Surface Science 418 (1998) 267–272
Structural analyses of the pure and cesiated
Ru(0001)–(2×2)-3O phase
Y.D. Kim, S. Wendt, S. Schwegmann, H. Over *, G. Ertl
Fritz-Haber-Institut der Max-Planck-Gesellschaft, Faradayweg 4–6, D-14195 Berlin, Germany
Received 17 June 1998; accepted for publication 31 August 1998
Abstract
The structures of the (2×2)-3O and the (2×2)-3O-Cs overlayers on the Ru(0001) surface were analyzed by LEED. The
Ru(0001)–(2×2)-3O phase is characterized by a (2×2) vacancy network in which CO molecules are not able to adsorb whereas
Cs atoms are. The averaged topmost Ru layer distance (2.21 Å) is slightly expanded with respect to the bulk value, which is
consistent with the observed trend that the topmost Ru layer distance increases with O-coverage. Coadsorbing Cs atoms reside in
the vacancies exposed by the (2×2)-3O overlayer, whereby the local structural parameters remain essentially unaltered. © 1998
Elsevier Science B.V. All rights reserved.
Keywords: Alkali metals; Chemisorption; Compound formation; Low energy electron diffraction (LEED); Low index single crystal
surfaces; Oxygen; Ruthenium; Surface chemical reaction; Surface structure, morphology, roughness, and topography
On most close-packed fcc(111) and hcp(0001)
transition metal surfaces, the exposure of molecular oxygen is restricted to form a (2×2)-O surface
structure at low coverages and either a (2×1) or
a (앀3×앀3)R30° overlayer at higher O coverage
dependent on the lateral interaction between the
O atoms [1]. A system which departs from this
general trend is O/Ru(0001). It has been known
for many years that oxygen adsorption produces
(2×2)-O and (2×1)-O overlayers on Ru(0001)
with O-coverages of 0.25 and 0.5 respectively [2].
Recent density functional theory (DFT ) calculations [3,4], however, predicted not only the existence of a (1×1)-O overlayer on Ru(0001), which
has been confirmed by a recent low energy electron
* Corresponding author. Fax: (+49) 30 84135106;
e-mail: [email protected]
diffraction (LEED) study [5], but also that of a
hitherto not known (2×2)-3O overlayer. The
(2×2)-3O overlayer is more stable than the separated (1×1)-O plus (2×1)-O domains by about
170 meV per (2×2) unit cell.
In the meantime it has been demonstrated that
the (2×2)-3O phase can be prepared by exposing
the Ru(0001) surface either to large amounts of
molecular oxygen [6 ] or alternatively to NO at
2
elevated sample temperatures [7]. At 500 K, NO
2
decomposes into O and NO at the Ru(0001)
ad
ad
surface, and NO desorbs immediately into the gas
phase since NO desorption from Ru(0001) is
already completed at about 500 K [8]. As a consequence, the Ru(0001) surface becomes enriched
with chemisorbed (atomic) oxygen, and oxygen
loads equivalent to more than 1 ML can be easily
produced [7]. Atomic oxygen from NO is a viable
2
0039-6028/98/$ – see front matter © 1998 Elsevier Science B.V. All rights reserved.
PII: S0 0 39 - 6 0 28 ( 98 ) 0 07 2 2 -5
268
Y.D. Kim et al. / Surface Science 418 (1998) 267–272
method which requires a much smaller gas load
than exposure of molecular oxygen.
The (2×2)-3O phase can be viewed as a wellordered ‘‘imperfect’’ (1×1)-O overlayer in which
CO molecules should be able to coadsorb and
react to form CO . Therefore, the study of the
2
(2×2)-3O phase may provide deeper insight into
the mechanism which leads to the exceptionally
high activity of the Ru(0001) surface under high
pressure and oxidizing conditions [9,10], assuming
the so-called Langmuir–Hinshelwood mechanism
to be operational. The Langmuir–Hinshelwood
mechanism states that both reactants have to
adsorb on the catalyst surface prior to the actual
reaction step [11].
In this contribution we present the atomic geometries of Ru(0001)–(2×2)-3O along with the
structure generated by coadsorbing Cs atoms onto
the (2×2)-3O overlayer by using quantitative
LEED analyses. The coadsorption system
(2×2)-3O+1Cs provides a simple means to monitor the defect density of the (2×2)-3O phase,
which actually turned out to be quite low. We
were not able to prepare a mixed (2×2)-3O+CO
overlayer, presumably due to the presence of an
energy barrier in the entrance channel for chemisorption of the CO molecules onto the (2×2)-3O
surface. Even after long CO exposures at low
sample temperatures (>200 K ) no CO uptake
could be detected on the Ru(0001)–(2×2)-3O
surface. This interesting issue will be studied in
the future by molecular beam experiments.
Unpublished DFT calculations by Stampfl [12]
suggest an energy barrier of about 0.4 eV in the
entrance channel.
The experiments were conducted in a UHV
chamber (base pressure: 1×10−10 mbar) equipped
with a display-type four-grid LEED optics and
with standard facilities for surface cleaning and
characterization. Details of the experimental set-up
and sample preparation can be found elsewhere
[13]. The various oxygen overlayers with coverages up to 1 ML were prepared by backfilling
the chamber with NO at a pressure of
2
1×10−7 mbar. The sample temperature was kept
at 500 K so that after decomposition of NO into
2
adsorbed O and NO (the dissociative sticking
coefficient of NO is about unity) only oxygen
2
remained on the surface. Concomitant LEED
intensity and LEED spot profile measurements of
the (1, 0) beam (see Fig. 1) indicated the appear2
ance of several distinct O-phases exhibiting LEED
patterns with apparent (2×2) periodicity. First,
the (2×2)-O phase is developed which transforms
with progressing NO exposure into a (2×1)-O
2
phase and then into the (2×2)-3O phase. The
transition region between consecutive phases is
characterized by a precipitous intensity drop along
with a broadening of the (1, 0) beam profile.
2
Further NO exposure deteriorates the quality of
2
the (2×2) LEED pattern and finally the surface
exhibits a well-ordered (1×1)-O structure, which
was investigated in a previous paper [5]. The
optimum NO exposure for the preparation of the
2
(2×2)-3O structure is 16 L. The coadsorption
phase (2×2)-3O+Cs was prepared by post-deposition of Cs on the Ru(0001)–(2×2)-3O surface
which was evaporated from a carefully outgassed
dispenser getter source. The sample temperature
during Cs deposition was kept fixed at 200 K in
order to maintain the (2×2) periodicity. It should
be noted that annealing the sample to above 400 K
destroyed irreversibly the ordering of the mixed
O+Cs overlayer. The LEED pattern changed from
a (2×2) in a (앀7×앀7)R19.1° structure indicating
the formation of a CsO surface phase [14]. The
2
actual Cs coverage of 0.25 in the coadsorbate
phase was evaluated from the integral of the Cs
Fig. 1. The LEED intensity and the full-width half maximum
( FWHM ) of the (1/2 , 0) beam versus the applied NO expo2
sure. Clearly, three ordered (2×2)-phases are discernible:
(2×2)-1O, (2×1)-1O, and (2×2)-3O.
Y.D. Kim et al. / Surface Science 418 (1998) 267–272
thermal desorption ( TD) spectrum which was calibrated with respect to the integrated TD signal of
the pure (2×2)-Cs overlayer on Ru(0001). From
a mere inspection of the LEED pattern the quality
of ordering in the (2×2)-3O+Cs phase was judged
as being as good as that of the (2×2)-3O phase
(up to energies of 350 eV ).
LEED intensities as a function of the incident
energy of the electrons were collected at normal
incidence of the primary beam, keeping the sample
at a temperature of 1OO K. A computer-controlled
video camera was used to record spot intensities
from the LEED fluorescence screen of three integral-order and five [six] fractional-order beams
(energy range 50 to 350 eV ) of the (2×2)-3O
[(2×2)-3O+Cs] structure. LEED I/E curves were
computed by using the program code by Moritz
[15] and compared with the experimental LEED
I/E curves by employing a least-squares optimization algorithm [16–18], based on Pendry’s r-factor
R [19]. Both experimental LEED data sets are
P
significantly different, as quantified by an r-factor
of R =0.45, already demonstrating a good (2×2)
P
ordering of the strong Cs scatterers.
The optimum atomic geometry of the
Ru(0001)–(2×2)-3O phase, as retrieved from the
experimental LEED data, is summarized in Fig. 2.
The oxygen atoms arrange themselves in a honey-
Fig. 2. Top view (a) and side view (b) of the optimized atomic
geometry (as determined by LEED) of the Ru(0001)–
(2×2)-3O with O sitting in the hcp-hollow site. The arrows
indicate the direction of the displacements of the substrate
atoms with respect to the bulk-terminated positions. The distances are given in ångströms.
269
comb structure with vacancies determining the
(2×2) periodicity, which confirms the previously
proposed model based on HREELS and preliminary STM measurements [6 ]. Oxygen atoms
occupy the hcp site, i.e. the threefold hollow site
where an Ru atom in the second layer is directly
underneath. The Ru–O bond lengths turned out
to be 1.98 Å and 2.03 Å (averaged value: 2.00 Å),
which is very close to the values found for the
(2×2)-O and (2×1)-O phases (2.03 Å and 2.02 Å
respectively) [20,21]. A significant feature of the
(2×2)-3O surface structure is the observed
(averaged) expansion of the topmost Ru layer
spacing of 2.21 Å (bulk value 2.14 Å). This value
fits nicely into the trend of the topmost Ru layer
spacing as a function of the O-coverage as provided
by previous LEED structure analyses of the
(2×2)-O, (2×1)-O [20,21] and (1×1)-O phases
[5] and which is illustrated in Fig. 3. A slight
expansion of the topmost Ru layer (2.22 Å) was
also found by DFT calculations [3,4].
The oxygen-induced local rearrangements of
substrate atoms are small. Most notably a buckling
of 0.08 Å in the second Ru layer occurs. The
second-layer Ru atom moves towards the three
first-layer Ru atoms directly above, which are only
coordinated to two O atoms. Within this four-Ru
cluster the layer distance (2.15 Å) is almost bulklike. The first-layer Ru atom which is attached to
three O atoms is slightly displaced inwardly by
0.02 Å. The layer distance (2.21 Å) between this
first-layer Ru atom and the three attached Ru
atoms in the second layer directly below is then
Fig. 3. The layer spacing d between the first and second Ru
12
layers as a function of the O-coverage. The value found for the
(2×2)-3O phase fits nicely into the general trend that the value
of d increases with increasing O-coverage.
12
270
Y.D. Kim et al. / Surface Science 418 (1998) 267–272
still substantially expanded with respect to the
bulk value.
The oxygen atoms move by 0.05±0.07 Å radially away from the high-symmetry hcp sites
towards the bridge sites, presumably to optimize
the electronic environment for the O-atoms. Note
that in the unrelaxed (2×2)-3O overlayer, one Ru
atom is coordinated to three O atoms whereas the
other three Ru atom in the unit cell are only
attached to two O atoms. From a recent LEED
analysis of the O-phases on an Ru(101: 0) surface
it was concluded that oxygen does not like to share
Ru atoms with other O atoms [22]. Therefore, it
seems plausible to attribute the observed relaxation
of the O atoms to this effect. The LEED intensity
spectra for the best fit geometry of the
Ru(0001)–(2×2)-3O surface are shown in Fig. 4.
The agreement between experimental and calculated LEED intensity data is quantified by an rfactor of R =0.25. For comparison, the best rP
factor reached for a model with the O atoms sitting
in fcc positions was 0.72. Also, a structure model
Fig. 4. The comparison between experimental and calculated
LEED-IV curves for the best-fit geometry of the (2×2)-3O
phase on Ru(0001). The overall r-factor is R =0.25.
P
for which the O atoms sit randomly with varying
concentrations in hcp positions within the (2×2)
unit cell was not able to improve the fit between
experimental and calculated data.
The structure analysis of the coadsorption phase
Ru(0001)–(2×2)-(3O+Cs) reveals that Cs atoms
fill up the vacant hcp-sites in the (2×2)-3O net,
as shown in Fig. 5. Other adsorption sites for Cs,
i.e. the fcc and the on-top position, result in rfactor values of 0.67 and 0.63 respectively, which
are significantly worse than for the case of hcp site
adsorption. The agreement between experimental
and calculated LEED intensity data for the best
fit geometry can be judged from Fig. 6; the overall
r-factor between these data is R =0.31. This
P
agreement is similar to that achieved for the pure
Cs-(2×2) phase on Ru(0001) [23].
Although the Ru–O layer spacing increases quite
substantially on coadsorption of Cs atoms, namely
from 1.21 Å in the (2×2)-3O phase to 1.28 Å in
the (2×2)-3O+Cs phase, the Ru–O bond length
does not change appreciably. The Ru–O bond
length is 2.03 Å in the coadsorbate phase, i.e. the
effective O-radius remains almost constant: the
O-radius of 0.65 Å in the (2×2)-3O phase changes
to 0.68 Å in the (2×2)-3O+Cs phase.
The mere existence of a well-ordered
Fig. 5. Top view and side view of the atomic geometry of the
Ru(0001)–(2×2)-3O+Cs with O and Cs sitting in the hcphollow sites. The arrows indicate the direction of the displacements of the substrate atoms with respect to the bulk-terminated
positions. The distances are given in ångströms.
Y.D. Kim et al. / Surface Science 418 (1998) 267–272
Fig. 6. The comparison between experimental and calculated
LEED-IV curves for the best-fit geometry of the (2×2)-3O+Cs
phase on Ru(0001). The overall r-factor is R =0.31.
P
(2×2)-3O+Cs surface indicates that the array of
vacancies formed by the oxygen overlayer is wellordered since the Cs atoms were found to reside
exclusively in those positions. As Cs is a much
stronger scatterer than O, a high defect density in
the (2×2) vacancy network would have inevitably
resulted in a high background LEED intensity at
higher energies which, however, was not observed
in the experiment. The main structural difference
between the pure Cs-(2×2) phase and that of the
mixed (2×2)-3O+Cs overlayer consists in the
adsorption site of Cs. Whereas Cs atoms occupy
on-top position on the clean surface [23], Cs atoms
prefer the hollow sites on the (2×2)-3O precovered
Ru(0001) surface. This might be related to the
more pronounced corrugation of the potential
energy surface which is imposed by the (2×2)-3O
overlayer. Yet, from a purely steric point of view
the Cs atoms have equally good access to the
on-top positions. One might even have guessed
that for the electropositive Cs atoms the on-top
sites would be more favorable than hcp hollow
271
sites due to the closer proximity of the strongly
electronegative O atoms.
In order to quantify the changes in the local
Cs–Ru bond configuration due to the presence of
oxygen, the effective Cs radius, i.e. the Cs–Ru
bond length minus the metallic Ru radius, was
evaluated. The calculated value of 2.16 Å for the
(2×2)-3O-Cs phase is identical to that found for
the (앀3×앀3)R30°-Cs phase on Ru(0001), i.e.
there is no strong effect of the oxygen coadsorbate
on the Cs adsorption geometry. Recall that the
effective radii of alkali metal atoms do not change
with coverage as long as the coordination number
is conserved [24]. Preliminary metastable
de-excitation spectroscopy (MDS) measurements
indicate that the electronic properties, in terms of
the metallicity of the Cs-(2×2) overlayer, are
almost identical for the clean and the (2×2)-3O
precovered Ru(0001) surfaces. Apparently the
electronic interaction between O and Cs in the
(2×2)-3O+Cs phase is weak. Upon annealing
the (2×2)-3O+Cs surface the metallic character
of the overlayer is lost, as evidenced by MDS, and
instead a CsO surface species is formed, as evi2
denced by the appearance of the (앀7×앀7)
R19.1° LEED pattern [14].
Quite different results were obtained for
the coadsorption system Ru(0001)–(앀3×앀3)
R30°-O+Cs [25]. LEED structure analysis indicates that the oxygen radius increases by about
0.12 Å, while that of Cs decreases by 0.06 Å in
comparison with the radii found for the pure phase
(2×2)-O [20,21] and (앀3×앀3)R30°-Cs [23].
Both changes are indicative of a net charge transfer
from the electropositive Cs to electronegative O
which is likely mediated by the Ru substrate. MDS
measurements indicate that this transformation is
accompanied by an electronic transition from metallic to non-metallic [26 ]. For the present coadsorption system this kind of charge transfer cannot
be inferred from either the structural data or
the electronic data. The reason for this might
be connected with the metastability of the
(2×2)-3O+Cs phase, as annealing the sample to
400 K destroys the (2×2) structure irreversibly.
The LEED pattern turns into a (7×7)R19.1°
structure which is indicative of a CsO surface
2
species [14].
272
Y.D. Kim et al. / Surface Science 418 (1998) 267–272
In conclusion, we have presented surface crystallographic data of the (2×2)-3O and the
(2×2)-3O-Cs overlayers on the Ru(0001) surface.
The (2×2)-3O phase consists of a (2×2) vacancy
network in which CO molecules do not adsorb at
temperatures higher than 200 K whereas Cs atoms
do. The first Ru layer separation (2.21 Å) is slightly
expanded, which is consistent with the observed
trend that the topmost Ru layer distance increases
with O-coverage. Coadsorbing Cs atoms do not
affect the (2×2)-3O overlayer, which might be
related to a weak interaction between O and Cs
and to the metastable character of the mixed
(2×2)-3O+Cs phase.
Acknowledgements
Cathy Stampfl is acknowledged for stimulating
discussions and for providing us with the unpublished optimum structural parameters of the
Ru(0001)–(2×2)-3O phase as obtained by recent
DFT calculations [3,4,12]. Ari Seitsonen is
acknowledged for fruitful discussions.
References
[1] P.R. Watson, M.A. Van Hove, K. Hermann, Atlas of surface structures—based on the NIST surface structure database (SSD), Journal of Physical and Chemical Reference
Data 1A and 1B (1994) 1ff.
[2] T.E. Madey, H.A. Engelhardt, D. Menzel, Surf. Sci. 48
(1975) 304.
[3] C. Stampfl, M. Scheffler, Phys. Rev. B 54 (1996) 2868.
[4] C. Stampfl, M. Scheffler, unpublished structural results of
the (2×2)-3O phase on Ru(0001).
[5] C. Stampfl, S. Schwegmann, H. Over, M. Scheffler, G. Ertl,
Phys. Rev. Lett. 77 (1996) 3371.
[6 ] K.L. Kostov, M. Gsell, P. Jacob, T. Moritz, W. Widdra,
D. Menzel, Surf. Sci. 377 (1997) L138.
[7] A. Böttcher, H. Niehus, S. Schwegmann, H. Over, G. Ertl,
J. Phys. Chem. 101 (1997) 11185.
[8] B.E. Hayden, K. Kretzschmar, A.M. Bradshaw, Surf. Sci.
125 (1983) 366.
[9] C.H.F. Peden, D.W. Goodman, J. Phys. Chem. 90
(1986) 1360.
[10] C.H.F. Peden, D.W. Goodman, M.D. Weisel, F.M.
Hoffmann, Surf. Sci. 253 (1991) 44.
[11] T. Engel, G. Ertl, Adv. Catal. 28 (1979) 1.
[12] C. Stampfl, unpublished DFT calculations of CO adsorption onto the Ru(0001)–(2×2)-3O surface.
[13] H. Over, H. Bludau, M. Skottke-Klein, G. Ertl, W. Moritz,
C.T. Campbell, Phys. Rev. B 45 (1992) 8638.
[14] H. Bludau, H. Over, T. Hertel, M. Gierer, G. Ertl, Surf.
Sci. 342 (1995) 134.
[15] W. Moritz, J. Phys. C: 17 (1983) 353.
[16 ] G. Kleinle, W. Moritz, G. Ertl, Surf. Sci. 226 (1990) 119.
[17] H. Over, U. Ketterl, W. Moritz, G. Ertl, Phys. Rev. B 46
(1992) 15438.
[18] M. Gierer, H. Over, W. Moritz, unpublished data.
[19] J.B. Pendry, J. Phys. C: 13 (1980) 937.
[20] H. Pfnür, G. Held, M. Lindroos, D. Menzel, Surf. Sci. 220
(1990) 43.
[21] M. Lindroos, H. Pfnür, G. Held, D. Menzel, Surf. Sci. 222
(1989) 451.
[22] S. Schwegmann, A.P. Seitsonen, V. De Renzi, H. Dietrich,
H. Bludau, M. Gierer, H. Over, K. Jacobi, M. Scheffler,
G. Ertl, Phys. Rev. B 57 (1998) 15487.
[23] H. Over, H. Bludau, M. Skottke-Klein, G. Ertl, W. Moritz,
C.T. Campbell, Phys. Rev. B 45 (1992) 8638.
[24] R.D. Diehl, R. McGrath, Surf. Sci. Rep. 23 (1996) 43.
[25] H. Over, H. Bludau, M. Skottke-Klein, W. Moritz, G. Ertl,
Phys. Rev. B 46 (1992) 4360.
[26 ] A. Böttcher, A. Morgante, R. Grobecker, T. Greber, G.
Ertl, Phys. Rev. B 49 (1994) 10607.