
point. At 200 K the CO-induced unreconstruction of the
(1ϫ2) phase is suppressed and the surface retains the
(1ϫ2) ordering, whereas at 300 K the (1ϫ2)→(1ϫ1)
transition occurs with increasing CO coverage, yet there are
only small differences between the two plots, confirming the
earlier finding of the experiments conducted only at 300 K
that the surface structural phase has little effect on the ab-
sorption cross section of the IRAS band. Notice that the de-
tailed measurements at the lower temperature shown here
only extend to 0.35 ML, but this is already much higher than
the coverage needed at room temperature to induce the struc-
tural phase transition ͑0.2 ML͒. Additional measurements at
1 ML coverage at both temperatures confirmed the essential
equivalence of the IRAS spectra at the two temperatures over
the full coverage range. As is shown below, the slightly
higher frequency and smaller FWHM observed at 200 K, is
attributable to less dephasing; at 200 K there is less ther-
mally induced occupation of low-frequency ͑frustrated trans-
lation͒ modes that are coupled to the C–O stretching mode.26
One issue in the interpretation of these IRAS data is the
role of CO island formation. At 200 K on the (1ϫ2) surface
the observed linear dependence of the frequency on coverage
is consistent with the effect of dynamic dipole–dipole cou-
pling in the absence of islanding.25,26,44,45 At 300 K the sur-
face transforms to the (1ϫ1) phase as the average CO cov-
erage increases above 0.2 ML. If this process were to
proceed with a significantly higher CO coverage on the
(1ϫ1) islands than on the surrounding (1ϫ2) surface one
would observe two distinct IRAS peaks at intermediate av-
erage coverages for a sample temperature of 300 K. This is
not observed, and indeed no significant difference is ob-
served in the coverage dependence of the frequency between
the measurements taken at sample temperatures of 200 and
300 K, for coverages between 0.2 and 0.5 ML. This is con-
sistent with the observation by STM that at 300 K the phase
transition occurs by homogeneous, rather than heterogeneous
nucleation, and that increasing CO coverage leads to an in-
crease in the density, rather than the size, of the (1ϫ1)
islands,37 but also shows that the local and average CO con-
centrations all over the surface are essentially the same. On
this basis we will assume that the ordinate of Fig. 1 does
represent local coverage and, combined with measured fre-
quency shift due to dipole–dipole coupling, we will therefore
use the C–O stretching frequency to determine the local cov-
erage under reaction conditions. Knowing the cross section
for adsorption at this local coverage, one can also use the
integrated absorbance to determine the total coverage.
indicating that the width is attributable to homogeneous
rather than inhomogeneous broadening.28 Finally, we should
note that our data show no evidence of any bridging CO that
would be expected to give rise to a significantly lower C–O
stretching frequency in the 1800–1900 cmϪ1 range. This is
in agreement with recent XPS and XPD measurements in
which bridging CO is only observed at the high coverage of
␪
COϭ1.09 ML which can only be obtained by cooling down
the crystal from 600 K to TϽ240 K in ambient CO.42
This initial calibration experiment provides characteriza-
tion of the IRAS for CO on Pt͕110͖ at sample temperatures
of 200 K and 300 K, but we wish to study CO on this surface
at the significantly higher temperatures at which CO oxida-
tion reaction oscillations occur. At these higher temperatures
we must account for the changes in the spectra due to
dephasing; i.e., the variations in the C–O stretching fre-
quency due to coupling to lower frequency frustrated trans-
lational vibrational modes, which become occupied to a sig-
nificant degree at higher temperatures.48–51 Two approaches
to this problem are possible. The first is to measure the tem-
perature dependence of the IRAS data at constant coverage
͑and thus at constant dipole–dipole coupling͒; in this experi-
ment one also obtains information on the temperature depen-
dence of the FWHM. A limitation to this method is that
thermal desorption of CO becomes significant at tempera-
tures above 350 K. The alternative method is to eliminate
dipole–dipole coupling by diluting the 12CO in 13CO. Since
it is already established that there is no coverage-dependant
chemical shift in this system, changes in coverage in the
absence of dynamic dipole coupling will have no effect on
the frequency, and all temperature-dependent variation will
then be due to dephasing. This method can be applied at
higher temperatures ͑and lower associated coverages͒.
Figure 2 summarizes the results of both of these experi-
ments. In one experiment we prepared a 1 ML coverage of
CO and varied the temperature in an ambient pressure of 2
ϫ10Ϫ7 mbar CO. The temperature was raised from 150 K to
500 K, in steps of 25 K, and spectra comprising 256 scans at
a resolution of 2 cmϪ1 were recorded. The maximum tem-
perature at which the saturated 1 ML CO coverage could
be maintained under this pressure of CO was 350 K. This
was deduced by subsequently raising the CO pressure to 1
ϫ10Ϫ5 mbar and noting that there was a change in the IRAS
spectra for TϾ350 K. The results of these experiments
showed no change in integrated absorbance with increasing
temperature. Increasing the temperature from 150 to 350 K
led to a decrease in the center frequency of the C–O stretch-
ing band by 7Ϯ2 cmϪ1. The form of the observed changes
in both the peak width and center frequency are as expected
due to dephasing and the peak shift is of a similar magnitude
to that seen on Pt͕111͖.24,51,52 Since there is very little change
in the FWHM of around 5 cmϪ1 when the CO coverage,
⌰co, increases beyond a value of 0.2 ML ͑see Fig. 1 and
Refs. 37, 46, 47, and 27͒, we should note that for a surface
exposed to 2ϫ10Ϫ5 mbar CO at 500 K the IRAS peak
The higher resolution of our measurements,27,38,46,47 and
the use of absorbance log(I /I) as opposed to transmission
͓
͔
0
(I0 /I) in the presentation of the spectra, allows small varia-
tions in spectral line shape to be observed and interpreted in
a more meaningful fashion. The slight increase in the
FWHM ͑full width at half maximum͒ with increasing cover-
age at very low coverage is as expected from dipole–dipole
coupling.26 Inspection of the original IRAS spectra also
shows that the absorption peak has a Lorenzian line shape,
showed a Lorenzian form with a FWHM of 7.8 cmϪ1
.
It is interesting to note the 4 cmϪ1 difference shown in
Fig. 2 between the center frequency values obtained from the
isotopic dilution experiments ͑full line͒ and the extrapolated
Downloaded 19 Aug 2013 to 128.118.88.48. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

889
J. Chem. Phys., Vol. 117, No. 2, 8 July 2002
COϩO2 reaction on Pt͕110͖
phase changes, or electronic CO–CO interactions, our objec-
tive is to apply IRAS to a study of the CO/O interaction
under reaction conditions, so we must also characterize any
possible chemical shift in the C–O vibrational frequency due
to the presence of coadsorbed oxygen. Indeed, it is known
from other systems that oxygen has a strong, repulsive inter-
action with CO, resulting in island formation and a chemical
shift of the C–O stretching frequency.52–55
Oxygen adsorption on Pt͕110͖ is dissociative at tempera-
tures above 175 K.56 It requires, therefore, two vacant ad-
sorption sites to adsorb. In contrast to CO, oxygen adsorption
does not result in the lifting of the (1ϫ2) reconstruction.40
Furthermore, although CO has a sticking coefficient of 1.0
on both surface phases,35 the sticking coefficient of O is
estimated to be 0.4 and 0.6 on the (1ϫ2) and the (1ϫ1)
phase, respectively.57 There are no conventional quantitative
structure determinations for the local adsorption geometry of
either CO or O on Pt͑110͒, and most of the models in the
literature are largely speculative. In the case of oxygen ad-
sorption on Pt 110 (1ϫ2) such speculation favors the three-
͕
͖
fold coordinated hollow sites on the ͑111͒-type walls of the
troughs, although at low coverage occupation of the four-
fold coordinated hollow sites located at the bottom of the
͗110͘ troughs has been proposed.56–58 A recent investigation
using a combination of STM and density functional theory
͑DFT͒ calculations, on the other hand, strongly favors the
͑111͒ nanofacet threefold coordinated hollow sites over these
trough sites.59 In the case of CO adsorption an early sugges-
tion was that these trough hollow sites were occupied by CO
at low coverage, but that bridge sites on the top of the rows
may be occupied at saturation coverage.35 However, the fact
that the C–O stretching frequency is always above
2000 cmϪ1 has been generally taken as a more reliable indi-
cation that local atop sites are always occupied by this spe-
cies; the fact that the saturation coverage of 1 ML is the same
on the (1ϫ2) and (1ϫ1) surfaces ͑at low and higher tem-
peratures, respectively͒ is also thought to imply that on the
(1ϫ2) phase CO must occupy local atop sites in the troughs
as well as on the ridges.27
FIG. 2. The temperature dependence of IRAS of 1 ML CO coverage on
Pt͕111͖. The two spectra at the top and the bold lines all relate to pure 12CO
layers. The light line corresponds to data collected with a mixture 1:10
12CO:13CO. The singleton ͑extrapolated zero coverage͒ frequencies at 200
and 300 K ͑crosses͒ are also depicted.
zero-coverage ͑singleton͒ values at 200 and 300 K, shown as
the two crosses. This arises because, for a total coverage of 1
ML and a 1:10 isotopic ratio 12CO:13CO, the local coverage
of 12CO is 0.1 ML, sufficient to give rise to a small but
significant dipole–dipole coupling shift. As may be seen in
Fig. 1, this coverage does lead to a shift due to dipole–dipole
coupling of precisely this value of 4 cmϪ1. Notice also, in
Fig. 2, that the pronounced fall in the center frequency at
higher temperatures (Ͼ350 K) seen in the results from the
isotopic dilution experiment is also largely attributable to this
same effect, the total CO coverage ͑and thus also the 12CO
coverage͒ falling at high temperatures.
Using the results of this experiment, we are able to ac-
count for the temperature induced change in the singleton
frequency, and thus relate spectra of CO on Pt͕110͖ recorded
at any temperature between 150 and 525 K to the calibration
measurement at 300 K, with an accuracy of Ϯ2 cmϪ1. From
the dipole–dipole coupling shift one can then determine the
local CO coverage.
In the absence of adsorbed oxygen, the adsorption of CO
on a clean Pt 110 (1ϫ2) surface at low coverage is homog-
͕
͖
enous as discussed above. In the presence of oxygen, how-
ever, not only could interaction of the coadsorbed species
lead to a change in the local site occupied by the CO, but one
may now have island formation with significant implications
for the macroscopic reaction of the two species.60 A separa-
tion of CO and O into distinct islands could also strongly
influence the ability of the two species to react, and perhaps
explain the origin of poisoning effects observed under reac-
tion conditions.19 In order to interpret the IRAS measure-
ments under reaction conditions it is therefore important to
first study the effects of coadsorption of oxygen and CO
under nonreaction conditions.
The results of the first of these coadsorption calibration
experiments are shown in Fig. 3; in the lower part of the
figure IRAS data are shown from a surface predosed with 0.2
L CO at 300 K, and then exposed to increasing doses of
molecular oxygen. The presence of the coadsorbed O follow-
ing a 2 L dose leads to an upward frequency shift and to
B. CO coadsorption with oxygen under nonreactive
conditions
While the calibration experiments on pure CO adsorp-
tion show no chemical shifts associated with the substrate
Downloaded 19 Aug 2013 to 128.118.88.48. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

890
J. Chem. Phys., Vol. 117, No. 2, 8 July 2002
Miners et al.
FIG. 3. Two sets of IRAS spectra obtained from Pt͕110͖ following the
coadsorption of CO and oxygen at 300 K. The lower group of spectra were
obtained by first adsorbing CO and then oxygen; this exposure sequence was
reversed to obtain the upper group of spectra.
FIG. 4. Two sets of IRAS spectra obtained from Pt͕110͖ following the
coadsorption of CO and oxygen at 300 K using an isotopic mixture of 1:10
12CO:13CO. Otherwise similar to Fig. 3.
around 1960 cmϪ1 associated with a component of 12C18O in
the gas mixture, which was nominally only of 12C16O and
13C16O; however, this is well removed from the spectral re-
gion of interest and the behavior of this species actually re-
inforces the primary results described below. We have al-
ready noted that, using a 1:10 isotopic ratio of 12CO:13CO,
the frequency shift due to dipole–dipole coupling in the peak
associated with the 12CO species can be no more that 4 cmϪ1
͑its maximum value at saturation coverage͒. As may be seen
in Fig. 4, however, the lowest oxygen exposure to a surface
predosed with this mixture of CO isotopes immediately leads
to an upward shift and a splitting of not only the
13CO-related peak, but also of that associated with 12CO
from the singleton value of 2058 cmϪ1 to 2066.7 and
2084 cmϪ1. Identical frequencies are observed on reversing
the dosing sequence, and it is clear that the magnitude of the
shifts is the same for the majority and minority CO species.
We therefore conclude that the entire change in frequency
due to O coadsorption observed in Fig. 3 is due to an
O-induced chemical shift, and not to dipole–dipole coupling.
This means that we can exclude the possibility of CO island
formation induced by the coadsorbed O.
While these experiments clearly identify the mechanism
of the upward frequency shift induced by the coadsorbed O,
a question remains as to the origin of the splitting into two
bands. Notice that the low CO coverages (Ͻ0.2 ML) used in
these experiments are insufficient to initiate nucleation of the
(1ϫ1) phase, so the two bands cannot be associated with
CO adsorbed on the two different surface phases of Pt͕110͖.
Similarly, while there have been suggestions in the literature
of subsurface oxygen species or local oxide formation during
CO/O2 reaction conditions over Pt͕110͖, we can exclude the
local interaction of some of the CO with one of these species
as the origin of a second band; conversion of O to subsurface
splitting of the band into components at 2069 and 2088 cmϪ1
from its previous value of 2061 cmϪ1. No CO2 was pro-
duced. Further oxygen exposure results in a small increase in
the frequency of both bands. In the upper part of Fig. 3 are
shown the results of a similar experiment in which the order
of gas dosing was reversed, a predose of 4 L O2 being fol-
lowed by successive dosing with CO. Again two distinct
C–O vibrational bands appear at similar energies to those
described above, while increasing CO exposure results in
increasing amplitudes of the bands at constant frequencies.
Heating the surface to 350 K results in the partial reaction of
CO with O. Note that the chemisorbed oxygen can only be
removed by total reaction with CO at temperatures in excess
of 400 K, or by sputtering; heating an oxygen covered sur-
face to 1000 K followed by dosing with CO yielded a similar
two-peak C–O vibrational spectrum characteristic of O/CO
coadsorption. No C–O stretching frequency in the
1800–1900 cmϪ1 range, characteristic of a bridging CO spe-
cies, was observed in either experiment.
There are two possible ways in which the coadsorbed O
could give rise to the upward shift of the C–O vibrational
frequency. One of these is that O–CO repulsion could lead to
isolated CO island formation; the shift would then arise from
dipole–dipole coupling in the islands which would have a
higher local coverage than in the absence of this O–CO re-
pulsion. The alternative explanation is that the increase in
vibrational energy is due to an O-induced chemical shift. Of
course, it is also possible that both of these effects may con-
tribute to the observed phenomena. In order to distinguish
these processes, experiments similar to those leading to the
results of Fig. 3 were repeated using isotopic mixtures of
CO, and the results are shown in Fig. 4. Notice that the
spectra are complicated slightly by the presence of a feature
Downloaded 19 Aug 2013 to 128.118.88.48. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

891
J. Chem. Phys., Vol. 117, No. 2, 8 July 2002
COϩO2 reaction on Pt͕110͖
oxygen can only be achieved by prolonged exposure at 400
K.12 There is also rather clear evidence that true oxide cannot
be formed on Pt at the low pressures relevant to these
experiments,61,62 but as CO adsorption upon the oxidized
surface is known to have a characteristic frequency of
2120 cmϪ1 ͑Refs. 63 and 64͒ which is not observed here, we
can clearly discount this possibility. We must therefore con-
clude that CO and O can form a mixed chemisorbed adlayer
on the (1ϫ2) reconstructed surface in which there are two
CO species which differ in some way.
Bearing in mind that chemical shifts in vibrational spec-
troscopy arise from a local electronic effect, one possible
explanation for two bands would be that the upper band cor-
responds to CO molecules adsorbed in sites with O near
neighbors, while the lower energy band arises from CO mol-
ecules which are relatively isolated from adsorbed O atoms.
If this were the case, however, the relative intensities of the
two bands would be dependent on the relative coverages of
adsorbed O and CO. For example, for a low CO coverage,
increasing the O coverage should transfer intensity from the
lower ͑isolated CO͒ band to the higher energy ͑O neighbor-
ing͒ band. Apart from the spectrum recorded at the lowest
CO exposure after oxygen predosing, Figs. 3 and 4 indicate
there is essentially no variation in the relative intensities of
the two bands, implying that this hypothesis is not correct.
The alternative explanation for the two bands is that they
correspond to two different local adsorption configurations
of CO which have different O-induced chemical shifts, ef-
fectively implying that there are two inequivalent sites for
CO adsorption relative to the position of the coadsorbed O
atoms. In the absence of any hard information concerning the
adsorption geometries of either coadsorbate, expanding on
this idea is necessarily somewhat speculative. We note, how-
ever, that the values of the C–O stretching frequencies
clearly favor all CO molecules being in local atop sites, and
that on the (1ϫ2) reconstructed phase we have already re-
marked that, at least at high coverages, this must imply atop
adsorption on both ridge and hollow Pt atoms. If the O atoms
do occupy ͑hollow͒ sites on the ͑111͒ trough walls as seems
to be favored by the only real attempt to establish the oxygen
adsorption site,59 one might then imagine that these two dis-
tinct CO sites would suffer different O-induced chemical
shifts, and could account for the two vibrational bands ob-
served.
FIG. 5. The IRAS spectra ͑left͒ and the rate of CO2 production ͑right͒
recorded continuously while a Pt͕110͖ surface, exposed to a partial pressure
of 3.8ϫ10Ϫ5 mbar O2 and a 1.3ϫ10Ϫ5 mbar CO, was heated and cooled at
a rate of 0.31 K sϪ1. Each spectrum was recorded at a resolution of 2 cmϪ1
and required 45 s to acquire.
when the dosing sequence is reversed and a coverage CO is
prepared, an induction period is required ͑presumably during
which CO desorption is occurring͒ before reaction can occur
(475 KϽTϽ595 K).18 However, as stated in the introduc-
tion, a variety of other models can also describe this behav-
ior. The bifurcation in the reaction rate exhibited by CO oxi-
dation over Pt͕110͖ has subsequently been well documented.
