Oxide-free oxygen incorporation into Ru(0001).

Abstract

A smooth Ru(0001) surface prepared under ultra-high vacuum conditions has been loaded with oxygen under high-pressure (p approximately 1 bar) and low-temperature (T < 600 K) conditions. Oxygen phases created in this way have been investigated by means of thermal desorption spectroscopy, low-energy electron diffraction, and ultraviolet photoelectron spectroscopy. The exposure procedures applied lead to oxygen incorporation into the subsurface region without creation of RuO2 domains. For oxygen exposures ranging from 10(11) to 10(14) L oxygen contents up to about 4 monolayer equivalent could be achieved. The oxygen incorporation is thermally activated. The CO oxidation reaction conducted at mild temperatures (T < 500 K) at a sample loaded with subsurface oxygen reaches CO --> CO2 conversion probabilities of 10(-3).

Full text

Page 1

Oxide-free oxygen incorporation into Ru„0001…
Raoul Blume and Horst Niehus
Institut fu ¨r Physik der Humboldt-Universita ¨t, Invalidenstr. 110, 10115 Berlin, Germany
Horst Conrad
Fritz-Haber-Institut der Max-Planck-Gesellschaft, Faradayweg 4-6, 14195 Berlin, Germany
Artur Bo ¨ttchera)
Inst. fu ¨r Physikalische Chemie, Univ. Karlsruhe, Kaiserstr. 12, 76131 Karlsruhe, Germany
?Received 8 January 2003; accepted 1 December 2003?
A smooth Ru?0001? surface prepared under ultra-high vacuum conditions has been loaded with
oxygen under high-pressure (p?1 bar) and low-temperature (T?600K) conditions. Oxygen
phases created in this way have been investigated by means of thermal desorption spectroscopy,
low-energy electron diffraction, and ultraviolet photoelectron spectroscopy. The exposure
procedures applied lead to oxygen incorporation into the subsurface region without creation of
RuO2domains. For oxygen exposures ranging from 1011to 1014L oxygen contents up to about 4
monolayer equivalent could be achieved. The oxygen incorporation is thermally activated. The CO
oxidation reaction conducted at mild temperatures (T?500K) at a sample loaded with subsurface
oxygen reaches CO→CO2conversion probabilities of 10?3.
Physics. ?DOI: 10.1063/1.1643724?
© 2004 American Institute of
I. INTRODUCTION
Oxidation procedures applied to metal surfaces result at
intermediate stages in the formation of extended, thin oxide
domains covering the surface. The physical and chemical
properties of such metal oxide layers are rather well known.1
The morphology of those oxides acting as terminating layers
of the metallic substrate underneath is element specific and
depends to a large extent on the oxidation conditions applied
?partial oxygen pressure p, substrate temperature T, surface
roughness ?, etc.?.2,3Usually, the crystallographic structure
of those oxide layers is very similar to the structure of the
corresponding solid oxides grown by applying conventional
methods of crystal growth.4Various spectroscopic methods
used to monitor the electronic properties of the surface ox-
ides also reveal overall agreement with those of the well-
known bulk oxides.
However, the very initial as well as almost all interme-
diate stages of the oxidation process are much less under-
stood. The importance of these transient states becomes evi-
dent if the incorporation of oxygen into the metal surface is
considered as the initial step for the restructuring of the
original metal lattice into the different crystallographic struc-
ture of the oxide lattice. In the case of Ru?0001?, the impor-
tance of the initial and intermediate oxidation stages became
evident when discovering that these phases, which consist
mostly of oxygen incorporated below the topmost layer, ex-
hibit extraordinarily high reactivity towards the CO oxida-
tion reaction.5–7A very efficient way to augment the amount
of incorporated oxygen has been found via applying oxida-
tion procedures at elevated substrate temperatures and high
partial pressures (p?10?2mbar). Such treatments result in
the formation of rather complex lateral structures, dominated
in the later stages by RuO2domains.3However, up to now,
neither the selective preparation nor the experimental char-
acterization of the particular intermediates preceding the for-
mation of a compact RuO2surface layer has been achieved
in detail.
In the following, we report on the oxygen incorporation
into the Ru lattice through the ?0001? surface at experimental
conditions where the oxide formation is essentially inhibited.
Such a particular state was achieved by exposing the clean
crystal ?cleaned under UHV-conditions? at high oxygen pres-
sures (p?1 bar) and low sample temperatures, T?600K.
Oxygen incorporation achieved with this treatment re-
sembles strongly those oxygen dissolution processes, which
were observed previously for various polycrystalline solids.8
Sample cleaning and all high-pressure experiments were
performed in an ultra high vacuum chamber ?UHV?. Use of
the established UHV procedures ?cycles of Ar ion sputtering,
heating at base pressure better than 10?9mbar) provides the
high level standards of cleanliness and smoothness of the
single-crystal surface.9Oxygen incorporation has been stud-
ied mainly by thermal desorption spectroscopy ?TDS?, which
allows to monitor the thermally driven removal of the incor-
porated oxygen and to quantify the oxygen load. In addition,
the electronic properties of the incorporated oxygen have
been characterized by means of UPS ?21.2 eV?. The tech-
niques applied reveal physical properties, which allow to
clearly differentiate between the oxygen incorporated and
that bound in RuO2domains. Moreover, incorporated oxy-
gen is characterized by its high reactivity towards the CO
oxidation reaction in the low temperature region (T
?550K).
a?Author to whom correspondence should be addressed. Telephone:
??0721-608-3254;Electronic
karlsruhe.de
mail: artur.boettcher@chemie.uni-
JOURNAL OF CHEMICAL PHYSICSVOLUME 120, NUMBER 822 FEBRUARY 2004
38710021-9606/2004/120(8)/3871/9/$22.00© 2004 American Institute of Physics
Downloaded 23 Feb 2004 to 129.13.167.7. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp

Page 2

II. EXPERIMENT
The experiments were performed in a UHV chamber at a
base pressure below 5?10?10mbar. The system was
equipped with standard facilities for crystal cleaning and
characterization ?LEED, Auger electron spectroscopy ?AES?,
UPS, and TDS?. The UHV chamber has been specially de-
signed for oxygen exposures with a partial pressure up to 1
bar, allowing to pump off the oxygen and to achieve UHV
conditions (p?10?9mbar) within about 15 min after expo-
sure.
The Ru crystal has been cleaned by means of Ar?sput-
tering and subsequent annealing according to established
procedures.9,10Argon sputtering with 2 kV ions ?3–10 ?A?
was applied for 10 minutes. The sample was subsequently
heated two times up to 1550 K. Surface cleanliness was con-
trolled by AES and UPS ?21.2 eV?. Surface roughness was
regularly monitored by using thermal energy diffuse He scat-
tering, thermal desorption spectroscopy ?TEAS?.11The
preparation cycles were repeated until the specularly scat-
tered He flux reached a maximum. Further temperature treat-
ment effected no more changes and thus indicated the high-
est surface smoothness achievable. However, even then we
cannot entirely exclude the presence of residual defects.
Oxygen exposure has been conducted at fixed tempera-
tures by keeping the Ru crystal in an oxygen bath at a con-
stant partial pressure varied between 10?7and 1000 mbar,
and controlled by using the IKR060 and TPR018 ?Balzers?
instruments for pressures below and above 10?3mbar, re-
spectively. The sample temperature has been measured and
stabilized by means of a K-type thermocouple, spot-welded
to the backside of the crystal. The oxygen amount leaving
the Ru?0001? sample has been determined by integrating the
O2-TD traces. Calibration has been performed by relating the
resulting areas to those derived from an O2-TD spectrum
corresponding to a saturated layer of chemisorbed oxygen,
which forms a 1?1-O LEED superstructure ?1 Monolayer
Equivalent?1.58?1015cm?2).5All TD spectra were taken
with a constant heating rate of 4 K/s. In order to achieve
oxygen contents above 1 MLE at the low temperatures used,
it was necessary to employ huge oxygen exposures
up to 1014L, a value, which requires a continuous oxygen
exposition at 1 bar over about 45 hours (1 L?1.33
?10?6mbar*s).
Heating the sample up to 1550 K removes all oxygen as
evidenced by TDS and confirmed by UPS andAES.Also, the
corresponding surface smoothness matches that of the
freshly prepared surface. Yet, in order to exclude any poten-
tial degradation of the subsurface region or any accumulation
of oxygen species buried deeper in the bulk, the above sput-
tering and annealing procedures were carried out prior to
each new oxidation run.
III. RESULTS
A. Evidence for subsurface oxygen
Figure 1?a? shows O2–TD spectra taken after exposing
the Ru?0001? surface, kept at 475 K, to increasing oxygen
exposures ranging from 10 to 5?1013L. As established by
previous studies,5the lower oxygen exposures lead to a layer
FIG. 1. ?a? O2–TD spectra taken after exposing the Ru?0001? surface kept at
475 K to various oxygen doses ranging from 10 to 5?1013L. ?b? O2–TD
spectra taken after exposing the Ru?0001? surface kept at 525 K to various
oxygen doses ranging from 104to 1011L. ?c? O2–TD spectra taken after
exposing the Ru surface to 1011L of oxygen at a constant partial oxygen
pressure of 1 bar at various sample temperatures as indicated in plot.
3872J. Chem. Phys., Vol. 120, No. 8, 22 February 2004Blume et al.
Downloaded 23 Feb 2004 to 129.13.167.7. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp

Page 3

of chemisorbed oxygen. The corresponding TD spectra are
broad and cover a relatively wide temperature range. Their
shape evolution is characterized by a temperature downshift
of the desorption onset with increasing lateral density of ad-
sorbed atoms. Such a behavior signifies a repulsive atom–
atom interaction in the layer, which leads to a lowering of the
activation energy for desorption as the mean distance be-
tween adjacent oxygen atoms decreases.9,12Eventually, a
complete 2D oxygen monolayer is formed with a saturation
density of 1.58?1015atomspercm2. The properties of this
phase have been characterized previously by means of vari-
ous spectroscopic methods12–15and density functional theory
?DFT?-calculations.15Here, we use the characteristic disap-
pearance of the ?1/2,1/2? LEED spots of the ?2?2? pattern as
indicator for the completion of the monolayer.12The corre-
sponding exposure of 5?105L provides the calibration of
the absolute amount of desorbing molecules from the respec-
tive TDS curve in monolayer equivalent, MLE ?see lower
panel of Fig. 6?.
For oxygen exposures between 103L and 5?105L, the
TD spectra exhibit a distinct peak growing at about 1040 K
?indicated as ? in Fig. 1?a??. With its appearance any further
downshift stops although the spectra still correspond to the
chemisorption layer regime. This shape progression can be
fully understood in terms of the formation of the chemi-
sorbed oxygen layer and has been modeled on the basis of
repulsive interaction.16
Unexpectedly, peak ? continues to grow even after
monolayer completion with increasing oxygen doses becom-
ing the dominating spectral feature for exposures up to
108L. Since the oxygen load corresponds then to about 1.5
MLE, we must conclude that the surplus oxygen is incorpo-
rated in the subsurface region. It should be noted that the
desorption maximum of peak ? is virtually not shifted; only
its size is increased.
For exposures higher than 107L, a new peak appears
centered between 900–950 K ?indicated as ? in Fig. 1?a??.
Peak ? also grows with oxygen exposure and finally, at ex-
posures above D?1012L, becomes comparable in size with
peak ?. Peak ? grows up to 2 MLE at 5?1013L exposure
where the presented experimental series was stopped because
even at 1 bar the exposure time approached about 16 h. It
must be emphasized that this feature is clearly developed and
distinct from other TDS features only if oxygen exposure is
performed at sample temperatures below 650 K. As known
from previous studies,7beyond this temperature substantial
oxide formation starts dominating totally the other spectral
features. However, below this threshold, we found the for-
mation of both peaks to be strongly temperature dependent.
We measured an exposure sequence similar to that in Fig.
1?a? at an increased temperature of 525 K. The resulting
series is presented in Fig. 1?b?. It is apparent that the se-
quence of appearance of the two peaks is basically indepen-
dent from temperature. Peak ? evolves only if the formation
of peak ? is practically completed and requires substantially
larger exposures. As the second important result, the com-
parison of the exposures necessary to produce a similar ?/?
intensity ratio is reduced by more than two orders of magni-
tude by increasing the sample temperature from 475 K ?Fig.
1?a?? to 525 K ?Fig. 1?b??. On the other hand, for preparation
temperatures only slightly above room temperature, peak ?
appeared for exposure conditions as applied in Fig. 1?a? just
as a weak shoulder at the low temperature side of peak ?.
Extrapolating the observed trend, we expect at room tem-
perature a substantial formation of peak ? only for pressures
of about 10 bar or even higher.
In order to show the correlation between exposure and
temperature in more detail, Fig. 1?c? displays the evolution
of both peaks for the case of constant exposure but increas-
ing sample temperature. The sequence of appearance is vir-
tually identical to the series in Figs. 1?a? and 1?b? implying
thereby that temperature and exposure are reciprocally
coupled, the higher the temperature the lower the exposure to
produce a similar state of incorporated oxygen.
At this point, we must consider how far the features in
the TD spectra represent different states of oxygen incorpo-
rated in the subsurface region. Usually, the temperature of a
TD peak maximum can be related to the binding energy of
the adsorbed particle involved in a desorption event via a
corresponding kinetic equation.17,18In the case of associative
desorption second order kinetics is valid since the migrating
atoms must first meet and form a transition molecular state
prior to the escape from the surface. Including the intralayer
repulsive interactions, the resulting TD spectra from the
chemisorbed layer are fully understood up to monolayer
coverage.16
For a desorption peak involving subsurface species the
situation is more complicated and peak ? and ? must be
subsurface mediated because the desorbing amount exceeds
the monolayer value. Peak ? appears on the first view as a
simple continuation of the chemisorption layer desorption for
larger coverage. However, it represents eventually about 0.5
MLE surplus oxygen desorbing at that temperature, which
signifies saturation of the chemisorption layer. A plausible
mechanism, which would lead to such a behavior, involves
oxygen atoms located closely below the surface. They can
occupy empty surface sites once these are created by the
onset of thermal desorption out of the chemisorption layer.
This kind of subsurface oxygen atoms ?denoted in the fol-
lowing as O?) then acts as a reservoir for maintaining the
monolayer coverage until depletion. Energetically, the corre-
sponding subsurface sites must be favored as long as all sur-
face sites are occupied. On the other hand, this mechanism
can also explain the high pressure–exposure necessary to
create the subsurface oxygen since the monolayer would
cause a high activation barrier for the incorporation process
with the O2dissociation as the most likely rate-determining
step. It must be pointed out, however, that we must assume a
precursor state for molecular oxygen to account for the ob-
served dependence on the substrate temperature, namely the
reciprocal coupling of substrate temperature and exposure as
reflected in Fig. 1?c?.
This also applies to the formation of the subsurface oxy-
gen, which gives rise to peak ? in the desorption spectra
?denoted as O?). In addition, the successive appearance of
the ?-peak strongly indicates a formation of O?just in the
formation channel. The possibility of a temperature induced
formation of the distinct O?and O?species can be ruled out
3873 J. Chem. Phys., Vol. 120, No. 8, 22 February 2004Oxide-free oxygen incorporation into Ru(0001)
Downloaded 23 Feb 2004 to 129.13.167.7. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp

Page 4

since we observed virtually no change in the peak ratio when
keeping the sample after exposure for an extended time pe-
riod close to the temperature of desorption onset. Thus, an
exchange between the two species is not taking place.
Presently we are performing a detailed study focused on
the role of surface defects on oxygen dissolution. For sur-
faces with intentionally created defects (Ar?sputtering, 500
V, 1.6 ?A, 120 s? the measured O2–TD spectra also exhibit
both peaks ?? and ?? very alike the spectra in Fig. 1?a?, but
for drastically smaller oxygen exposures.19Figure 2 demon-
strates this finding. It shows two O2–TD spectra taken after
performing an identical oxidation procedure (D?1011L, T
?475K, p?1 bar) at a rough and a smooth surface, dashed
and solid trace, respectively. For that oxygen exposure,
the defect-enriched surface shows peak ? with an intensity
exceeding even that of the topmost curve in Fig. 1?a?
(5?1013L) while ? for the smooth surface appears as in
Fig. 1?a?. Thus, the formation of the ?-related oxygen spe-
cies is closely connected with the presence of surface de-
fects. The apparent correlation between the defect density
and the amount of O?species indicates a mediating effect of
the defects onto the formation kinetics. In consequence, we
must assume a precursor state of oxygen atoms, which dif-
fuse over the saturated ?1?1? oxygen layer and may either
permeate the chemisorption layer and occupy a subsurface
site (O?) or encounter a defect, where they form O?. Given
a certain residence time of the precursor oxygen atom, the
probability of the latter event depends strongly on the mean
distance between the defects, i.e., their density. As the total
amount of O?of 2 MLE rules out an incorporation confined
in the vicinity of the defects, we propose that the structural
distortions due to the defects trigger the formation of O?,
which then spreads further into the subsurface region. On
desorption, the process may be reverted. O?atoms diffuse to
the defects where they recombine to O2and leave the sur-
face. It must be noticed, however, that the clean surface an-
nealing is in effect at the temperature of O?desorption,
which supports our interpretation of the kinetically governed
formation of the O?species.
The TD spectra presented so far correspond to a state of
the substrate where the presence of Ru-oxide can be defi-
nitely ruled out on basis of the LEED pattern. No indication
of the characteristic RuO2superstructure spots has been ob-
served. However, the increase of diffuse scattering was
clearly reflected by an enhanced but unstructured back-
ground. On the other hand, both the topmost spectrum in
Figs. 1?a? and 1?c? exhibits an apparent shift of peak ? to
higher temperatures. As this is indicative of the onset of ox-
ide formation, we measured the temperature dependence of
the TD spectra at the highest reasonably accessible exposure
(5?1013L) to shed some light on the transition region. The
spectra shown in Fig. 3?a? demonstrate that over an ex-
tremely narrow temperature interval an additional TD peak
?denoted as ?? develops first at 1120 K ?exposure at 500 K?,
shifts upwards by about 30 K ?exposure at 525 K? and back
downwards in the topmost spectrum. It must be noted that
this spectrum is shown with reduced scale. Apparently, fur-
ther intensity increase of peak ? will eventually totally cover
peaks ? and ? as is apparently the case in the full oxidation
sequences presented in our previous papers where we per-
formed the exposures at much higher temperatures.20,21Cor-
respondingly, we assign the appearance of peak ? to the very
initial formation stage of RuO2nuclei, which for further evo-
lution probably start to grow in size, and in succession lead
to the absolute predominance of oxide-mediated desorption.
This isalsosupported by
RuO2-superstructure in LEED for only slightly higher prepa-
ration temperatures at about 600 K.
The appearance of the RuO3fragment in TDS has been
turned out as a very convenient signature for the presence of
RuO2on the substrate.7Accordingly, we have measured the
RuO3desorption ?Fig. 3?b?? corresponding to the preparation
procedure, which gave rise to the three-peak spectra of Fig.
3?a?. For means of a direct comparison, these are also shown
in Fig. 3?b?. The RuO3features are denoted as ? and ? peaks.
The apparently identical evolution behavior of the O2and the
RuO3traces of peak ? and ?, respectively, confirm our as-
signment of peak ? as due to RuO2nuclei and/or islands.
Surprisingly, an additional RuO3maximum ??? appears at the
trailing edge of peak ?. Since we know that this state of
incorporated oxygen is not genuine RuO2, at least does not
show its structural features, we have to conclude that a dy-
namical transition occurs in the course of the temperature
sweep of the TDS acquisition, and RuO2is formed as an
intermediate.
As recently demonstrated, RuO2(110) domains created
by high temperature oxidation, offer coordinatively unsatur-
ated Ru atoms (Rucus) as additional adsorption sites for oxy-
gen atoms.22,23Such adsorbates (Ocusatoms) are stable on
the RuO2domains up to about 600 K. Moreover, they exhibit
anunusually highreactivity
reaction.24,25We have used the existence of Ocussites as fin-
the appearanceof the
forthe CO-oxidation
FIG. 2. O2–TD spectra taken from two surfaces oxidized using the same
parameters (D?1011L, T?475 K, p?1 bar). The solid curve corresponds
to a nearly defect-free smooth surface, the dashed curve to a surface after a
mild Ar ion bombardment ?500 V, 1,6 ?A, 120 s?. For the defective surface
a roughness coefficient of 0.4 has been measured by He-TEAS.
3874J. Chem. Phys., Vol. 120, No. 8, 22 February 2004 Blume et al.
Downloaded 23 Feb 2004 to 129.13.167.7. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp

Page 5

gerprint feature to further elaborate the constitutive differ-
ences between the subsurface O?and O?species observed
here and the O atoms present in RuO2domains. Figure 4
shows O2–TD spectra in the range between 300 and 600 K
after adsorbing 10 L of oxygen at room temperature on three
differently treated surfaces. The middle panel displays the
spectrum obtained for a surface, which consists of RuO2is-
lands surrounded by regions containing subsurface oxygen.
The applied oxidation conditions were the same as used in
Ref. 7 (T?750K, p?5?10?3mbar, D?105L) and the
TDS data are in good agreement with the findings of Ref. 7.
The spectrum shown in the lower panel has been obtained
from the surface, which contained only subsurface oxygen,
i.e., the surface kept at 475 K was exposed to 1011L of
oxygen at partial oxygen pressures of 1 bar ?see Fig. 1?. Note
that virtually no oxygen desorption is detected. Evidently, the
Ocusadsorption phase does not exist at a surface, which ex-
clusively contains subsurface oxygen. The situation illus-
trated in the upper panel presents desorption from a surface
terminated by a thick oxide layer ?oxidation procedure: T
?750K, p?100mbar, D?1010L) and shows a fully devel-
oped Ocusdesorption feature. Evidently, at room temperature
oxygen molecules adsorb dissociatively only on surfaces
which are at least partially terminated by RuO2. Thus, from
the evidence presented in Fig. 4 we can rule out that a sur-
face exposed to oxygen at low-temperature is terminated by
either RuO2domains or an oxide layer.
B. Electronic structure
In order to explore the effects of the subsurface oxygen
species on the electronic structure of the surface, we have
monitored the features in the Ru?0001? valence band region
for the various treatments by means of UPS using the HeI
line of a discharge lamp ?45° incidence, normal emission
detection?. In Fig. 5, we compare the ultraviolet ?UV? spectra
of clean Ru?0001? ?lower panel? and a surface covered with
RuO2domains ?upper panel? with the features developing for
a sequentially increased oxygen load, in correspondence with
the treatment used for TDS in Fig. 1 ?middle panel?. The
spectrum of the clean surface exhibits three dominating fea-
tures ?marked as I, II, and III? centered at about 5.0, 2.6, and
FIG. 3. ?a? O2–TD spectra after exposing the Ru surface to an oxygen dose
of 5?1013L O2at constant sample temperatures of T?475, 500, 525, and
550 K, from bottom to top, respectively. ?b? O2- and RuO3–TD spectra
taken from the surface exposed to 5?1013L of O2at three representative
sample temperatures T?475, 500, and 525 K, lower, middle, and upper
panel, respectively.
FIG. 4. O2–TD spectra taken after exposing differently oxidized surfaces at
room temperature to 10 L of oxygen. Three different oxidation procedures
have been applied: ?a? Upper panel: The surface was covered by a thick
RuO2 film created by oxidizing the Ru?0001? surface under high-
temperature and high-pressure conditions (T?750 K, p?100 mbar, D
?1010L). ?b? Middle panel: The surface consisted of RuO2domains sur-
rounded by subsurface oxygen. The oxidation procedure has been conducted
at following conditions: T?750 K, p?5?10?3mbar, D?105L ?Ref. 7?.
?c? Lower panel: the surface contained the subsurface oxygen only, i.e., the
surface kept at 475 K was exposed to 1011L of oxygen at 1 bar, yielding an
initial oxygen content of about 2.7 MLE.
3875 J. Chem. Phys., Vol. 120, No. 8, 22 February 2004Oxide-free oxygen incorporation into Ru(0001)
Downloaded 23 Feb 2004 to 129.13.167.7. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp

Page 6

0.3 eV, which are due to direct transitions from Ru-4d bands
and have recently been fully assigned in detail.26,27Due to a
low cross section for photoionization, the emission from the
Ru-5s orbital is very weak and practically not visible.28
For the initial stage of oxygen uptake at 475 K, D
?105L, where just the oxygen chemisorption layer is built
up to the monolayer coverage, peaks I and II show continu-
ous chemical shifts. While the binding energy of peak II
decreases by about 0.3 eV, that of peak I increases by almost
0.5 eV ?not shown?. Further exposures, 105?D?108L, con-
tinue this trend up to the situation shown in the bottom spec-
trum of the middle panel. The binding energy shifts reach a
maximum value of ?0.4 eV and ?0.7 eV for peaks II and I,
respectively. Recalling the corresponding TD spectra of Fig.
1, at that exposure the total oxygen content exceeds 1 MLE.
Incorporation into the subsurface region has started and peak
? is clearly developed. For exposures higher than 108L, cor-
responding to an oxygen load exceeding 1.5 MLE ?see also
Fig. 6?, a new peak emerges at a binding energy of 3.6 eV.
Concurrently, all Ru-4d peaks become attenuated and, at an
oxygen exposure of 1011L, are just distinguishable from the
background. For oxygen exposures above 1011L, the new
peak becomes more prominent and dominates for exposures
beyond 1012L. The total oxygen content at this stage ex-
ceeds 3 MLE.
The overall spectral features differ substantially from the
characteristics of the valence band region of a clean Ru. The
initial stages of oxygen incorporation into the subsurface re-
gion are reflected in UPS by a continuation of the shift of the
Ru-4d peaks already apparent for the chemisorption layer.
This process is probably induced by the incorporation of
oxygen atoms into the interstitial sites of the Ru lattice ?see
above? what in turn proceeds to reduce the electronic overlap
of the 4d orbitals between adjacent Ru atoms ?now to the
second layer? and, as a consequence, causes the observed
further shift of the Ru-4d features.26,27Their eventual disap-
pearance for oxygen exposures higher than 1011L may be
related to a change in the electronic structure due to the
beginning of the formation of microscopic oxide nuclei as
indicated by the emergence of the TDS-peak ?. In recent
DFT-calculations, performed by Reuter et al.29,30the forma-
tion is expected for oxygen loads ?4 MLE. The appearance
of a further peak at 3.6 eV may be induced by the accumu-
lation of additional oxygen via surface defects since its evo-
lution matches the growth of TDS-peak ?. However, it is not
clear whether this peak is induced by a change in the band
structure or is to be attributed to a localized O-2p state. For
comparison, the upper panel in Fig. 5 shows the spectrum of
the valence band region after substantial oxidation at 750 K
with an exposure of 105L. Here the surface contains large
oxide domains24as manifested by the strong characteristic
peak IV centered at about 4.4 eV.7When comparing spectra
of a substrate predominated by subsurface oxygen ?middle
panel? with spectra of RuO2domains, there are no matching
spectral features.
As shown above, the oxygen deposition considerably
modifies the valence band of the surface. This is reflected not
only in binding energy shifts and the appearance of new
photoelectron peaks but also in changes of the average work
function. Figure 6 shows a compilation of data assembled
from UPS, TDS, and LEED. Both the work function change
FIG. 5. ?a? Lower panel: UP ?21.2 eV? spectrum of a clean Ru?0001? sur-
face. ?b? Middle panel: modifications of the valence-band structure as in-
duced by exposing the Ru surface kept at 475 K to increasing oxygen. ?c?
Upper panel: valence band of the Ru surface consisting of RuO2domains
surrounded by regions terminated by the O?1?1? layer ?applied oxidation
conditions T?750 K, p?5?10?3mbar, D?105L).
FIG. 6. Work function change ??, relative to a clean surface, and oxygen
content COvs oxygen exposure ?full lines?. Both functions result from ap-
plying the following oxidation procedure, T?475 K, p?1 bar. The expo-
sure region where the LEED pattern showed the O?1?1? structure is indi-
cated. The work function change for a high temperature oxidation, T
?750 K, p?1 bar, is plotted for comparison ?dashed line?.
3876J. Chem. Phys., Vol. 120, No. 8, 22 February 2004Blume et al.
Downloaded 23 Feb 2004 to 129.13.167.7. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp

Page 7

?determined from the cut-off edge of the UV spectra? and the
oxygen uptake are presented as functions of increasing oxy-
gen exposures. Note the logarithmic scale of the exposure.
The range of exposures where the O?1?1? LEED pattern is
observed, is additionally indicated in the figure. It is present
as soon as the oxygen chemisorption layer is completed at an
exposure of ?105L ?see arrow line 1?. At saturation, the
dipole layer reaches its maximum strength15,29and the work
function its highest value. For exposures up to 108L, where
a steady increase of the oxygen content to 1.5 MLE occurs,
the work function decreases by about 0.3 eV concurrently
with the maximum binding energy shift of peak I ?arrow line
2?. Interestingly, for higher exposures up to 1011L where the
oxygen content rises more steeply from 1.5 to about 3 MLE,
and the new photoelectron peak appears in the UP spectra,
the work function remains essentially unchanged ?arrow line
3?. Likewise, the LEED pattern is unaffected by the increas-
ing oxygen content. With further increase of the oxygen ex-
posure, another change of the slope of the uptake curve is
observed due to the onset of oxide formation ?see Fig. 3?a??.
While the work function still remains unchanged, the LEED
pattern fades until no spots are distinguishable from the
background. In stark contrast, the work function decreases
by more than 0.7 eV when the high-temperature oxidation
procedure is used ?dotted line?. For that case, the surface is
covered with RuO2islands and the averaged work function,
as obtained from the photoelectron cutoff in the spectra, is
predominantly determined by emission from the low work
function patches. This domain formation is also apparent
from the additional RuO2oxide spots in the LEED pattern23
and directly in the PEEM/SPEM images.3,21
C. Reactivity of the subsurface oxygen
As an additional means, we use the CO oxidation reac-
tion as a probe for the reactivity of the subsurface oxygen
species. For this aim, O2–TD spectra taken before and after
exposing the sample to a particular CO dose are compared.
Thus, the amount of oxygen atoms removed from the surface
via the CO?Oad→CO2reaction is directly calculated from
the resulting difference of the desorption spectra, and used
for an estimation of the integral reaction yield. Figure 7
shows the O2–TD spectra taken before and after exposing
the Ru surface to a sequence of increasing CO doses. Oxy-
gen incorporation has been performed at 475 K by exposing
the surface to 1011L of oxygen resulting in about 2.7 MLE of
oxygen in total without any traces of RuO2domains ?see Fig.
1?a??. The CO–CO2-reaction has been accomplished at a
sample temperature of 475 K.
Apparently, the intensity of peak ? decreases preferen-
tially with increasing CO exposure. Whereas the oxygen rep-
resented by peak ? disappears entirely at an exposure of
about 4000 L, peak ? appears to be at most slightly affected.
Even for higher CO exposures, its intensity is practically
stable with remaining oxygen content of about 1.5 MLE.
Analysis of the corresponding peak area of ? yields an inte-
gral probability for the CO?Oad→CO2event of roughly
10?3per impinging CO molecule. Further experiments at el-
evated reaction temperatures show that the reaction effi-
ciency increases with temperature.
Since the reaction necessarily takes place on top of the
surface, the Osub→Osurfacetransition must be the preceding
step being just the reversal of the formation mechanisms ?see
above?. The reaction yield is then most likely governed by
the diffusion flow of Osubatoms in the vicinity of surface
defects. In consequence, O?, as the defect-mediated species
is efficiently consumed. This interpretation is supported by
the behavior of surfaces with intentionally created defects
where even higher amounts of O??see Fig. 2? are substan-
tially reduced by a CO exposure as used here.19On basis of
the data available, the reaction channel manifested by the
much weaker but resolvable depletion of peak ? cannot be
quantified yet. However, it is likely to take place on flat
terraces. A more detailed model will be presented after final-
izing our molecular-beam study of this reaction.19
It should be noted, that the reaction probability of 10?3
determined here represents the upper limit for the reaction
with exclusively subsurface oxygen. On the other hand, the
reaction performed on Ru surfaces containing RuO2domains
exhibits conversion probabilities of the same order.30
IV. DISCUSSION
One of the main results of the experiments presented
here concerns the existence of oxygen phases, which must be
considered as precursors of the ‘‘true’’ oxide formation.
While the transformation into the stable RuO2configuration
requires not only a change of the chemical state of a Ru atom
but equally well a substantial restructuring of the whole Ru
lattice ?at least on a mesoscopic scale?, the precursor oxygen
species conserve the original lattice structure and very likely
occupy interstitial sites below the surface. By using specific
exposure conditions, in particular temperatures below 550 K,
we have been able to prepare the different phases separately
FIG. 7. O2–TD spectra taken after exposing the substrate, loaded with the
subsurface oxygen phase, to a series of different CO doses. The initial oxy-
gen content was about 2.7 MLE. The CO exposure has been performed at a
partial CO pressure of 10?3mbar and a sample temperature of 475 K, both
kept constant during the reaction.
3877J. Chem. Phys., Vol. 120, No. 8, 22 February 2004Oxide-free oxygen incorporation into Ru(0001)
Downloaded 23 Feb 2004 to 129.13.167.7. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp

Page 8

and to demonstrate their precursor character. The mostly
used temperatures for oxidation7,23tend to hide their pres-
ence, because the competing growth of the RuO2islands
predominates totally almost all spectroscopic data. However,
close inspection of TD spectra for instance, reveal clear in-
dication of features, we have shown to be due to the presence
of precursor oxygen.7,31Even when using NO2as the oxi-
dant, which is not directly comparable, the consecutive for-
mation of the O?and O?species is evident in the TD spectra
published by Weinberg et al.32In general, we are confident
that the formation of RuO2always proceeds via the precursor
species O?and O?correlated with the smooth and the defect
areas of the surface, respectively.
Since the O?species is formed at the flat areas of the
substrate ?see above? it should be the more appropriate can-
didate for a theoretical study, because an ideal surface is
naturally easier accessible to modeling. Very recently, Reuter
et al.29,31reported on a model description as to how oxygen
atoms in a first step are incorporated below the Ru?0001?
surface when the chemisorption layer is saturated. Based
upon density-functional theory ?DFT? calculations, a detailed
pathway towards the oxide formation has been proposed. Ac-
cording to this model, oxidation proceeds via a sequence of
three distinct Ru–O phases involving oxygen atoms located
within the three topmost Ru layers. Initially, the surface of
the top Ru layer becomes saturated with a ?1?1?-O adlayer.
Then, penetration of further oxygen atoms into sites directly
below the top Ru layer becomes energetically favorable.
Thus, an O–Ru?I?–O sandwich layer is formed. Further-
more, it turned out that such an configuration becomes de-
coupled from the bulk Ru lattice when subsequently the sec-
ond Ru?II? layer becomes saturated in the sense that an
oxygen layer is formed similar to the chemisorption layer.
With further oxygen incorporation this process is repeated
with the saturation of the deeper lying Ru?III? and Ru?IV?
layers. Structural changes leading to a RuO2-(110) bulk for-
mation are expected to occur for oxygen loads ?4 MLE.
As a logical consequence, we identify the O?species as
the subsurface part of a O–Ru?I?–O?sandwich layer. We
expect, however, that on a real surface the extension of such
a sandwich in its perfect form reaches a length in the meso-
scopic scale. Thus, the measured number of O?species
amounts to 0.5 MLE while on an ideal surface the proposed
sandwich layer should reach 1 MLE.
The detailed configuration of the O?species cannot be
elucidated on basis of our data. However, the definite con-
nection of its formation probability with the defect density,
and the deduced kinetic role of individual defects underline
the importance of this phase in real systems. In particular, the
high catalytic activity apparent from the CO oxidation reac-
tion measurements ?Fig. 7? gives rise to expect a consider-
able contribution under conditions of real world catalysis,
which is operated at high pressures and medium tempera-
tures, comparable to the settings used here. It must also be
noted that the peak at 3.6 eV binding energy in the UV
spectra ?Fig. 5? is definitely induced by O?possibly indicat-
ing a substantially shifted contribution from the 2p states of
oxygen. A more detailed interpretation requires further ex-
perimental and theoretical efforts.
As the last stage of the preparation sequence presented,
the ? peak appears in the TD spectra ?Fig. 3?, which we
identify with the presence of nuclei of RuO2. In contrast to
the peak shapes observed in the TD spectra from substan-
tially oxidized surfaces, the ? peak is rather narrow, a char-
acteristic, representative for well-defined processes. In our
interpretation, this can be traced back to the fact that the
oxide nuclei are predominantly of the same size, in particular
at that initial oxidation stage prepared here. Upon further
oxidation, these nuclei grow in size and, in consequence,
their size distribution should become broader. The corre-
sponding TD peak also broadens and shifts slightly to lower
maximum temperatures, as the nuclei become islands. Thus,
the transition to the TD spectra in the literature dealing with
higher oxidation stages up to closed oxide layer becomes
apparent.33
V. SUMMARY
By exposing a Ru?0001? sample to oxygen at high pres-
sures ??1 bar? and low temperatures ?350–450 K?, we have
deposited substantial amounts of oxygen ??3.5 MLE? in the
subsurface region, without creating RuO2domains. The non-
oxidic nature of the subsurface oxygen phase has been dem-
onstrated by several experimental methods. The incorpora-
tion of the oxygen atoms proceeds via two reaction channels,
which lead to well distinguishable TD features, O?and O?.
Peak ? has been found to be due to surface defects by com-
paring the O2–TDS taken from a smooth and a surface with
intentionally created defects.
The distinct properties of the subsurface phases, which
definitely rule out the presence of RuO2, are evident from
the LEED data, which show the ?1?1? pattern, and the fact,
that no Ocusspecies are created upon room temperature ad-
sorption. In addition, the induced UPS features are clearly
different from the established RuO2signature. However, the
integral reaction yields of the O?and the RuO2phases are in
the same order of magnitude.
ACKNOWLEDGMENTS
The authors thank Karsten Reuter, Matthias Scheffler,
and Herbert Over for stimulating discussions. Raoul Blume
and Horst Niehus acknowledge the financial support by the
DFG through project Ni-452.
1V. E. Henrich and P. A. Cox, The Surface Science of Metal Oxides ?Cam-
bridge University Press, Cambridge, 1993?.
2A. Bo ¨ttcher, B. Krenzer, H. Conrad, and H. Niehus, Surf. Sci. Lett. 466,
811 ?2000?.
3A. Bo ¨ttcher, B. Krenzer, H. Conrad, and H. Niehus, Surf. Sci. 504, 42
?2002?.
4G. Renaud, Surf. Sci. Rep. 32, 1 ?1998?.
5A. Bo ¨ttcher, H. Niehus, S. Schwegmann, H. Over, and G. Ertl, J. Phys.
Chem. B 101, 11185 ?1997?.
6A. Bo ¨ttcher, M. Rogozia, H. Niehus, H. Over, and G. Ertl, J. Phys. Chem.
B 103, 6267 ?1999?.
7A. Bo ¨ttcher, H. Conrad, and H. Niehus, J. Chem. Phys. 112, 4779 ?2000?.
8D. Korsch, Z. Angew. Phys. 28, 43 ?1969?.
9T. E. Madey, H. A. Engelhardt, and D. Menzel, Surf. Sci. 48, 304 ?1975?.
10I. Surnev, G. Rangelov, and G. Bliznakov, Surf. Sci. 159, 299 ?1985?.
11B. Poelsema and G. Comsa, Scattering of Thermal Energy Atoms from
Disordered Surfaces, Springer Tracts in Modern Physics, Vol. 15
?Springer-Verlag, Berlin, 1989?.
3878J. Chem. Phys., Vol. 120, No. 8, 22 February 2004Blume et al.
Downloaded 23 Feb 2004 to 129.13.167.7. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp

Page 9

12C. Stampfl, S. Schwegmann, H. Over, M. Scheffler, and G. Ertl, Phys.
Rev. Lett. 77, 3371 ?1996?.
13P. He and K. Jacobi, Phys. Rev. B 55, 4751 ?1997?.
14K. I. Kostov, M. Gsell, P. Jacob, T. Moritz, W. Widdra, and D. Menzel,
Surf. Sci. 394, L138 ?1997?.
15C. Stampfl and M. Scheffler, Phys. Rev. B 54, 2868 ?1996?.
16C. Stampfl, H. J. Kreutzer, S. H. Payne, H. Pfnu ¨r, and M. Scheffler, Phys.
Rev. Lett. 83, 2993 ?1999?.
17P. A. Redhead, Vacuum 12, 203 ?1962?.
18K. Christmann, Introduction to Surface Physical Chemistry ?Steinkopff,
Darmstadt, Springer, New York, 1991?.
19R. Blume, H. Niehus, H. Conrad, and A. Bo ¨ttcher ?unpublished?.
20A. Bo ¨ttcher and H. Niehus, J. Chem. Phys. 110, 3186 ?1999?.
21A. Bo ¨ttcher, U. Starke, H. Conrad, R. Blume, L. Gregoriatti, B. Kaulich,
A. Barinov, and M. Kiskinova, J. Chem. Phys. 117, 8104 ?2002?.
22H. Over, Y. D. Kim, A. P. Seitsonen, E. Lundgren, M. Schmid, P. Varga, A.
Morgante, and G. Ertl, Science 287, 1474 ?2000?.
23Y. D. Kim, A. P. Seitsonen, and H. Over, Surf. Sci. 465, 1 ?2000?.
24A. Bo ¨ttcher and H. Niehus, Phys. Rev. B 60, 14396 ?1999?.
25C. Y. Fang, J. Wang, K. Jacobi, and G. Ertl, J. Chem. Phys. 114, 10058
?2001?.
26P. Hofmann and D. Menzel, Surf. Sci. 152Õ153, 382 ?1985?.
27T. Pelzer, G. Ceballos, F. Zbikowski, B. Willerding, K. Wandelt, U. Tho-
mann, Ch. Reuss, Th. Fauster, and J. Braun, J. Phys.: Condens. Matter 12,
2193 ?2000?.
28J. J. Yeh and I. Lindau, At. Data Nucl. Data Tables 32, 1 ?1985?.
29K. Reuter, M. V. Ganguglia-Paravano, C. Stampfl, and M. Scheffler, Phys.
Rev. B 65, 165403 ?2002?.
30A. Bo ¨ttcher, H. Conrad, and H. Niehus, Surf. Sci. 452, 125 ?2000?.
31I. J. Malik and J. Hrbek, J. Phys. Chem. 95, 10188 ?1991?.
32W. J. Mitchell and W. H. Weinberg, J. Chem. Phys. 104, 9127 ?1996?.
33K. Reuter, C. Stampfl, M. V. Ganguglia-Paravano, and M. Scheffler,
Chem. Phys. Lett. 352, 311 ?2002?.
3879J. Chem. Phys., Vol. 120, No. 8, 22 February 2004 Oxide-free oxygen incorporation into Ru(0001)
Downloaded 23 Feb 2004 to 129.13.167.7. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp