Kinetics of the CO oxidation reaction on Pt(111) studied by in situ high-resolution x-ray photoelectron spectroscopy

Article abstract of DOI:10.1063/1.1669378

The coadsorption and reaction of CO with atomic oxygen on Pt(111) was examined by in situ high-resolution core level spectroscopy using synchrotron radiation. A atomic oxygen layer with a coverage of 0.25 ML was exposed to CO from a supersonic molecular beam at various sample temperatures and pressures. X ray photoelectron spectra were collected at photon energies of 380 eV for C 1s and 650 eV for the O 1s region in normal emission geometry. From an Arrhenius plot of the rate constant between 275 and 305 K an activation energy of (0.53±0.04) eV and a prefactor of 4.7×10 ^6±0.7s^-1 were determined.

Full text of DOI:10.1063/1.1669378

Kinetics of the CO oxidation reaction on Pt(111) studied by in situ high-resolution x-ray
photoelectron spectroscopy
M. Kinne, T. Fuhrmann, J. F. Zhu, C. M. Whelan, R. Denecke, and H.-P. Steinrück
Citation: The Journal of Chemical Physics 120, 7113 (2004); doi: 10.1063/1.1669378
View online: http://dx.doi.org/10.1063/1.1669378
View Table of Contents: http://scitation.aip.org/content/aip/journal/jcp/120/15?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Mechanism of CO oxidation reaction on O-covered Pd(111) surfaces studied with fast x-ray photoelectron
spectroscopy: Change of reaction path accompanying phase transition of O domains
J. Chem. Phys. 124, 224712 (2006); 10.1063/1.2205856
Kinetics of CO oxidation on high-concentration phases of atomic oxygen on Pt(111)
J. Chem. Phys. 123, 224703 (2005); 10.1063/1.2126667
Kinetic parameters of CO adsorbed on Pt(111) studied by in situ high resolution x-ray photoelectron
spectroscopy
J. Chem. Phys. 117, 10852 (2002); 10.1063/1.1522405
The CO oxidation kinetics on supported Pd model catalysts: A molecular beam/in situ time-resolved infrared
reflection absorption spectroscopy study
J. Chem. Phys. 114, 4669 (2001); 10.1063/1.1342240
High-resolution x-ray photoelectron spectroscopy study of the ethanol oxidation reaction on Pd(110)
J. Chem. Phys. 110, 8703 (1999); 10.1063/1.478777
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.159.70.209 On: Wed, 03 Dec 2014 11:56:49

JOURNAL OF CHEMICAL PHYSICS
VOLUME 120, NUMBER 15
15 APRIL 2004
Kinetics of the CO oxidation reaction on Pt„111… studied by in situ
high-resolution x-ray photoelectron spectroscopy
a)
b)
M. Kinne, T. Fuhrmann, J. F. Zhu, C. M. Whelan, R. Denecke, and H.-P. Steinr u¨ ck
Physikalische Chemie II, Universit a¨ t Erlangen-N u¨ rnberg, Egerlandstr. 3, 91058 Erlangen, Germany
͑
Received 18 September 2003; accepted 20 January 2004͒
High-resolution x-ray photoelectron spectroscopy has been used to study the kinetics of the CO
oxidation reaction on a Pt͑111͒ surface in situ. The study focuses on the interaction of a preadsorbed
p(2ϫ2) layer of atomic oxygen with CO dosed using a supersonic molecular beam. Measurements
of O 1s and C 1s spectra at 120 K show that CO adsorbs on the oxygen precovered substrate, but
no reaction occurs. A maximum CO coverage of 0.23 ML ͑monolayer͒ is observed, with CO
exclusively bound on on-top sites. In accordance with the literature, bridge sites are blocked by the
presence of atomic oxygen. The reaction of CO with preadsorbed O to CO is studied isothermally
2
in a temperature range between 275 and 305 K. The reaction rate initially increases with CO
Ϫ7
pressure, but saturates at 9ϫ10 mbar. The data indicate that a certain amount of disordered
oxygen within the p(2ϫ2) layer acts as a starting point of the reaction and for a given temperature
reacts with a higher rate than O in the well-ordered oxygen p(2ϫ2) phase. For the reaction of CO
with this ordered phase, the results confirm the assumption of a reaction mechanism, which is
restricted to the edges of compact oxygen islands. The activation energy of the reaction is
determined to (0.53Ϯ0.04) eV, with a prefactor of 4.7ϫ106
of Physics. ͓DOI: 10.1063/1.1669378͔
Ϯ0.7 Ϫ1
s
. © 2004 American Institute
I. INTRODUCTION
eV in the limit of low coverages from isothermal molecular
beam experiments, but also a strong coverage dependence of
the energy. The latter value was obtained with the assump-
In the field of heterogeneous catalysis the oxidation of
carbon monoxide to CO on the Pt͑111͒ surface is probably
2
tion of a first-order dependence of the rate on the CO and O
the most extensively studied example of an activated
Langmuir–Hinshelwood reaction.1 The common interest in
this particular system is on the one hand due to its techno-
logical relevance in automotive catalysts; on the other hand,
its relative simplicity makes it an ideal model system for the
understanding of surface reactions.
4,10
coverages. In contrast to that, Gland and Kollin
proposed
,2
that the reaction is restricted to the edges of oxygen islands.
More recently, this was confirmed by real-time scanning tun-
neling microscopy ͑STM͒ experiments of Wintterlin and
7,8
co-workers, resulting in a reaction order of the atomic oxy-
gen coverage near 0.5. This finding disproves the conclusion
The reaction mechanism is known to consist of the fol-
11
from isotopic labeling experiments of Akhter and White
2
lowing substeps: ͑1͒ adsorption of O and CO, ͑2͒ dissocia-
2
that the reaction occurs within the mixed p(2ϫ2) O–CO
tion of O ͑ad͒ to 2 O ͑ad͒, ͑3͒ reaction COϩO→CO , and
2
2
coadsorption phase. The geometric structure of the mixed
finally ͑4͒ desorption of CO . In order to investigate the
2
12
phase has been examined by Bleakley and Hu using den-
fundamental properties of the reaction step ͑3͒ it was found
helpful to exclude the influence of the oxygen adsorption and
sity functional theory calculations. Wintterlin and co-workers
calculated an activation energy of 0.49 eV, which is found
constant with respect to the oxygen partial coverage. How-
ever, this experimental approach does not provide informa-
tion about different adsorption sites and it is limited to rather
low temperatures ͑up to 274 K͒. Spectroscopic information
on the O/CO coadsorbate phase, mainly from high-resolution
dissociation process by starting out with a surface covered
with atomic oxygen, on which CO is dosed subsequently.3–8
Then the adsorption of CO remains the only additional lim-
iting factor, as the reaction product is known to desorb very
rapidly at temperatures above ϳ270 K at which the reaction
9
takes place.
1
0,13
electron-energy-loss spectroscopy ͑HREELS͒
and infra-
As a common method to study reaction kinetics, mass
red absorption spectroscopy ͑IRAS͒,14 is available from
static experiments—i.e., for low temperatures—where no re-
action takes place. Agreement is found that CO adsorbs on a
saturated atomic oxygen layer only on linearly coordinated
spectrometry has been used to detect desorbing CO . For
2
example, Gland and Kollin4 found by temperature-
programmed reaction experiments activation energies rang-
ing from 1.72 eV for small coverages to 0.70 eV for larger
coverages of CO and O. More recently, Zaera and
͑on-top͒ sites, whereas on the clean Pt͑111͒ surface also
5
,6
twofold-coordinated ͑bridge͒ sites are populated.
co-workers obtained a strongly disagreeing value of 0.39
Despite the thorough research performed so far, the out-
lined discrepancies in the recent literature indicate that the
CO oxidation is still not completely understood and should
be examined by new surface science methods. Since the
a͒
Present address: IMEC, Kapeldreef 75, B-3001 Leuven, Belgium.
Author to whom correspondence should be addressed. Electronic mail:
Reinhard.Denecke@chemie.uni-erlangen.de
b͒
0021-9606/2004/120(15)/7113/10/$22.00
7113
© 2004 American Institute of Physics
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
30.159.70.209 On: Wed, 03 Dec 2014 11:56:49
1

7114
J. Chem. Phys., Vol. 120, No. 15, 15 April 2004
Kinne et al.
availability of third-generation synchrotron radiation the use
of in situ high-resolution x-ray photoelectron spectroscopy
sample cleaning͒ and cooled by a LN cryostat to 95 K.
2
During XPS, the sample is heated radiatively by a filament
positioned at the back in order to avoid disturbing magnetic
fields caused by resistive heating. The temperature is mea-
sured by a K-type thermocouple spot-welded to the edge of
the sample. A linear temperature ramp can be applied by a
programmable temperature controller ͑Eurotherm͒. The
͑
HR-XPS͒ has been proven to be a powerful tool in the in-
1
5
vestigation of surface reactions, as it provides site-specific
quantitative information on the adsorbates. In particular, CO
oxidation has already been studied on other surfaces—e.g.,
Rh͑110͒ ͑Ref. 16͒ or Pd͑110͒ ͑Ref. 17͒—by this method.
In this paper, time-resolved in situ HR-XPS measure-
ments on the coadsorption and reaction of CO with atomic
oxygen on Pt͑111͒ are presented. First, the adsorption of CO
on a p(2ϫ2) atomic oxygen structure at low temperature
ϩ
sample was cleaned by repeated cycles of Ar ion bombard-
Ϫ7
ment ͑1 kV͒, heating in oxygen (1.3ϫ10 mbar, 300–800
K, approximately 2 K/s ramp͒, and annealing to 1300 K.
After the sample was bulk cleaned, carbon contaminants on
the surface could be removed efficiently by a single cycle of
heating in oxygen and annealing.
XP spectra were collected at photon energies of 380 eV
for C 1s and 650 eV for the O 1s region in normal emission
geometry. The light incidence angle in that case was 50° with
respect to the surface normal. As the oxygen coverage is the
main observable for the determination of kinetic properties,
mainly O 1s spectra have been taken. For time-resolved mea-
surements, spectra in this region could be collected within a
͑100–120 K͒ is examined and compared to results obtained
near room temperature ͑275 and 295 K͒. The dependence of
the reaction rate on the CO gas phase pressure at a tempera-
ture of 295 K is the subject of the next section. The reaction
order and activation energy are extracted from isothermal
reaction measurements in the range between 275 and 305 K
Ϫ6
for high sample pressure (1.3ϫ10 mbar). Different ki-
netic models for the reaction mechanism are discussed. Fi-
nally, results of a temperature-programmed reaction experi-
ment are used to compare with the result of the isothermal
measurement.
time as short as 5 s per spectrum (E ϭ527–536 eV) at a
B
resolution of ϳ250 meV. The reproducibility of binding en-
ergy values within this study is Ϯ30 meV; the calibration of
the absolute binding energy scale compared to other studies
has an uncertainty of typically Ϯ150 meV.
II. EXPERIMENT
All experiments were performed at the BESSY II syn-
chrotron radiation facility in Berlin, Germany, in a transport-
able UHV chamber, which has already been described
III. RESULTS AND ANALYSIS
A. General aspects
1
8–20
elsewhere.
Briefly, it comprises three separate chambers:
͑1͒ An analysis chamber, equipped with a hemispherical
electron energy analyzer ͑Omicron EA 125 U7 HR͒; syn-
chrotron radiation from beamlines U49/1-SGM or U49/2-
PGM1 enters at an angle of 50° with respect to the analyser
lens system in the plane of linear light polarization. ͑2͒ In the
molecular beam chamber a collimated supersonic beam is
produced by expanding CO through a 100-␮m-diam Mo
nozzle into the first of three successive differential pumping
stages. The molecular beam enters the analysis chamber at an
angle of 45° with respect to the analyzer lens system; it runs
through the focal point of the analyzer in order to allow for
in situ measurement of adsorption processes. A flag, mounted
on a pneumatic feedthrough in the second stage, allows for
fast ͑Ͻ 0.1 s͒ switching of the beam. The CO ͑purity 4.7͒
was further purified by passing the gas line through a cooling
trap at Tϳ220 K in order to remove Ni carbonyls, which can
be produced in the stainless-steel supply tube. In this work
the nozzle is held at a constant temperature of 303 K; the
kinetic energy of the CO molecules is calculated as 0.09 eV
In order to examine the properties of the CO oxidation
reaction without the influence of the dissociation of oxygen
on the sample, all experiments were performed by starting
out with a surface precovered with atomic oxygen onto
which CO was dosed subsequently. During the reaction, no
further oxygen is added to the surface. At a coverage of 0.25
ML ͑monolayer͒ ͑1 ML is defined as one adsorbate atom/
molecule per substrate surface atom͒ atomic oxygen is
2
2
known to form a p(2ϫ2) structure where O atoms are
adsorbed in threefold fcc hollow sites.2 This structure was
prepared by offering a saturation dose of oxygen ͓here a dose
3,24
Ϫ6
of 20 L (1 Lϭ10 Torr s) was applied͔ at 120 K and sub-
sequent annealing to 300 K, which results in dissociation and
desorption of part of the oxygen.23 After cooling the sample
in vacuum to ϳ100 K a sharp p(2ϫ2) LEED structure could
be observed, which was checked after each preparation.
The bold lines in Figs. 1͑a͒ and 1͑b͒ show the O 1s XP
spectrum of the O-p(2ϫ2) structure; it consists of a single
asymmetric peak at a binding energy of 529.88 eV, in very
good agreement with the literature25 ͑all binding energy val-
ues are derived by fitting the data with model functions—see
below͒. Subsequent dosing of CO at a background pressure
(
Ekinϭ7/2 kT; see, e.g., Ref. 21 and references therein͒. ͑3͒
Finally, a preparation chamber contains low-energy electron
diffraction ͑LEED͒ optics, an ion gun, and a quadrupole
mass spectrometer. The sample holder is mounted on a xyz
manipulator with rotational feedthrough, which also allows
for variation of the polar angle orientation of the sample.
Alternatively to the supersonic molecular beam, gas can also
be dosed into the analysis and preparation chambers by a
conventional dosing system, which in this work was used for
oxygen dosing ͑Messer Griessheim, purity 4.8͒.
Ϫ9
of 4.4ϫ10 mbar and a sample temperature of 100 K leads
to the evolution of a second peak located at 532.88 eV ͓Fig.
1͑a͔͒, which shows a slight shift of 0.08 eV towards larger
binding energies during uptake. This peak is identified as
on-top CO by comparison with spectra of CO on clean
Pt͑111͒,1
9,20
where a value of 532.80 eV was observed. Dur-
The Pt͑111͒ sample ͑diameter 10 mm͒ is spot-welded to
Ta wires and can be heated resistively up to 1500 K ͑for
ing CO adsorption, the height of the oxygen-related peak is
attenuated by less than 10% ͑for exact values see below͒,
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.159.70.209 On: Wed, 03 Dec 2014 11:56:49

J. Chem. Phys., Vol. 120, No. 15, 15 April 2004
Kinetics of CO-oxidation on Pt(III)
7115
atomic oxygen layer could be due to the preparation proce-
dure ͑see above͒: During heating of the saturated molecular
adlayer, oxygen desorbs and, depending on the dissociation
and desorption kinetics, vacancies within the remaining or-
dered atomic oxygen layer can be created. A lower local
oxygen coverage could also be present at domain boundaries
of the p(2ϫ2) adlayer.
For temperatures of 275 K and above, a reaction is ob-
served, as can be seen from the O 1s spectra in Fig.
1
͑b͒: the oxygen peak decreases while a third peak evolves
1
9
at 530.93 eV, arising from CO adsorbed on bridge sites.
During the reaction, again a small change in binding energies
of the on-top and bridge CO peaks is observed; the on-top
peak shifts from 532.93 eV to lower binding energy by 0.10
eV, while the bridge peak increases its binding energy by
about the same value. Peak shifts of that magnitude can be
explained by adsorbate–adsorbate interactions, which can be
different for the different local surroundings ͑on-top CO in
the mixed CO–O layer and bridge mainly in pure CO re-
gions͒. The binding energy of the oxygen peak does not
show any significant changes. Note that due to the high CO
Ϫ6
pressure of 1.3ϫ10 mbar in this case, the on-top CO peak
reaches a high intensity immediately after switching on the
molecular beam. CO , which is produced during the reac-
2
tion, cannot be observed on the sample as it desorbs rapidly
at temperatures where the reaction takes place.22 C 1s spectra
during reaction at higher temperatures ͑spectra not shown͒
reflect an increase of the bridge related peak at 286.06 eV,
similar to O 1s results. The position of this peak is not al-
tered by the presence of oxygen, whereas the on-top species
is slightly shifted downwards from 286.76 eV by 0.06 eV.
In order to extract quantitative information and exact
peak positions from the raw data, the spectra are fitted with
model line shapes. The following procedure is applied to the
O 1s spectra: First a straight line fitted to the O 1s region of
the clean substrate is subtracted from all spectra, and after
FIG. 1. Normal emission XP spectra during uptake of CO on an O-p(2
ϫ2)-precovered Pt͑111͒ surface. ͑a͒ O 1s region, Tϭ100 K, pCOϭ4.4
Ϫ9
ϫ10 mbar, time between spectra: 13 s. ͑b͒ O 1s region, Tϭ275 K, p
CO
Ϫ6
26
ϭ1.3ϫ10 mbar, time between spectra: 5 s. Bold lines: spectra of pure
O-p(2ϫ2) adlayer. ͑c͒ C 1s region, Tϭ100 K, pCOϭ4.4ϫ10 mbar, time
that, a Shirley-type background subtraction routine is ap-
Ϫ9
plied. Then the spectra can be described by a convolution of
between spectra: 13 s. The gas phase pressure was switched on after the first
spectrum in all cases. Excitation energies: h␯ϭ650 eV ͑O 1s͒ and 380 eV
27
ˇ
asymmetric Doniach–Sunji c´ functions with Gaussians. The
shape of the asymmetric atomic oxygen peak is determined
from spectra where only this species is present ͓bottom spec-
tra in Figs. 1͑a͒ and 1͑b͔͒. Spectra where only atomic oxygen
and CO on-top species are present ͓topmost spectrum in Fig.
1͑a͔͒ are used to obtain the shape parameters of the on-top
CO peak, leading to a symmetric profile. Keeping these pa-
rameters fixed, from spectra with a large CO coverage ͓top
spectra in Fig. 1͑b͔͒ finally the profile of the CO-bridge con-
tribution is calculated, with a slightly asymmetric shape.
Once determined, the same peak shape parameters have been
used for all spectra, assuming that the peak shape is not
influenced by any coadsorbate interaction. The peak posi-
tions are used as fitting parameters, which leads to the above-
mentioned binding energies. The C 1s spectra are treated
similarly, but here only a linear and no Shirley-background
subtraction leads to the best description of the data.19 In that
case both CO peaks are fitted with asymmetric shapes in a
similar manner as was done for the O 1s spectra. Peak areas
were normalized with respect to the count rate of the clean
surface spectrum, such that any decrease in the photon flux
͑C 1s͒. For binding energy values, see text.
which is in agreement with the reported observation that no
reaction takes place.4 No clear peak corresponding to
bridge CO is observed, as oxygen is known to block the
adsorption of CO on bridge sites.10 For comparison, the CO
uptake at 100 K is also followed by C 1s spectra, which are
shown in Fig. 1͑c͒. Like in the O 1s spectra, one CO-related
peak evolves at a binding energy of 286.74 eV and shifts
towards higher binding energy during uptake. This value is
very similar to the value observed for on-top CO on the clean
,6
1
9,20
surface.
At higher exposures, a small second peak at
1
9
ϳ286.1 eV is found, corresponding to CO on bridge sites.
This small peak amounts to only 0.01–0.02 ML ͑from quan-
titative analysis: see below͒ and was not resolved in the O 1s
spectra, as in the latter the corresponding peak is shadowed
by the asymmetric signal of atomic oxygen. The occupation
of small amounts of bridge sites by CO is attributed to a not
completely filled oxygen p(2ϫ2) adlayer. Vacancies in the
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.159.70.209 On: Wed, 03 Dec 2014 11:56:49

7116
J. Chem. Phys., Vol. 120, No. 15, 15 April 2004
Kinne et al.
͑
resulting from a decrease of the storage ring current͒ is ac-
counted for.
It should be mentioned that in particular for the O 1s
spectra the set of parameters describing the different peaks is
not unique. One can find different sets of parameters, which
would lead to a similar quality of the fit, but would result in
different peak areas of the individual species. Furthermore,
due to photoelectron diffraction ͑PED͒ effects that are not
negligible at the low kinetic energies used, the coverages of
the various adsorbate species cannot be directly calculated
from the peak areas obtained by fitting the XP spectra.
Therefore individual calibration factors were determined for
the three different species. This calibration was done using
structures with known coverage, which are the p(2ϫ2)
structure of atomic oxygen (␪ϭ0.25 ML) and the c(4ϫ2)
structure of CO (␪ϭ0.5 ML; both bridge and on-top sites
are covered with 0.25 ML͒; the latter can be prepared by
dosing 2 L of CO at 200 K and subsequent cooling down to
2
3,28
100 K.
The ratio of peak areas of on-top and bridge
bound species in the O 1s spectra is determined to 1.3 ͑for
h␯ϭ650 eV); this value differs from a ratio of unity, which
would have been expected for equal coverages, without PED
effects. It agrees qualitatively with the observation of
Bj o¨ rneholm et al.,29 although they do not quantify the effect.
The intensity of the oxygen peak for the p(2ϫ2) structure
also leads to a different scaling factor compared to both CO
peaks; the ratio of areas of on-top CO and atomic oxygen
peaks is found to be 0.65. For the quantitative analysis the
peak areas have been corrected using these calibration fac-
tors. In the same way also the peak areas of C 1s spectra
have to be weighted with individual scaling factors, the ratio
of which ͑on-top/bridge͒ has been determined in Ref. 19 has
been determined to 1.26 ͑for h␯ϭ380 eV).
Considering the same quality of fit results for a certain
peak shape parameter range and the PED effects, the error
bars for the coverages denoted below are difficult to esti-
mate. By testing different fit parameters and considering the
reproducibility of the adsorbate structures used for calibra-
tion as well as the experimental geometries ͑e.g., electron
emission angle͒ we estimate the error bars for the absolute
coverages given in this paper to Ϯ8%, which corresponds to
Ϯ0.02 ML for a coverage of 0.25 ML. One should note that
the relative uncertainty within one experimental run ͑e.g.,
uptake or isothermal desorption͒ is smaller. Being aware of
these sources of uncertainty, care has been taken that none of
the results stated in the following are influenced by experi-
mental artifacts.
FIG. 2. Coverage vs time for CO uptake on O-p(2ϫ2)/Pt(111) with pCO
ϭ1.3ϫ10 mbar. CO pressure is switched on at tϭ0. ͑a͒ Tϭ120 K; solid
symbols from C 1s, open symbols from O 1s spectra. ͑b͒ Tϭ295 K; solid
symbols: C 1s, open symbols: O 1s. Bold dashed lines: result after uptake on
clean sample under equal conditions.
Ϫ6
The results of O 1s and C 1s measurements at 120 K are
shown in Fig. 2͑a͒. Immediately after opening the CO beam
at tϭ0, the O 1s data ͑open symbols͒ show a small decrease
of the oxygen coverage by about 0.02 ML. After that, no
significant change in the oxygen signal occurs up to a total
CO dosing time of 160 s ͑ϭ160 L͒. The initial loss in the
oxygen signal is interpreted as a damping or a photoelectron
diffraction effect caused by neighboring CO molecules. A
reaction as an alternative explanation for the decrease in the
oxygen signal seems to be unlikely at a temperature of 120
K; no evidence for an effect like that can be found in the
6,14
literature. After the start of the CO exposure, the coverage
of the on-top CO species immediately jumps to 0.21 ML and
does not increase any further after applying higher doses of
CO. Note that the high sample pressure is the reason that
saturation of the CO on-top coverage had already been
reached before the first spectrum was measured. As already
evident from the raw spectra in Fig. 1͑a͒, the contribution of
bridge-coordinated CO is almost negligible ͑0.01–0.02 ML͒
during the whole uptake. The quantitative analysis of the
corresponding C 1s spectra yields very similar results ͓solid
symbols in Fig. 2͑a͔͒: Upon the start of adsorption, the on-
top CO coverage immediately builds up, whereas the bridge
site occupation is almost negligible ͑0.01 ML͒. Like in the O
1s spectra, no further change is observed both in bridge and
on-top site occupation at higher exposures. The absolute
value of the on-top site occupation is determined to 0.25 ML,
which is somewhat larger than obtained from the O 1s mea-
surements, but within the margins of error mentioned above.
B. Comparison between 120 and 295 K
In the following the results of the quanitative analysis of
XP spectra are presented. Figure 2 shows the coverages of
the different species present on the surface during the ad-
sorption of CO on the O-p(2ϫ2)-precovered surface, as de-
termined both from O 1s and C 1s spectra. The sample tem-
peratures were 120 K ͓Fig. 2͑a͔͒ and 295 K ͓Fig. 2͑b͔͒ and
Ϫ6
the pressure in the CO beam was set to 1.3ϫ10 mbar in
both cases ͑note that for this pressure an exposure time of
1
s corresponds to 1 L͒.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.159.70.209 On: Wed, 03 Dec 2014 11:56:49

J. Chem. Phys., Vol. 120, No. 15, 15 April 2004
Kinetics of CO-oxidation on Pt(III)
7117
By averaging the two values we obtain a saturation coverage
of CO on the O-p(2ϫ2) layer of 0.23 ML. This value is in
very good agreement with the value of 0.25 ML one would
expect for a coadsorbate structure, where each O-p(2ϫ2)
unit cell contains one on-top CO molecule, as suggested,
e.g., in Ref. 8. It is obtained on the surface immediately after
opening the CO beam and no further increase of the CO
coverage up to an exposure of 160 L was observed in the
present study. This behavior is different compared to CO
adsorption on a clean Pt͑111͒ surface, where at the same
temperature on-top coverages significantly above 0.25 ML
are observed at comparable exposures.19 It will be shown
below that for oxygen coverages below 0.25 ML during the
reaction also an increase of CO on-top species to coverages
larger than 0.25 ML can be found. Obviously, oxygen atoms
not only block bridge sites for CO adsorption, but also a part
of the on-top sites available in the absence of O.
oxidation. This channel could be due to a certain amount of
disordered oxygen, which might be present on the surface
even though a sharp LEED pattern has been observed, e.g.,
at the boundaries of ordered p(2ϫ2) domains. As this
amount might vary slightly from one preparation of the oxy-
gen layer to another, the mentioned differences for different
runs are not surprising. Such a reaction between CO and
disordered oxygen was found by Yoshinobu and Kawai14 in
TPD measurements already at a temperature of ϳ225 K. At
Tϭ295 K, one would expect that this reaction is very fast
and thus could cause the rapid change at the beginning of the
CO uptake in our data. After the disordered oxygen species
have reacted to CO2 , only ordered p(2ϫ2) domains of oxy-
gen coadsorbed with CO should be present on the surface.
The reaction of CO with these islands then occurs with a
lower reaction rate, which gives rise to the observed slow
decrease of the O coverage and increase of CO bridge cov-
erage. In other words, the disordered parts in the oxygen
layer act as a starting point of the reaction, which then con-
tinues at the edges of compact oxygen islands, as will be
outlined below.
The finding of a maximum CO coverage of 0.23 ML on
the O-p(2ϫ2) layer is in contrast to the results from
temperature-programmed desorption ͑TPD͒ experiments by
1
3
Kostov et al., where after a dose of 100 L at 120 K a CO
coverage of 0.41 ML was obtained. Similar observations of a
higher CO saturation coverage are reported by Zaera and
From Fig. 2͑b͒ it becomes evident that also the evolution
of the CO on-top species at higher temperatures ͑295 K͒ is
different as compared to low temperatures ͓120 K, Fig. 2͑a͔͒:
After the immediate fast increase of the coverage from zero
to ϳ0.24 ML at the first data point after the start of adsorp-
tion a slow, but significant increase up to 0.30 ML at ϳ120 L
is observed. This behavior is similar to the adsorption of CO
on the clean Pt͑111͒ surface: There, for total coverages above
5
,6
co-workers, with values ranging up to 0.4 ML in the pres-
ence of the O-p(2ϫ2) layer. These differences are too large
to be explained by damping or diffraction effects of photo-
electrons of the coadsorbed CO. A reason for a discrepancy
could be different preparation procedures of the O-p(2ϫ2)
structure: In the cited contributions6,13 the O-p(2ϫ2) layer
0
0
.5 ML, the CO on-top population increases slowly from
.25 ML to ϳ0.30 ML for equal experimental conditions,
was prepared by applying high exposures of oxygen at tem-
peratures above the dissociation temperature of O2 on
Pt͑111͒ which could lead to a different domain structure
compared to our experiments. However, both preparation
methods yield a sharp p(2ϫ2) LEED pattern. Uncertainties
in the calibration procedures could be another possible
source of error.
whereas the bridge site occupation of CO stays almost con-
stant at 0.25 ML.1
9,20,29
The fact that the slow increase of
on-top CO in Fig. 2͑b͒ goes along with the decrease of
atomic O again suggests that oxygen atoms partially block
the on-top adsorption of CO. For comparison, we have indi-
cated the CO bridge and on-top coverages after exposure
onto the clean Pt͑111͒ surface under the same conditions as
used here as dashed lines at the right margin of Fig. 2͑b͒. As
expected, these values coincide with the corresponding cov-
erages after the reaction (tϭ120 s), where no oxygen is left
on the surface.
In Fig. 2͑b͒ the results of O 1s ͑open symbols͒ and C 1s
͑solid symbols͒ measurements at 295 K are shown—i.e., at a
temperature at which a reaction between O and CO occurs.
Within the margin of error, the CO coverages obtained from
the two data sets are in very good agreement with each other.
Overall, a steady decrease of the oxygen coverage is ob-
served that is accompanied by a very fast increase in CO
on-top coverage, similar to the situation at 120 K, and a
much slower increase of the CO bridge coverage. A closer
look at the data shows that there is a sudden drop of the
oxygen signal ͑by ϳ0.05 ML͒ from its starting value to the
first data point after tϭ0, followed by a smoothly declining
curve. This effect is always observed in our experiments be-
tween 275 and 305 K, with the magnitude of the drop vary-
ing between 0.03 and 0.06 ML. An inverse behavior is ob-
served for the CO population on bridge sites: after the
molecular beam has been switched on, the signal suddenly
rises and thereafter increases smoothly with a slope compa-
rable to the negative slope of the oxygen curve. The quantity
of this initial rise is comparable with the drop of the oxygen
coverage: i.e., it lies between 0.03 and 0.06 ML. The sudden
and opposite intensity changes of the O and CO bridge sig-
nals indicate the existence of a fast reaction channel for CO
C. Pressure dependence of the reaction rate
To simplify the determination of the rate constant of CO
oxidation from our experiments it is helpful to exclude the
influence of other rate limiting factors, such as the flux of the
incoming CO molecules or diffusion processes on a not fully
occupied surface. To rule out these effects, a CO pressure
should be used, where the impingement rate of CO is much
higher than the reaction rate, as has been done in the above
experiments. As a consequence, one would expect that
bridge sites and additional on-top sites, which after the re-
moval of oxygen become available for CO adsorption, are
immediately filled and the CO surface population is satu-
rated. Thus the total speed of the reaction, which is a func-
tion of the CO coverage, should reach a saturation value for
high pressures. In the following, both the saturation of the
CO surface population as well as the reaction rate are
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.159.70.209 On: Wed, 03 Dec 2014 11:56:49

7118
J. Chem. Phys., Vol. 120, No. 15, 15 April 2004
Kinne et al.
FIG. 3. Reaction rate as determined from the negative average slope of the
oxygen coverage for 0.05 MLϽ␪ Ͻ0.15 ML during reaction (Tϭ295 K)
O
vs CO beam pressure ͑solid circles, left axis͒. Average total coverage (CO
ϩO) for 0.05 MLϽ␪ Ͻ0.15 ML in the same experiment ͑open squares,
FIG. 4. Sum of the atomic oxygen and bridge CO coverage vs atomic
oxygen coverage during reaction at 295 K for different CO pressures. Dotted
lines: interpolating splines, as guides to the eye. The solid line indicates the
bisector of the graph; the scale of the left axis does not start at zero and
therefore the bisecting line does not run through the lower left corner of the
graph.
O
right axis͒.
checked for a sample temperature of 295 K and different
molecular beam pressures by the same type of experiment as
described above ͓Fig. 2͑b͔͒, starting out with an O-p(2ϫ2)
layer.
ϭ0.10–0.15 ML] from a constant value of 0.25 ML, which
is expected if one oxygen atom blocks the adsorption on one
bridge site and free sites are efficiently filled by CO. The
reason for the smaller value is most likely that at the edges of
the islands not every O can be replaced by CO molecules,
which is expected to be most relevant for intermediate O
coverages, as is observed. Concerning the low-pressure ex-
periments, at the beginning of the reaction ͑i.e., for high
oxygen coverage͒ the data points coincide with the bisector
of the graph ͑solid line in Fig. 4͒, which means that at first
almost no CO is found on bridge sites. Later, the data points
diverge from the bisector and finally come near the value of
As a measure of the reaction rate, the slope of the oxy-
gen signal between 0.05 and 0.15 ML is displayed in Fig. 3
as a function of the gas phase pressure ͑solid circles, left
axis͒. Clearly, a saturation effect is observed for pressures
Ϫ7
above ϳ9ϫ10 mbar, indicating that the pressure acces-
sible by the molecular beam is indeed high enough that the
reaction rate is not limited by the adsorption of CO. Note
that even for the lowest pressure the impingement rate of CO
would nominally exceed the reaction rate. In the limit of zero
pressure the reaction rate must go to zero.
To check the saturation of the CO population with in-
creasing pressure, we investigate the behavior of the
CO bridgeϩO coverage as well as the total COϩO cover-
age. Both values should saturate for a sufficiently high pres-
sure. On the other hand, at lower pressures, for which the
impingement rate becomes comparable with the reaction
rate, not all vacant sites can be immediately populated by CO
and these coverages should drop after the start of the reac-
tion. In Fig. 3, the averaged total COϩO population between
0
.25 ML at the end of the reaction ͑i.e., for zero oxygen
coverage͒, as expected.
Figure 3 shows that saturation of CO coverage and reac-
Ϫ7
tion rate occur at different pressures—i.e., 2.5ϫ10 and 9
Ϫ7
ϫ10 mbar, respectively. The reason for this difference
could be a repulsive interaction of oxygen and CO, as pro-
8
posed by V o¨ lkening and Wintterlin: Since sites in the vicin-
ity of oxygen atoms are the reactive ones, but simultaneously
are energetically unfavorable for the adsorption of CO, the
surface can be almost saturated with CO and still a consid-
erable part of the reactive sites would be vacant. This is
especially true if the number of reactive sites is small com-
pared to the number of unreactive adsorption sites for CO as
would be the case if the reaction were to occur only at the
edges of large oxygen islands. As small differences ͑р0.01
ML͒ in occupation are not detectable in XPS, it might appear
as if the surface is already saturated at a certain pressure,
while the reaction speed still increases with pressure.
0
.05 MLϽ␪ Ͻ0.15 ML ͑with ␪O being the atomic oxygen
O
coverage͒ is included as open circles. The data clearly show
the saturation effect starting at pressure of 2.5
a
Ϫ7
ϫ10 mbar, which is significantly lower than for the reac-
tion speed ͑see above͒. Within the margin of error an identi-
cal behavior is encountered if the sum of bridge-CO and
oxygen is analyzed ͑data not separately shown in Fig. 3͒. In
Fig. 4 the corresponding data are shown as a function of
oxygen coverage for the different pressures. Note that for
clarity of the presentation the scale of the left axis in the
graph is expanded and does not start at zero coverage and,
therefore, the bisecting line does not run through the lower
left corner of the graph. Clearly, the data for low pressures
D. Temperature dependence of the reaction rate
In order to determine the kinetic parameters of the CO
oxidation, isothermal reaction experiments have been per-
formed for different temperatures between 275 and 305 K,
͑
open and solid circles͒ lie below the data for pressures at
Ϫ7
and above 2.5ϫ10 mbar, which are identical within the
scatter of Ϯ0.01 ML. Furthermore, for high pressures the
sum shows only small deviations ͓(0.025Ϯ0.01) ML at ␪O
Ϫ6
using a CO pressure of 1.3ϫ10 mbar. The experimental
values of the oxygen coverage are displayed in Fig. 5 as a
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.159.70.209 On: Wed, 03 Dec 2014 11:56:49

J. Chem. Phys., Vol. 120, No. 15, 15 April 2004
Kinetics of CO-oxidation on Pt(III)
7119
FIG. 6. Ratio k*ϭ␪O /(␪O␪CO) as a function of time during reaction at T
˙
Ϫ6
ϭ300 K. Dashed line: CO beam (pCOϭ1.3ϫ10 mbar) impinging on the
FIG. 5. Coverage of atomic oxygen vs time during reaction for different
temperatures. pCOϭ1.3ϫ10 mbar in all cases. Solid lines: fits of the
function given in Eq. ͑6͒. The time axis has been shifted so that the fit
sample during the whole reaction. Solid line: only a short ͑3 s͒ CO pulse
Ϫ6
(pCOϭ2ϫ10Ϫ6 mbar) is applied at the beginning. ␪O has been calculated
˙
after smoothing the data curves by spline functions.
curves coincide in the point (tϭ0,␪ ϭ0.16 ML) ͑see text͒.
O
function of time. In the graph, the time scale for each experi-
ment is shifted by an offset value, such that at tϭ0 all curves
one where the O-p(2ϫ2)-covered sample is exposed only to
a short, intense pulse ͑duration 3 s, pϭ2.0ϫ10 mbar) of
Ϫ6
coincide in the point ␪ ϭ0.16 ML, which was done for clar-
O
˙
CO at the beginning. The ratio ␪/␪ ␪ for both choices is
O
CO
ity of presentation because of the following reason: The first
data points in each run show that despite equal preparation
procedures of the p(2ϫ2) atomic oxygen layer the observed
coverages exhibit some variation. Additionally, since the ini-
tial drop of the oxygen coverage also differs slightly from
one experiment to another ͑see Sec. III B͒, the starting point
of the smoothly decreasing part of the reaction curve differs
between the experiments, but in any case is found for cover-
ages larger than 0.16 ML. Therefore, taking 0.16 ML as the
starting coverage ensures that the initial reaction of disor-
dered oxygen does not affect the following results.
presented in Fig. 6 as a function of time, clearly showing that
the values in the case of the pulse experiment are lower by a
factor of 3–13 compared to steady CO exposure. Obviously,
k* is not a constant if it is determined in that way; it varies
also significantly within the high-pressure experiment, as
7
was already pointed out by Wintterlin et al. One interpreta-
tion could be that k* is strongly coverage dependent, as was
6
suggested by Zaera et al. Alternatively, the simple assump-
tion of an isotropic distribution of adsorbates could be
wrong. In the last section we could show that the existence
of oxygen islands could explain the pressure dependence of
the reaction speed if the reaction is restricted to the island
edges. The latter assumption also provides a simple explana-
tion for the difference in speed between the measurements in
Fig. 6: As it is known that CO can adsorb inside the O-p(2
ϫ2) structure, the possibility of a reaction within this struc-
ture would be in contradiction with the observed slow reac-
tion rate after the CO pulse. Within this model Eq. ͑1͒ can be
written as
For quantitative analysis of the data in Fig. 5 an appro-
priate rate equation has to be found, which in the most gen-
eral case for a Langmuir–Hinshelwood reaction30 is of the
form
d␪ /dtϭϪkf͑␪ ,␪CO͒,
͑1͒
O
O
where ␪O and ␪CO are the coverages of atomic oxygen and
CO, respectively, and k is the reaction rate constant. The
function f not only depends on the given coverages, but also
on the local arrangement of the adsorbates on the surface. A
straightforward choice of f is a simple first-order dependence
␣
O
d␪ /dtϭϪkc␪ g͑␪CO͒,
͑3͒
O
␣
O
where c␪ ϭ␪ is the number of oxygen atoms at island
O,edge
on ␪O and ␪ , assuming an isotropic distribution of
CO
7
edges per substrate atom, with ␣ϭ0.5 for smooth ͑nonfrac-
tal͒ edges and g a still unknown function depending on ␪CO
molecules —i.e.,
.
d␪ /dtϭϪk*␪ ␪
.
͑2͒
O
O CO
In the simple case where all oxygen islands have the same
size, ␪O,edge is given by
˙
By calculating the ratio ␪ /␪ ␪ , the value of k* can
O
O CO
be obtained during the whole reaction. To do that the deriva-
␪O,edgeϭaz N /z/N ϭa zN /N ,
ͱ ͱ
O m O m
˙
tive ␪ has to be calculated numerically, which due to statis-
O
where Nm is the number of substrate atoms, z the total num-
ber of islands, and NO the total number of oxygen atoms. a
depends on the shape of the islands; in the case of circular
tical noise is only possible after smoothing the data curves
with interpolating spline functions. If k* is assumed to be
independent of the coverages, Eq. ͑2͒ should hold for differ-
ent surface coverages during the reaction—i.e., for different
CO pressures. In order to check this, two extreme cases are
compared: an experiment with a CO pressure of 1.3
islands, aϭ2 ␲. Setting N ϭ␪ N , this yields
ͱ
O O m
␪O,edgeϭa
ͱ
z␪ /
ͱ
Nm
O
Ϫ6
ϫ10 mbar present during the whole reaction and another
and leads by comparison finally to
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.159.70.209 On: Wed, 03 Dec 2014 11:56:49

7120
J. Chem. Phys., Vol. 120, No. 15, 15 April 2004
Kinne et al.
cϭa
ͱ
z/Nm.
͑4͒
This means in other words that without knowledge of the
island size and shape it is only possible to determine the
product kЈϭkc from our measurement. The function g(␪CO
)
in Eq. ͑3͒ accounts for the density of CO molecules on reac-
tive sites—i.e., at the edge of oxygen islands. In the most
general case it depends on the reaction rate, the incoming
flux of CO particles, the diffusion constant of adsorbates, and
interaction energies between adsorbates. Fortunately, in the
high-pressure limit it becomes greatly simplified: then all
reactive sites are always fully populated and g reduces to
unity. In the last paragraph we could show that this limit is
fulfilled in our experiments. Finally, the equation
␣
O
d␪ /dtϭϪkЈ␪
͑5͒
O
would be suitable to evaluate our isothermal reaction data,
but would involve numerical calculation of the derivative
and smoothing of the data curves, as mentioned above.
Therefore, it seems better to use the integral form of Eq.
FIG. 7. Arrhenius plot of kЈ as determined from the fits in Fig. 5 vs inverse
temperature. Solid line: exponential fit.
͑
5͒—that is,
At this point the error bars of these numbers have to be
discussed. When comparing the reaction rates for 295 K in
Figs. 2͑b͒ and 5, one finds a difference of a factor of 2 in
these rates. There are two possible reasons for this differ-
ence. It can be attributed to an incorrect calibration of the
temperature scale due to a bad connection of the thermo-
couple to the sample: All data in Figs. 2, 3, 4, and 8 were
collected in one experimental period, whereas Figs. 5 and 6
are taken from another period. Between both periods the
sample and thermocouples had been remounted. If the tem-
perature deviation were 10 K, the data of the two periods
would be in very good agreement with each other. A second
source of the observed difference could be a different num-
ber of oxygen islands, z, due to different surface quality.
According to Eq. ͑4͒, c is proportional to z1 and thus a
change in the number of oxygen islands by a factor of 4
would lead to a difference in the reaction rate by a factor of
1
Ϫ␣
Ϫ͑1Ϫ␣͒kЈ͑tϪt ͔͒1/͑1Ϫ␣͒,
␪Oϭ͓␪
͉
͑6͒
O tϭt
0
0
which can be fitted to the curves in Fig. 6͑a͒. As mentioned
above, ␪ is chosen to be 0.16 ML and t is the time at
͉
O tϭt
0
0
which this coverage is observed. Data points near the sudden
initial decrease of the oxygen signal are thereby neglected in
the following analysis.
In order to check if the assumption of compact islands—
i.e., ␣ϭ0.5—can be justified, first a fit of Eq. ͑6͒ with three
variable parameters—namely, ␣, kЈ, and t —is performed,
yielding a mean value of
0
␣
ϭ0.63Ϯ0.15.
/2
Within the given statistical error margin this result is consis-
tent with the hypothesis of a one-dimensional reaction front.
The large uncertainty indicates that the shape of the curve
shows a relatively weak dependence on ␣. This value is com-
parable with the one of 0.55 given by V o¨ lkening and
2
, as is observed. These uncertainties, either in the tempera-
ture measurements or in sample preparation, contribute to the
error bars given above.
8
Wintterlin obtained from STM observations. In a second fit
As c is not known (cӶ1 for extended islands͒, in any
case ␯ alone cannot be determined. From the STM work of
V o¨ lkening and Wintterlin ͓Fig. 1͑a͒ of Ref. 8͔ we can esti-
procedure ␣ is held fixed at 0.63 and kЈ and t are deter-
mined again, resulting in the curves displayed in Fig. 5 to-
gether with the data points. Note that on the time axis (t
0
mate that c is of the order of 0.05, so by simply taking the
slope of the Arrhenius plot one would underestimate ␯ by a
Ϫt ) rather than t is used, causing all curves to coincide in
0
the point (tϭ0,␪ ϭ0.16 ML). Now kЈ can be used to ex-
O
8
factor of ϳ20. In this STM study an activation energy of
tract the activation energy E of the reaction by assuming an
a
7
Ϫ1
0
.49 V and a prefactor of 1.6ϫ10 s
are obtained in a
Arrhenius-type behavior of the reaction rate—i.e.,
temperature region from 237 to 274 K, in good agreement
with our observation. Within their quoted uncertainty, Zaera
kЈϭc␯ exp͑ϪE /k T͒,
͑7͒
a
B
6
et al. found comparable values of E ϭ0.39 eV and ␯ϭ2.5
where ␯ is the preexponential factor and c the parameter
defined in Eq. ͑4͒. Figure 7 shows kЈ as a function of the
inverse temperature as well as a fit according to Eq. ͑7͒,
which results in the parameters
a
5
Ϫ1
ϫ10 s in the limit of zero coverage from their molecular
beam study in the range of 300–400 K. In this case, how-
ever, an incorrect assumption of an isotropic distribution of
adsorbates has been used to model the coverage-dependent
data. From a theoretical point of view, Eichler31 obtained
from density functional theory calculations an activation
E ϭ͑0.53Ϯ0.04͒ eV,
a
c␯ϭ4.7ϫ106Ϯ0.7 sϪ1
.
energy of 0.74 eV and a preexponential factor of 5
Ϫ1
Using a fixed value of ␣ϭ0.5 for the fit yields results that
ϫ1012
s
, values which differ substantially from our and
6
Ϫ1
are within the given error bars ͑0.52 eV, 2.4ϫ10 s ).
other experimental data. The very high prefactor leads to the
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.159.70.209 On: Wed, 03 Dec 2014 11:56:49

J. Chem. Phys., Vol. 120, No. 15, 15 April 2004
Kinetics of CO-oxidation on Pt(III)
7121
Eq. ͑7͒ can be used to calculate the change of oxygen cov-
erage during the temperature increase. Instead of a constant
temperature, a linear behavior T(t)ϭT ϩ␤t ͑with ␤ being
0
the heating rate and T the starting temperature͒ is used and
0
Eq. ͑5͒ is solved numerically. With a starting coverage of
0
.17 ML, corresponding to the coverage of ordered oxygen
regions, one observes a function which is shifted upwards by
0 K as compared to the data points. This discrepancy can
1
easily be explained by either the above-mentioned uncer-
tainty in the temperature scale ͑Figs. 5 and 8 stem from
different measurement periods͒ or by different surface qual-
ity resulting in reaction rates different by a factor of 2. If the
simulated curve is shifted down by 10 K, one obtains the
solid line in Fig. 8, which above 250 K fits the data points
very well.
FIG. 8. Coverages of CO and oxygen from O 1s data during temperature-
programmed reaction ͑symbols͒. CO pressure was set to 1.3ϫ10 mbar
during the whole experiment; a heating rate of 0.1 K/s was applied, starting
at 120 K. Lines: calculations without ͑solid line͒ and with ͑dotted line͒
contribution of disordered oxygen ͑see text for details͒.
To also describe the behavior below 250 K, rate equation
Ϫ6
͑5͒ can be extended by a term, which accounts for the reac-
tion in disordered regions. Thereby one assumes that the
oxygen coverage in ordered and disordered regions can be
treated independently. The ordered part is still described by
d
O
Eq. ͑5͒, whereas the coverage of disordered oxygen, ␪ , can
be modeled by
prediction of a reaction rate which is higher than the experi-
mentally observed one.
d
O
d
d
a
d
O
d␪ /dtϭϪ␯ exp͑ϪE /k T͒␪ ,
͑8͒
B
where Ea and ␯d are the activation energy and prefactor
d
E. Temperature-programmed reaction
corresponding to disordered oxygen. This first-order depen-
To obtain further insight into the process of CO oxida-
tion, we performed a temperature-programmed reaction ex-
d
dence on ␪ assumes that all oxygen atoms in the disordered
O
areas are reactive and are surrounded by CO molecules as
reaction partners. Note that with the present data basis this
assumption is somehow arbitrary, but is the simplest choice
which still can describe the experiment. Assuming an initial
coverage of 0.22 ML and an identical preexponential factor
periment: The O-p(2ϫ2)-precovered sample is cooled to
Ϫ6
1
20 K and is exposed to 1.3ϫ10 mbar CO from the mo-
lecular beam. After 120 s, the sample temperature is in-
creased linearly with a heating rate of 0.1 K/s, while O 1s
spectra are taken approximately every 6 K. The CO impinge-
ment rate was the same during the whole experiment. The
resulting O and CO coverages are presented in Fig. 8. Up to
d
as for the reaction with oxygen islands—i.e., ␯ ϭ␯ ͑with the
estimated value of cϭ0.05)—a reasonable fit to the experi-
mental data above 180 K ͑dotted line in Fig. 8͒ can be ob-
200 K only a slow decline of the oxygen signal of about 0.02
d
a
tained with E ϭ0.48 eV. Note that also in this case the tem-
ML is observed, while the amount of on-top CO increases
approximately by the same amount. As a thermally activated
reaction seems unlikely for that temperature, photon-
stimulated reaction or desorption could serve as an interpre-
tation. Vacant sites are effectively filled by CO molecules
due to the high CO pressure. Above 200 K a steeper decrease
perature scale of the simulated curve is shifted down by 10
d
a
K. From the obtained value for E , the calculated rate con-
stant at 300 K for the reaction of disordered oxygen—e.g.,
d
O
for ␪ ϭ0.05 ML—would be higher by a factor of approxi-
mately 20 compared to the rate of oxygen in islands at ␪O
ϭ0.17 ML. Even considering the large error margins pos-
sible in the evaluation, always a considerably higher rate
constant is found.
of ␪ follows, with a simultaneous increase of CO on on-top
O
and bridge sites. This effect cannot be explained by the ki-
netic parameters of the reaction at oxygen islands that have
been derived in Sec. III D: at the given time scale they would
predict a negligible reaction rate below ϳ250 K. However, it
fits to the fast reaction channel at the start of isothermal
experiments discussed in Sec. III C ͑between 275 and 305
K͒, which has been explained by a reaction of CO with dis-
ordered oxygen atoms. The amount of oxygen involved in
this reaction was determined to be 0.03–0.06 ML, which is
in agreement with the decrease of ␪O by 0.05 ML between
IV. SUMMARY
We have examined the coadsorption and reaction of CO
with atomic oxygen on Pt͑111͒ by in situ high-resolution
core level spectroscopy using synchrotron radiation. A p(2
ϫ2) atomic oxygen layer with a coverage of 0.25 ML was
exposed to CO from a supersonic molecular beam at various
sample temperatures and pressures. O 1s spectra and in se-
lected cases C 1s spectra have been used to determine the
surface coverage of the adsorbed species.
At a temperature of 120 K or below, no reaction takes
place between the preadsorbed oxygen and adsorbed CO; in
accordance with the literature, bridge sites are blocked for
CO adsorption in the presence of the O-p(2ϫ2) adlayer. For
200 and 250 K in the temperature-programmed reaction in
Fig. 8. Assuming that the reaction of disordered oxygen oc-
curs within that temperature window, it seems plausible, that
at 275 K and above this effect would happen faster than the
temporal resolution of our XPS measurement.
In order to compare the temperature-programmed ex-
periment with the isothermal reaction, Eq. ͑5͒ together with
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.159.70.209 On: Wed, 03 Dec 2014 11:56:49

7122
J. Chem. Phys., Vol. 120, No. 15, 15 April 2004
Kinne et al.
CO on-top sites, a saturation coverage of (0.23Ϯ0.02) ML is
observed at 160 L. Compared to CO on clean Pt͑111͒, this
implies that also part of the on-top sites is blocked by the
presence of atomic oxygen. This finding stands in contrast to
the TPD results of Kostov et al.,13 where after 100 L already
a CO coverage of 0.41 ML on the O-p(2ϫ2) adlayer is
reported. The origin of this discrepancy remains unclear.
The kinetics of the reaction were examined for sample
temperatures between 275 and 305 K. Immediately after the
start of the CO exposure at high pressure (1.3
of the experiment. This supports the evidence of a higher
reaction rate of disordered oxygen as compared to compact
islands.
ACKNOWLEDGMENTS
This work was supported by the Deutsche Forschungs-
gemeinschaft through Grant No. STE 620/4-2. We would
like to thank the BESSY staff for technical support.
1
R. L. Palmer and J. N. Smith, J. Chem. Phys. 60, 1453 ͑1974͒.
T. Engel and G. Ertl, Adv. Catal. 28, 1 ͑1979͒.
T. Matshushima, D. B. Almy, and J. M. White, Surf. Sci. 67, 89 ͑1977͒.
J. L. Gland and E. B. Kollin, J. Chem. Phys. 78, 963 ͑1983͒.
M. D. Xu, J. Y. Liu, and F. Zaera, J. Chem. Phys. 104, 8825 ͑1996͒.
F. Zaera, J. Y. Liu, and M. D. Xu, J. Chem. Phys. 106, 4204 ͑1997͒.
2
3
Ϫ6
ϫ10 mbar), a sudden drop of the oxygen coverage and,
4
simultaneously, a fast increase of CO on bridge sites have
been observed, with both changes amounting to 0.03–0.06
ML. This process is attributed to the reaction of CO with
disordered oxygen atoms, which at the temperatures studied
has a higher reaction rate than the reaction of CO with or-
dered p(2ϫ2) oxygen islands. The disordered areas, which
are present although a sharp p(2ϫ2) LEED pattern is ob-
served, act as a starting point of the reaction.
5
6
7
J. Wintterlin, S. Vo
¨
lkening, T. V. W. Janssens, T. Zambelli, and G. Ertl,
Science 278, 1931 ͑1997͒.
S. V o¨ lkening and J. Wintterlin, J. Chem. Phys. 114, 6382 ͑2001͒.
D. Kulginov, M. Persson, C. Akerlund, I. Zoric, and B. Kasemo, J. Vac.
8
9
˚
Sci. Technol. A 13, 1511 ͑1995͒.
10
J. L. Gland and E. B. Kollin, Surf. Sci. 151, 260 ͑1985͒.
S. Akhter and J. M. White, Surf. Sci. 171, 527 ͑1986͒.
K. Bleakley and P. Hu, J. Am. Chem. Soc. 121, 7644 ͑1999͒.
K. L. Kostov, P. Jakob, and D. Menzel, Surf. Sci. 377–379, 802 ͑1997͒.
J. Yoshinobu and M. Kawai, J. Chem. Phys. 103, 3220 ͑1995͒.
A. Baraldi, G. Comelli, S. Lizzit, M. Kiskinova, and G. Paolucci, Surf.
Sci. Rep. 49, 169 ͑2003͒.
A. Baraldi, S. Lizzit, D. Cocco, G. Comelli, G. Paolucci, M. Kiskinova,
and R. Rosei, Surf. Sci. 385, 376 ͑1997͒.
I. Z. Jones, A. Bennet, and M. Bowker, Surf. Sci. 439, 235 ͑1999͒.
R. Denecke, M. Kinne, C. M. Whelan, and H. P. Steinr u¨ ck, Surf. Rev.
1
1
12
By investigating the reaction behavior for different pres-
sures, it was found that around 300 K the reaction rate satu-
13
14
Ϫ7
15
rates for CO pressures higher than ϳ9ϫ10 mbar, while it
decreases if the pressure is reduced. From the data collected
under different experimental conditions, a reaction of the ad-
sorbed CO molecules within the ordered oxygen p(2ϫ2)
islands—i.e., a first-order reaction kinetics—can be ex-
cluded. From our measurements for sufficiently high pres-
sure ͑where the rate is independent of pressure͒, we obtain a
reaction order ␣ with respect to oxygen of 0.63Ϯ0.15. This
value is in good agreement with a previous STM study by
16
1
1
7
8
Lett. 9, 797 ͑2002͒.
M. Kinne, T. Fuhrmann, C. M. Whelan, J. F. Zhu, J. Pantf o¨ rder, M. Probst,
19
G. Held, R. Denecke, and H. P. Steinruck, J. Chem. Phys. 117, 10 852
͑2002͒.
¨
2
2
0
1
M. Kinne, Ph.D. thesis, Universit a¨ t Erlangen-N u¨ rnberg, 2004.
S. Kneitz, J. Gemeinhardt, H. Koschel, G. Held, and H. P. Steinru
¨
ck, Surf.
7
Wintterlin et al. at somewhat lower temperatures and can be
Sci. 435, 27 ͑1999͒.
2
2
2
2
3
4
explained if the reaction between O and CO is restricted to
the edges of extended ordered oxygen p(2ϫ2) islands—in
that case one would expect an ideal value of ␣ϭ0.5. From
an Arrhenius plot of the rate constant between 275 and 305
K, an activation energy of (0.53Ϯ0.04) eV and a prefactor
J. L. Gland, B. A. Sexton, and G. B. Fisher, Surf. Sci. 95, 587 ͑1980͒.
H. Steininger, S. Lehwald, and H. Ibach, Surf. Sci. 123, 1 ͑1982͒.
N. Materer, U. Starke, A. Barbieri, R. D o¨ ll, K. Heinz, M. A. Van Hove,
and G. A. Somorjai, Surf. Sci. 325, 207 ͑1995͒.
C. Puglia, A. Nilsson, B. Hernn a¨ s, O. Karis, P. Bennich, N. Materer, and
N. M a˚ rtensson, Surf. Sci. 342, 119 ͑1995͒.
25
26
6
Ϯ0.7 Ϫ1
of 4.7ϫ10
s
are determined. These values are again
D. A. Shirley, Phys. Rev. B 5, 4709 ͑1972͒.
2
2
7
8
ˇ
7
S. Doniach and M. Sunjic, J. Phys. 3, 285 ͑1970͒.
in good agreement with the results of Wintterlin et al.
The obtained parameters were checked by following the
reaction in a temperature-programmed way and comparing
the calculated change of oxygen coverage with the data
points. If, additionally to the reaction rate of oxygen islands,
a rate law for the reaction of disordered regions of oxygen is
taken into account, the simulation leads to a good description
H. Froitzheim, H. Hopster, H. Ibach, and S. Lehwald, Appl. Phys. 13, 147
͑
1977͒; I. Zasada and M. A. Van Hove, Surf. Rev. Lett. 7, 15 ͑2000͒; F.
Bondino, G. Comelli, F. Esch, A. Locatelli, A. Baraldi, S. Lizzit, G. Pa-
olucci, and R. Rosei, Surf. Sci. 459, L467 ͑2000͒.
O. Bj o¨ rneholm, A. Nilsson, H. Tillborg, P. Bennich, A. Sandell, B. Hern-
n a¨ s, C. Puglia, and N. M a˚ rtensson, Surf. Sci. 315, L983 ͑1994͒.
T. Engel and G. Ertl, J. Chem. Phys. 69, 1267 ͑1978͒.
A. Eichler, Surf. Sci. 498, 314 ͑2002͒.
2
9
30
31
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
30.159.70.209 On: Wed, 03 Dec 2014 11:56:49
1