The high rate branch is associated with a higher temperature
and a greater O2 :CO partial pressure ratio.1,6,7,19,65 It is
known that the surface is in the (1ϫ2) state during the high
rate branch and in the (1ϫ1) state in the low rate
branch.16,65 There are, however, no experimental data avail-
able which identify the nature and coverage of the surface
species during either rate branch.
Figure 5 displays IRAS spectra and mass spectroscopy
͑MS͒ measurements recorded concurrently while the tem-
perature of a Pt͕110͖ crystal, exposed to a constant flow of
oxygen and CO, was successively lowered and raised. The
IV. CO¿OXYGEN COADSORPTION UNDER
REACTION CONDITIONS
A. IRAS
rate of CO2 production in this figure is expressed in ML sϪ1
,
Two distinct regions of kinetic behavior have been iden-
tified in the reaction of COϩO2 on Pt͕110͖. Bonzel and
Ku18,19 observed that, for a given temperature and oxygen
pressure, the rate of CO2 production initially increased as the
partial pressure of CO increased up to a critical value, where-
upon the rate rapidly decreased. They were able to model
this behavior by considering that adsorbed CO molecules
block sites for oxygen absorption, but not vice versa. This
model was supported by their observation that an oxygenated
surface can react immediately when it is exposed to CO, but
calibrated by preparing a surface with 1 ML of adsorbed CO
at 300 K and then warming the crystal in 3.75ϫ10Ϫ5 mbar
O2 while recording the CO2 and CO partial pressures. The
measurements of Fig. 5 show a small hysteresis in the tem-
perature at which the transition between the two rate
branches occurs on cooling ͑500 K͒, and heating ͑504 K͒.
Two vibrational bands are observed, one associated with
each of the two reaction rate states. A high intensity band at
ϳ2085 cmϪ1 is seen in the low rate branch with a FWHM of
7.7 cmϪ1 and a Lorentzian line shape associated with life-
Downloaded 19 Aug 2013 to 128.118.88.48. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

892
J. Chem. Phys., Vol. 117, No. 2, 8 July 2002
Miners et al.
absorbance of 0.025 A cmϪ1. When using isotopic dilution it
has a frequency of 2056 cmϪ1. We can therefore conclude
that the frequency of 2062 cmϪ1 includes a shift due to
dipole–dipole coupling of 6Ϯ3 cmϪ1, while accounting for
the temperature dependence of the C–O stretching frequency
shown in Fig. 2 indicates that there is also a chemical shift of
5Ϯ2 cmϪ1 indicative of coadsorption with oxygen. Figure 1
also allows us to equate the dipole–dipole coupling shift to a
local CO coverage of 0.05Ϯ0.04 ML. The large Gaussian
component in the broadening of this band implies inhomo-
geneous broadening and is indicative of a large number of
adsorbate environments for the CO, as would be expected of
a steady state CO coverage, in a mixed O–CO adlayer. This
intermixing is confirmed by a determination of the average
CO coverage from the measured integrated absorption ͑using
the cross section of CO at this CO local coverage͒; the value
obtained is 0.025 ML. Despite the limited precision of the
local coverage determination, it is clear that these two values
of the local and average coverage are most obviously recon-
ciled by a fully intermixed rather than islanded CO layer.
The rather accurate determination of the average CO cover-
age as a low value of 0.025 ML clearly implies, from earlier
LEED studies, that the surface in this high rate branch of the
reaction is in the (1ϫ2) reconstructed phase.65
The IRAS band associated with the low rate branch is
observed at a frequency of 2083 cmϪ1, with an integrated
absorbance of 0.211 A cmϪ1. When using isotopic dilution it
has a frequency of 2051 cmϪ1. Indeed the IRAS band asso-
ciated with the low rate branch is identical to that obtained in
the absence of any O2 in the gas phase. The shift of 32
Ϯ3 cmϪ1 due to dynamic dipole coupling yields a value for
the local CO coverage of 0.5Ϯ0.1 ML while comparison
with the average coverage obtained from the integrated ab-
sorbance shows that this average coverage is present on 90
Ϯ10 % of the surface; i.e., there is no islanding. This cover-
age means that there can be no pairs of adjacent vacant sites
upon which O2 can dissociatively adsorb, thus providing
the first direct evidence that the low rate branch corresponds
to a CO-poisoned surface. Prior LEED studies indicate that
at this CO coverage the surface is in the unreconstructed
(1ϫ1) phase.65
FIG. 6. The IRAS spectra from Pt͕110͖ recorded under reaction conditions
at constant temperature and a CO partial pressure of 8ϫ10Ϫ6 mbar, but
with varying partial pressure of O. The associated rates of CO2 production
are also noted. In the upper set of spectra, a mixture 1:10 12CO:13CO was
used.
time broadening.28 The high rate branch is characterized by
an absorption band at lower frequency (ϳ2062 cmϪ1) with a
FWHM of 12 cmϪ1 and Voigt line shape ͑of which 8 cmϪ1
is Lorentztan͒. The difference in frequency of the two bands
arises from a combination of dynamic-dipole coupling with
increasing CO coverage and the effect of O coadsorption
producing a chemical shift. These can be separated by using
isotopic mixtures as described below. No IRAS band associ-
ated with the presence of either CO adsorbed in the bridge
configuration ͑around 1800 cmϪ1͒ or CO adsorbed on an ox-
ide surface (2120 cmϪ1) was observed.
The local CO coverage in these two states can be de-
duced from the shift due to dipole–dipole coupling, and the
local chemical environment from the chemical shift. The re-
sults of experiments to separate these components are shown
in Fig. 6. In order to keep any spectral shift due to dephasing
constant during this experiment, the system was kept at con-
stant temperature and the system varied from high rate
branch to low rate branch by varying the oxygen partial pres-
sure while maintaining the CO partial pressure constant. In
the lower half of the figure are the spectra for the high and
low rate branch using 100% 12CO ͑cf. Fig. 5͒, while the
upper half of the figure shows the results of repeating the
experiment using an isotopic mixture of 13CO and 10%
12CO. In the uppermost spectrum, recorded in the absence of
oxygen, the isotopically diluted 12CO band is centered at a
frequency of 2051 cmϪ1. This value corresponds to the fre-
quency of a band with no chemical shift and a maximum of
4 cmϪ1 dipole–dipole shift.
It is expected that the two states observed in the bifur-
cation of the reaction rate are the same two states which
coexist on the surface during pattern formation. Neverthe-
less, we may ask whether our IRAS experiments correspond
to the situation of homogeneous switching of the surface
between these two states or pattern formation. Figure 5 con-
tains two spectra in which the C–O stretching bands charac-
teristic of both of the two states are observed, but this does
not prove that the two states are present on the surface at the
same time as would be the case in pattern formation; an
alternative explanation is that the entire surface switched
from one state to the other during the time taken to accumu-
late these spectra.
To provide further information on this aspect, kinetic
rate oscillations were established and monitored simulta-
neously by time-resolved IRAS and mass spectrometry. A
representative set of IRAS spectra and the associated CO2
production rate derived from the mass spectrometry are dis-
The IRAS band associated with the high rate branch is
observed at a frequency of 2062 cmϪ1, with an integrated
Downloaded 19 Aug 2013 to 128.118.88.48. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

893
J. Chem. Phys., Vol. 117, No. 2, 8 July 2002
COϩO2 reaction on Pt͕110͖
surface in each state oscillating with time, thus indicating
that microscopic local pattern formation is occurring. The
macroscopic reaction is presumably coupled either via the
gas phase or faceting.15,65,66 Similar measurements of oscil-
lations were recorded under a wide range of conditions
(450 KϽTϽ530 K, 1ϫ10Ϫ5 mbarϽptotalϽ1ϫ10Ϫ4 mbar)
with periods from seconds to minutes, and all of these ex-
hibited the behavior shown in these two figures. Notice too
that Fig. 8 shows particularly clearly the in-phase behavior
of the CO2 partial pressure and the 2066 cmϪ1 vibrational
band, and the fact that the 2081 cmϪ1 band has an intensity
which is in antiphase to this monitor of the reaction rate,
providing a particularly clear correlation of these monitors of
the surface state with the rate of reaction.
Note that in these experiments the pumping speed was
high, ensuring that that the measured partial pressures in the
gas phase relate directly relate to the rate of reaction or pro-
duction. Furthermore, the oscillations in the pressure of the
product CO2 are exactly out of phase with those of the reac-
tants, CO and O2 . The highest rate of production of CO2
thus occurs when the rate of adsorption of the reactants is
highest, which indicates that the high rate branch in the os-
cillation reaction is occurring by adsorption and immediate
reaction on the surface, rather than by slow accumulation of
adsorbed CO followed by a ‘‘surface explosion’’ of the reac-
tion. This conclusion is in contrast to the results of the lower
pressure measurement by Meißen et al.15 Throughout the os-
cillations the frequency of the IRAS band characteristic of
high rate branch remains approximately constant at a value
of 2066 cmϪ1, but there is a small variation in the frequency
of the band associated with the low rate branch around
2081 cmϪ1. This variation can be assigned the effects of
dynamic dipole coupling either through the growth in the
size of the CO poisoned (1ϫ1) domains ͑and thus the de-
creasing proportion of island-edge molecules͒, or a slight
increase in the local coverage.
FIG. 7. A set of IRAS spectra recorded from Pt͕110͖ under conditions
giving rise to an oscillating reaction rate with a period of 63 seconds. The
spectra were recorded with 7.2 s temporal and 2 cmϪ1 spectral resolution.
Also shown is the CO2 production rate obtained from monitoring the mass
44 partial pressure on the mass spectrometer.
played in Fig. 7, while the amplitudes of the main IRAS and
mass spectrometer peaks are shown as a function of time in
Fig. 8. The oscillations in the mass spectrometer signals
clearly show that a macroscopic kinetic oscillation in the
reaction rate is occurring. In addition, however, the IRAS
data show the presence on the surface of the same two states
which are observed on either side of the reaction rate bifur-
cation. The fact that both IRAS bands are present at approxi-
mately constant frequency throughout the oscillations, but
change their relative intensities, indicates that the two asso-
ciated species are spatially separated; they coexist on differ-
ent parts of the surface but with the relative areas of the
B. PEEM
The results of repeating the conditions of our IRAS ex-
periments using PEEM are presented in Fig. 9. To relate the
setup to the previous PEEM calibration measurements30 and
account for day-to-day fluctuations in the measured PEEM
intensity, the intensities obtained from a clean surface, a sur-
face of known CO coverage,32 and the O-saturated surface
were measured.31,32 The numerical values of these average
intensities ͑shown in arbitrary units͒ are expressed as the
darkness, so the value is zero for the clean surface which
shows a bright image and is large for the darkest image ͑cor-
responding to the largest work function͒. The value of the
saturated oxygen coverage is taken as being 0.7 ML. Follow-
ing these initial calibration experiments, a partial pressure of
CO was introduced and the oxygen partial pressure varied to
obtain the reaction conditions corresponding to low rate
branch, the high rate branch and pattern formation. The work
function increase associated with O adsorption is much
larger than that due to CO adsorption ͑0.8 and 0.15 eV, re-
spectively͒, so the PEEM intensity is far more sensitive to
the presence of oxygen, especially at low coverages. As
shown in Fig. 9, the measured PEEM intensity when the
FIG. 8. Summary of the main results of the experiment shown in raw data
form in Fig. 7, notably the temporal variation of the partial pressures of O2 ,
CO, and CO2 , the integrated absorbance of the two IRAS bands and the
exact frequency of the IRAS peak around the 2081 cmϪ1 band.
Downloaded 19 Aug 2013 to 128.118.88.48. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

894
J. Chem. Phys., Vol. 117, No. 2, 8 July 2002
Miners et al.
The measurements were repeated at different tempera-
tures and pressures (440 KϽTϽ530 K, 1ϫ10Ϫ5 mbar
ϽptotalϽ5ϫ10Ϫ4 mbar), both with PEEM and IRAS. Al-
though in IRAS, the same two coverages of CO were ob-
served, and the PEEM intensity of the low rate branch re-
mained constant, the intensity of the high rate branch showed
considerable variation under different reaction conditions,
and the full range of results of obtained indicate that the high
rate branch can correspond to oxygen coverages in the wide
range from 0.3–0.7 ML. The large range also reflects not
only a real variation in the oxygen coverage but also a re-
duced precision for the coverage determination when the
work function becomes very large and the PEEM intensity is
far less sensitive to its exact value. One clear example of the
real variation in oxygen coverage in those parts of the sur-
face reacting under the high rate branch conditions can be
seen in the PEEM image obtained under the conditions of
pattern formation in Fig. 9͑F͒. While the bright regions cor-
responding to the low rate branch have a brightness very
similar to those of panels ͑B͒ and ͑D͒ as remarked earlier, the
dark regions of these images clearly are less dark than in
panels ͑C͒ and ͑E͒, indicating a subsaturation oxygen cover-
age. This effect has previously been reported in the work
function measurements of Eiswirth and Ertl.6,7
We should remark that the variability of the oxygen cov-
erage in the high reaction rate condition and the more con-
stant high CO coverage of the surface in the low rate condi-
tion are consistent with expectations based on the site
blocking and reconstruction model of the reaction.2,18,19,60,67
In this model the low reaction rate condition is attributed to
the fact that two neighboring vacant sites are required for
molecular oxygen dissociation subsequent atomic oxygen re-
action, but a local CO coverage of 0.5 ML blocks adjacent
sites and thus inhibits oxygen dissociation and poisons the
reaction. In the high rate condition, however, the oxygen
steady state coverage simply needs to be sufficient to prevent
the coverage of CO rising above the level required to cause
statistical nucleation of (1ϫ1) unreconstructed-surface
islands.68–70
FIG. 9. PEEM images from Pt͕110͖ recorded at 483 K: ͑A͒ clean surface;
͑B͒ CO-covered surface ͑exposed to pCOϭ1.5ϫ10Ϫ5 mbar͒, 0.85 ML CO
͑Ref. 64͒; ͑C͒ oxygen-covered surface ͑exposed to pO ϭ4.5ϫ10Ϫ5 mbar͒;
2
͑D͒ surface undergoing reaction in the high rate branch ͑pO2ϭ4.5
ϫ10Ϫ5 mbar, pCOϭ1.5ϫ10Ϫ5 mbar͒; ͑E͒ surface undergoing reaction in
the low rate branch ͑pO ϭ4.0ϫ10Ϫ5 mbar, pCOϭ1.5ϫ10Ϫ5 mbar͒. Note
2
that the oxygen coverage quoted in this panel of the figure relates to the
results of a range of different experiments and not only the specific image
shown here; ͑F͒ surface undergoing reaction under conditions of pattern
formation ͑pO ϭ4.2ϫ10Ϫ5 mbar, pCOϭ1.5ϫ10Ϫ5 mbar͒.
2
surface is entirely reacting under the conditions of the low
rate branch ͓Fig. 9͑D͔͒ is almost the same as when the sur-
face is only exposed to CO ͓Fig. 9͑B͔͒. Furthermore the
same intensity is observed in the light patches of the PEEM
images during the conditions of pattern formation ͓Fig.
9͑F͔͒. These results are consistent with the conclusions of
our IRAS study that under these conditions the surface is
essentially free of adsorbed oxygen. Based on the IRAS data
we concluded that in the low reaction rate branch the surface
has ⌰COϭ0.5Ϯ0.1 ML and the PEEM intensity then indi-
cates that ⌰Oϭ0.03Ϯ0.01 ML.
When the surface is reacting in the high rate branch, the
IRAS data indicate a CO coverage of only 0.05Ϯ0.03 ML,
and the small effect which this is expected to have on the
numerical value of the PEEM intensity ͑1.6Ϯ1.0 arbitrary
units͒ is clearly an almost negligible contribution to the mea-
sured total value of 140 arbitrary units. The dominant inten-
sity change must therefore be ascribed to a high oxygen cov-
erage. The high rate branch is identified as being an
oxygenated surface in the (1ϫ2) phase; strictly the results
of panels ͑C͒ and ͑E͒ of Fig. 9 imply an oxygen coverage of
⌰Oϭ0.7 ML, but the coverage range quoted in Fig. 9 re-
flects the average of many measurements under different
conditions as described below.
V. CONCLUSIONS
Our primary objective in this investigation has been to
use IRAS, together with some additional PEEM experiments,
to obtain quantitative information on the CO and oxygen
coverage of a Pt͕110͖ surface as it undergoes an oscillatory
CO oxidation reaction, and in particular to provide this char-
acterization of the surface when in the low and high reaction
rate branches of this process. Using IRAS to provide this
information has required the results of extensive initial cali-
bration experiments to determine how the C–O stretching
frequency and integrated absorbance vary as a function of
CO coverage, oxygen coadsorption and temperature, and to
separate out the roles of dynamic dipole–dipole coupling,
chemical shifts and dephasing. In the process of collecting
this information, a number of important further results have
been obtained. In particular, our results show clearly that CO
is distributed rather evenly over the surface at all coverages
with and without coadsorbed atomic oxygen, and does not
form extended islands. Coadsorption experiments at 300 K
Downloaded 19 Aug 2013 to 128.118.88.48. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

895
J. Chem. Phys., Vol. 117, No. 2, 8 July 2002
COϩO2 reaction on Pt͕110͖
also indicated that there are two different sites for CO ad-
sorption in presence of oxygen. We propose that these may
be assigned to CO molecules which adsorb atop hollow and
ridge Pt atoms on the (1ϫ2) surface; molecules occupying
these two sites may then be rendered inequivalent in terms of
oxygen-induced chemical shifts by the specific adsorption
sites of the O atoms, probably on the ͕111͖-like trough sides
of this missing-row reconstructed surface. We should per-
haps note here, however, that under reaction conditions there
was no evidence of co-occupation of these two states. The
reason for this is not known, but may be related to the sig-
nificantly higher temperature under reaction conditions
which may cause one of these sites to be more strongly fa-
vored.
Using the results of these calibration experiments, how-
ever, we have also been able to interpret the results of similar
studies of the surface under reaction conditions, IRAS pro-
viding primary information on the CO coverage and the local
coordination site, while PEEM has provided information on
the oxygen coverages. Changes on the surface have been
related directly to the CO2 production rate under both steady
state and oscillatory reaction conditions. These measure-
ments allow us to clearly exclude some processes which
have previously been considered in trying to understand this
system. For example, we find no evidence for an oxide phase
or subsurface oxygen, nor for any local CO coordination site
other than atop ͑such a bridge͒.
The CO and oxygen coverages on the Pt͕110͖ surface
when undergoing the CO oxidation reaction in both low and
high reaction rate branches have been determined. In the low
rate branch the surface has a high CO coverage and low
oxygen coverage (⌰COϭ0.5Ϯ0.1 ML, ⌰Oϭ0.03Ϯ0.01)
which must therefore be in the (1ϫ1) unreconstructed
phase. By contrast, under high reaction rate branch condi-
tions, the surface has a high oxygen coverage and low CO
coverage (⌰COϭ0.05Ϯ0.03 ML, ⌰Oϭ0.3–0.7 ML) which
must therefore be in the reconstructed (1ϫ2) phase. The
states observed to make up the high and low reaction
branches are responsible for the two main phases observed
during pattern formation. These findings are entirely consis-
tent with the site blocking and reconstruction model of the
reaction.2,18,19,60,67
Based on this new information and the prior models of
the system which are consistent with it, we can relate the two
states observed in the hysteresis and in the spatiotemporal
pattern formation exhibited by this system as observed by
PEEM, IRAS, and LEEM. The high rate phase observed in
the ͑spatio-͒temporal pattern formation is an oxygenated
(1ϫ2) surface; the low rate corresponds to regions of
(1ϫ1) surface with a high ͑0.5 ML͒ coverage of CO. The
key processes are the role of high CO coverages in blocking
sites for O2 dissociation and the influence of the surface
reconstruction and unreconstruction.2,15,18,19,60,67 It appears
that oxygen dissociation and adsorption is inhibited when
data are inconsistent with the heterogeneous nucleation and
growth of islands of this coverage.
ACKNOWLEDGMENTS
The authors would like to thank F. Meißen and A.
Patchett for their close cooperation and useful discussions
and likewise the group of H. H. Rotermund.
1 R. Imbihl and G. Ertl, Chem. Rev. 95, 697 ͑1995͒.
2 G. Ertl, Surf. Sci. 299Õ300, 742 ͑1994͒.
3 H. H. Rotermund, Surf. Sci. Rep. 29, 265 ͑1997͒.
4 Oscillations and Traveling Waves in Chemical Systems, edited by R. J.
Field and M. Burger ͑Wiley, New York, 1985͒.
5 I. Ferino and E. Rombi, Catal. Today 52, 291 ͑1999͒.
6 M. Eiswirth and G. Ertl, Surf. Sci. 177, 90 ͑1986͒.
7
¨
M. Eiswirth, Doctorate thesis, University of Munchen, 1987.
8 S. Ladas, R. Imbihl, and G. Ertl, Surf. Sci. 198, 42 ͑1988͒.
9 W. Tappe, U. Korte, and G. Meyer-Ehmsen, Surf. Sci. 388, 162 ͑1997͒.
10 G. Ertl, Science 254, 1750 ͑1991͒.
11 S. Jakubith, H.-H. Rotermund, W. Engel, A. von Oertzen, and G. Ertl,
Phys. Rev. Lett. 65, 3013 ͑1990͒.
12 A. von Oertzen, A. S. Mikhailov, H. H. Rotermund, and G. Ertl, J. Phys.
Chem. B 102, 4966 ͑1998͒.
13 K. C. Rose, D. Battogtokh, A. Mikhailov, R. Imbihl, W. Engel, and A. M.
Bradshaw, Phys. Rev. Lett. 76, 3582 ͑1996͒.
14 K. C. Rose, B. Berton, R. Imbihl, W. Engel, and A. M. Bradshaw, Phys.
Rev. Lett. 79, 3427 ͑1997͒.
15 F. Meißen, A. J. Patchett, W. Engel, A. M. Bradshaw, and R. Imbihl
͑unpublished͒.
16 A. J. Patchett, F. Meißben, W. Engel, R. Imbihl, and A. M. Bradshaw,
Surf. Sci. 454–456, 341 ͑2000͒.
17 M. P. Cox, G. Ertl, and R. Imbihl, Phys. Rev. Lett. 54, 1725 ͑1985͒.
18 H. P. Bonzel and R. Ku, Surf. Sci. 33, 91 ͑1972͒.
19 H. P. Bonzel and R. Ku, J. Vac. Sci. Technol. 9, 663 ͑1972͒.
20 G. S. Yablonskii and V. I. Elokhin, in Perspectives in Catalysis, edited by
J. A. Thomas and K. I. Zamaev ͑Blackwell Scientific, Boston, 1994͒.
21 B. C. Sales, J. E. Turner, and M. B. Maple, Surf. Sci. 114, 381 ͑1982͒.
22 N. A. Collins, S. Sunaresan, and Y. J. Chabal, Surf. Sci. 180, 272 ͑1987͒.
23 J. H. Miners, R. Martin, P. Gardner, R. Nalezinski, and A. M. Bradshaw,
Surf. Sci. 377–379, 791 ͑1997͒.
24 A. M. Bradshaw and E. Schweizer, Adv. Spectrosc. ͑N.Y.͒ 23, 413 ͑1988͒.
25 M. Scheffler, Surf. Sci. 81, 562 ͑1979͒.
26 B. N. J. Persson and R. Ryberg, Phys. Rev. B 24, 6954 ͑1981͒.
27 R. K. Sharma, W. A. Brown, and D. A. King, Surf. Sci. 414, 68 ͑1998͒.
28 J. W. Gadzuk and A. C. Luntz, Surf. Sci. 144, 429 ͑1984͒.
29 W. Engel, M. E. Kordesch, H. H. Rotermund, S. Kubala, and A. von
Oetzen, Ultramicroscopy 36, 148 ͑1991͒.
30
¨
¨
A. von Oertzen, Doctorate thesis, Frei-Universitat, Berlin, 1992.
31 S. Cerasari, Doctorate thesis, Frei-Universitat, Berlin, 2000 ͑http://
www.diss.fu-berlin.de/2000/61/indexe.html͒.
32 S. Cerasari, A. von Oertzen, and H. H. Rotermund ͑unpublished͒.
33 A. von Oertzen, H. H. Rotermund, and S. Nettesheim, Surf. Sci. 311, 322
͑1994͒.
34 P. Fery, W. Moritz, and D. Wolf, Phys. Rev. B 38, 7275 ͑1988͒.
35 C. M. Comrie and R. M. Lambert, J. Chem. Soc., Faraday Trans. 72, 1659
͑1976͒.
36 R. Imbihl, S. Ladas, and G. Ertl, Surf. Sci. 206, L903 ͑1988͒.
37 T. Gritsch, D. Coulman, R. J. Behm, and G. Ertl, Phys. Rev. Lett. 63, 1086
͑1989͒.
38 B. E. Hayden, A. W. Robinson, and P. M. Tucker, Surf. Sci. 192, 163
͑1987͒.
39 P. Hofmann, S. R. Bare, and D. A. King, Surf. Sci. 117, 245 ͑1982͒.
40 S. Ferrer and H. P. Bonzel, Surf. Sci. 119, 234 ͑1982͒.
41 P. Hofmann, S. R. Bare, N. V. Richardson, and D. A. King, Solid State
Commun. 42, 645 ͑1982͒.
42 M. Nowicki, A. Emundts, G. Pirug, and H. P. Bonzel, Surf. Sci. 478, 180
͑2001͒.
⌰
⌰
CO(local)Ͼ0.3 ML and that the reaction is poisoned when
CO(local)ϭ0.5 ML; moreover, homogeneous nucleation of
43 M. I. Ban, M. A. van Hove, and G. A. Somorjai, Surf. Sci. 185, 355
͑1987͒.
the unreconstructed (1ϫ1) phase occurs at ⌰CO(local)
ϭ0.2 ML and locally these small islands are thought to have
a coverage of CO of 0.5 ML. Notice, however, that our IRAS
44 R. M. Hammaker, S. A. Francis, and R. P. Eischens, Spectrochim. Acta 21,
1295 ͑1965͒.
45 G. D. Mahan and A. A. Lucas, J. Chem. Phys. 68, 1344 ͑1978͒.
Downloaded 19 Aug 2013 to 128.118.88.48. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

896
J. Chem. Phys., Vol. 117, No. 2, 8 July 2002
Miners et al.
58
46 S. R. Bare, P. Hofmann, and D. A. King, Surf. Sci. 144, 347 ͑1984͒.
A. V. Walker, B. Klotzer, and D. A. King, J. Chem. Phys. 109, 6879
͑1998͒.
¨
47
¨
C. Klunker, M. Balden, S. Lehwald, and W. Daum, Surf. Sci. 360, 104
59 S. Helveg, H. T. Lorensen, S. Horch, E. Lægsgaard, I. Stensgaard, K. W.
Jacobsen, J. K. Nørskov, and F. Besenbacher, Surf. Sci. 430, L533 ͑1999͒.
60 T. Engel and G. Ertl, Adv. Catal. 28, 1 ͑1979͒.
͑1996͒.
48 C. B. Harris, R. M. Schelby, and P. A. Cornelius, Phys. Rev. Lett. 38, 1415
͑1977͒.
49 B. N. J. Persson, F. M. Hoffmann, and R. Ryberg, Phys. Rev. B 34, 2266
͑1986͒.
61 M. Peuckert and H. P. Bonzel, Surf. Sci. 145, 239 ͑1984͒.
62 G. Pirug, R. Dziembaj, and H. P. Bonzel, Surf. Sci. 221, 553 ͑1989͒.
63 T. H. Lindstrom and T. T. Tsotsis, Surf. Sci. 150, 487 ͑1985͒.
64 D. J. Kaul and E. E. Wolf, J. Catal. 91, 216 ͑1985͒.
65 J. Falta, R. Imbihl, M. Sander, and M. Henzler, Phys. Rev. B 45, 6858
͑1992͒.
50 B. N. J. Persson, Chem. Phys. Lett. 149, 278 ͑1988͒.
51
¨
E. Schweizer, B. N. J. Person, M. Tushaus, D. Hoge, and A. M. Bradshaw,
Surf. Sci. 213, 49 ͑1989͒.
52
53
¨
M. Tushaus, E. Schweizer, P. Hollins, and A. M. Bradshaw, J. Electron
Spectrosc. Relat. Phenom. 45, 57 ͑1987͒.
66 J. Lauterbach and H. H. Rotermund, Catal. Lett. 27, 2 ͑1994͒.
67 R. Imbihl, Prog. Surf. Sci. 44, 185 ͑1993͒.
¨
H. Conrad, G. Ertl, and J. Kuppers, Surf. Sci. 76, 332 ͑1978͒.
54 G. E. Thomas and W. H. Wienberg, J. Chem. Phys. 70, 954 ͑1979͒.
55 F. M. Hoffman, M. D. Weisel, and C. H. F. Peden, Surf. Sci. 253, 59
͑1991͒.
68 V. P. Zhdanov, Surf. Sci. 426, 345 ͑1999͒.
69
¨
O. Kortluke, V. N. Kuzovkov, and W. von Neissen, Phys. Rev. Lett. 83,
56 J. Schmidt, Ch. Stuhlmann, and H. Ibach, Surf. Sci. 284, 121 ͑1993͒.
57 N. Freyer, M. Kiskinova, G. Pirug, and H. P. Bonzel, Surf. Sci. 166, 206
͑1986͒.
3089 ͑1999͒.
70
¨
O. Kortluke, V. N. Kuzovkov, and W. von Neissen, J. Chem. Phys. 110,
11523 ͑1999͒.
Downloaded 19 Aug 2013 to 128.118.88.48. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions