Adsorption and NiO(1 0 0) formation by atomic and molecular oxygen on Ni(1 1 0)

Article abstract of DOI:10.1016/j.susc.2008.07.004

The structure of the ordered p(2 × 1) and p(3 × 1) phases of adsorbed oxygen as well as the formation of ultrathin NiO(1 0 0) layers on a Ni(1 1 0) single crystal are investigated by grazing scattering of fast hydrogen atoms. Via ion beam triangulation based on the detection of the number of emitted electrons, we obtain direct information on the structure of oxygen adsorbates and ultrathin nickel oxide layers. For oxidation using atomic instead of molecular oxygen, the gas exposure can be reduced by almost two orders of magnitude. We compare the experimental results with computer simulations based on classical projectile trajectories for grazing scattering of fast hydrogen atoms and test structure models for oxygen adsorbed on Ni(1 1 0) and NiO(1 0 0).

Full text of DOI:10.1016/j.susc.2008.07.004

Surface Science 602 (2008) 2808–2815
Contents lists available at ScienceDirect
Surface Science
journal homepage: www.elsevier.com/locate/susc
Adsorption and NiO(100) formation by atomic and molecular oxygen on Ni(110)
*
M. Busch , D. Blauth, H. Winter
Humboldt-Universität zu Berlin, Institut für Physik, Newtonstraße 15, D-12489 Berlin, Germany
a r t i c l e i n f o
a b s t r a c t
Article history:
The structure of the ordered p(2 ꢀ 1) and p(3 ꢀ 1) phases of adsorbed oxygen as well as the formation of
ultrathin NiO(100) layers on a Ni(110) single crystal are investigated by grazing scattering of fast hydro-
gen atoms. Via ion beam triangulation based on the detection of the number of emitted electrons, we
obtain direct information on the structure of oxygen adsorbates and ultrathin nickel oxide layers. For oxi-
dation using atomic instead of molecular oxygen, the gas exposure can be reduced by almost two orders
of magnitude. We compare the experimental results with computer simulations based on classical pro-
jectile trajectories for grazing scattering of fast hydrogen atoms and test structure models for oxygen
adsorbed on Ni(110) and NiO(100).
Received 17 March 2008
Accepted for publication 3 July 2008
Available online 11 July 2008
Keywords:
Oxygen adsorption
Nickel oxide
Atomic oxygen
Ion beam triangulation
Ion–solid interactions
Ó 2008 Elsevier B.V. All rights reserved.
1. Introduction
above the topmost layer of Ni atoms was estimated to 0.25–0.3 Å
[11,12,14] and 0.59 Å [15]. For the primarily designated missing-
Studies on chemisorption at metal surfaces are important for
the understanding of heterogeneous catalysis. In particular, the
interaction of oxygen with surfaces of 3d transition metals, e.g.
nickel and copper, have been extensively studied [1]. On the
Ni(110) surface oxygen chemisorption induces a sequence of
adsorbate superstructures at substrate temperatures T P 300 K.
With increasing oxygen exposure, one observes: p(3 ꢀ l)i for an
row structure model a long-range transport mechanism is in-
volved: Ni atoms are removed either from step edges or by the cre-
ation of troughs on the terraces and react with the O adatoms,
resulting in the formation of –Ni–O– rows. Therefore, this recon-
struction with respect to the nucleation and growth mechanism
is of an added-row rather than of a missing-row type.
The reconstruction of the saw-tooth type is due to a displace-
ment of Ni [001] rows from the topmost layer of the unrecon-
structed Ni(110) surface on top of this surface layer. Hence, the
oxygen coverage of
H
= 1/3 ML, p(2 ꢀ l) for
H = 1/2 ML, and
p(3 ꢀ 1)f for = 2/3 ML (1 ML of oxygen atoms on the Ni(110)
H
surface is equivalent to 1.14 ꢀ 1015 atoms/cm2). For higher oxygen
exposure the sequence proceeds with a p(9 ꢀ 5) ‘‘suboxide” struc-
ture and ends up in the formation of disordered NiO(100) as ob-
served from LEED (low energy electron diffraction) studies [2–5].
The formation of the superstructures p(3 ꢀ l)i, p(2 ꢀ l), and
p(3 ꢀ 1)f involves an oxygen-induced reconstruction of the
Ni(110) surface. There was some debate, whether these recon-
structions are of the missing-row, added-row, or saw-tooth type.
For these types, the superstructures are due to the formation of
low-coordinated –Ni–O– rows along the [001] direction, analo-
gous to the known O-induced reconstructions on Cu(110) [6–8],
Ag(110) [9], and Cu(100) [10]. In particular, the –Ni–O– rows
are arranged one for each third, one for each second, and two for
saw-tooth surface has only one mirror plane along the ½110ꢁ direc-
ꢀ
tion, whereas the added-row surface has two mirror planes along
ꢀ
the ½110ꢁ and [001] directions.
Several investigations using Auger electron spectroscopy (AES)
induced by surface channeled ions [11], surface extended X-ray-
absorption fine-structure (SEXAFS) [15], scanning tunneling
microscopy (STM) [16], and electron energy loss spectroscopy
(EELS) [17] favoured the saw-tooth model. However, the predom-
inant part of investigations which used low energy ion scattering
(LEIS) [12,13], He diffraction [14], Rutherford backscattering
(RBS) [18], structural studies including LEED [19], high resolution
electron energy loss spectroscopy (HREELS) [20], dynamic-visuali-
zation scanning tunneling microscopy (DVSTM) [21], and angle re-
solved ultraviolet photoemission spectroscopy (ARUPS) [22]
support an added-row model.
ꢀ
each second ½110ꢁ lattice spacing, corresponding to p(3 ꢀ l)i,
p(2 ꢀ 1), and p(3 ꢀ 1)f superstructures, respectively (Fig. 1b and
1c). For these three structures the O adatoms are localized at
long-bridge sites of the Ni lattice. The height of the O adatoms
A third model for the p(2 ꢀ 1) superstructure was proposed by
Masuda et al. [23] for room and higher temperatures using LEED
and HREELS. The experimental results were interpreted in terms
of oxygen atoms adsorbed at room temperature in two different
stages. An unreconstructed p(2 ꢀ 1) superstructure with oxygen
* Corresponding author. Tel.: +49 30 2093 7697; fax: +49 30 2093 7899.
E-mail address: marco.busch@physik.hu-berlin.de (M. Busch).
0039-6028/$ - see front matter Ó 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.susc.2008.07.004

M. Busch et al. / Surface Science 602 (2008) 2808–2815
2809
a
[01]
[12]
[11]
b
[01]
[11]
[43]
[21]
[21]
[31]
[41]
[41]
[10]
[10]
c
[01]
[11]
d
[11]
[43]
[01]
[21]
[31]
[21]
[41]
[10]
[10]
Fig. 1. Structural models of (a) clean Ni(110) surface, oxygen superstructure (b) p(2 ꢀ 1) and (c) p(3 ꢀ 1)f as added-row type with long-bridge adsorption sites and (d)
p(2 ꢀ 1) superstructure without reconstruction and short-bridge adsorption sites on Ni(110). Gray bars indicate channels along low-index directions [uv] in the primitive
rectangular surface lattice of Ni(110). Widths of gray bars indicate widths of axial channels. Full circles: nickel (gray) and oxygen (black) atoms in topmost layer, open circles:
nickel atoms in second layer. Grid lines denote Ni lattice. For structural models see also Refs. [1,12,14,18–23].
atoms being located on the short-bridge sites (cf. Fig. 1d) and the
added-row type with oxygen atoms at long-bridge sites. For
annealing of the adsorbate layer after formation at room tempera-
ture and for adsorption at about 540 K only the reconstructed
added-row p(2 ꢀ 1) superstructure is present. This behaviour is
accompanied by a sharpening of the LEED spots of the p(2 ꢀ 1)
pattern.
NiO(100) square unit cell (lattice constant gNiO = 4.17 Å and O–O
pffiffiffi
nearest-neighbor spacing gNiO
=
2 ¼ 2:95 Å) is somewhat dis-
torted, with the side length varying between 2.8 and 3.1 Å [21].
The thickness of the islands corresponds to about 3–4 layers of
NiO(100) [24,27].
In this paper we report on results for the structural properties of
the oxygen p(2 ꢀ 1) and p(3 ꢀ 1)f superstructures on Ni(110) as
well as the formation of NiO(100) by exposure to atomic and
molecular oxygen. Making use of ion beam triangulation (IBT),
based on the observation of electron emission induced by grazing
scattering of ions or atoms, we obtained real-space information
on the geometrical arrangement of atoms in the topmost surface
layer. In this work we present first IBT studies on oxygen adsorbate
structures on a metal and the formation of a metal oxide film [28].
Of particular interest here is the effect of atomic and molecular
oxygen exposure on the adsorption process. We compare the
experimental results with computer simulations based on classical
trajectories for grazing scattering of atoms.
For oxygen exposure after the formation of the p(3 ꢀ 1)f recon-
struction, a p(9 ꢀ 5) structure evolves at substrate temperatures of
about 400 K. From DVSTM investigations [21] it was concluded,
that the p(9 ꢀ 5) structure is indeed commensurate and well or-
dered within domains of about 100 ꢀ 100 Å2. For the p(9 ꢀ 5)
structure, Ni atoms are locally rearranged significantly compared
to their positions in the Ni fcc lattice. Therefore, the p(9 ꢀ 5) struc-
ture can be considered as a genuine oxide structure [24], and as a
precursor for growth of NiO(100) at further oxygen exposure. In
the early stage of the NiO(100) formation, oxide islands nucleate
and grow on top of the p(9 ꢀ 5) structure. Initially these islands
nucleate at the step edges and then also on the terraces. The is-
lands grow to a size, which is comparable to the p(9 ꢀ 5) domains
(100 ꢀ 100 Å2) [21]. Continued oxidation proceeds via nucleation
of further islands, which eventually coalesce, rather than by lateral
growth of already existing NiO islands [25,26]. NiO grows at sub-
strate temperatures of about 400 K almost epitaxially, with the
NiO(100) planes parallel to the Ni(110) surface. However, the
2. Experiment
The experiments were performed in an ultrahigh vacuum (UHV)
chamber at a base pressure of 5 ꢀ 10ꢂ11 mbar, attached via differ-
ential pumping stages to the beam lines of two electrostatic ion
accelerators with energies from 5 to 23 kV and from 30 to

2810
M. Busch et al. / Surface Science 602 (2008) 2808–2815
400 kV. The Ni(110) single crystal disk was prepared by cycles of
grazing sputtering with 50 keV Ar+ ions and subsequent annealing
to 850 K for 10 min. The cleanness and flatness of the surface was
checked by low energy electron diffraction (SPA-LEED, Omicron),
Auger electron spectroscopy (AES), and polar angular distributions
of scattered ions [29].
within a small stainless steel tube of 50 mm diameter and
50 mm length with a ring electrode as a focusing element and a
highly transparent grid (transparency ꢃ95% at a potential of about
1.5 kV). The distance between grid and target was about 60 mm.
Electrons emitted from the target surface are collected by the
resulting electric field and accelerated towards the entrance aper-
ture of the SBD. A small axial magnetic field of some Gauss be-
tween target and SBD improves the collection efficiency for
emitted electrons.
The exposure (1 L = 1 ꢀ 10ꢂ6 Torr s = 1.33 ꢀ 10ꢂ6 mbar s) of the
well-prepared and atomically clean Ni substrate to molecular
oxygen for the formation of the oxygen p(2 ꢀ 1) and p(3 ꢀ 1)f
superstructures was performed at
a
partial pressure of
5 ꢀ 10ꢂ9 mbar and a substrate temperature of T = 460 K. The inves-
tigations on the NiO formation by exposure to molecular oxygen
were performed at partial pressures of 5 ꢀ 10ꢂ8 mbar (630 L), 1 ꢀ
10ꢂ7 mbar (40 ꢂ 120 L) and 5 ꢀ 10ꢂ7 mbar (300 L) at T = 420 K.
For atomic oxygen the partial pressure was 1 ꢀ 10ꢂ8 mbar.
3. Results and discussion
3.1. Exposure to molecular oxygen
Fig. 2 shows LEED patterns for clean Ni(110) and the oxygen
p(2 ꢀ 1) and p(3 ꢀ 1)f superstructures on Ni(110). The primitive
rectangular surface lattice is always identified. The two superstruc-
tures were produced via exposure to molecular oxygen of 0.8 L and
7 L, respectively, at a pressure of 5 ꢀ 10ꢂ9 mbar and a substrate
temperature of T = 460 K [16,20,21]. The elevated temperature is
applied, since the LEED patterns for the superstructures formed
at room temperature are somewhat diffuse, indicating the exis-
tence of small domains. The quality of the LEED spots is improved
by annealing the surface to above 400 K after the formation of the
superstructures or by adsorption of oxygen above this temperature
[21].
Fig. 3 shows SBD counts, normalized to the total emission
events, versus azimuthal angle of incidence h for grazing scattering
under a polar angle of incidence /in = 1.2° of 15 keV hydrogen
atoms from the surface of clean Ni(110) and of the p(2 ꢀ 1) and
p(3 ꢀ 1)f superstructures. The data shows pronounced minima, if
the direction of the incident projectile beam coincides with low-in-
dex directions in the surface plane (axial surface channeling [36]).
For the clean Ni(110) surface the [uv] directions in the primitive
rectangular surface lattice are identified up to the 4th order (i.e.
[41] and [14]). In the channels along low-index directions, the tra-
jectories of projectiles differ from those for planar symmetry met
for a random azimuthal orientation of the surface plane. Then re-
gions with higher electron densities and subsurface layers are
probed, resulting in an enhanced number of electrons emitted
per projectile [30].
The source for atomic oxygen, a commercial available thermal
cracker (Model TC 50, Oxford Applied Research), exploits thermal
dissociation on a ThO2-coated filament. At relatively low tempera-
tures (<1000 °C) the fraction of oxygen atoms is about 60%.
For IBT, we investigated the electron emission induced by graz-
ing scattering of ions or atoms by means of a surface barrier detec-
tor (SBD) [30], which is a passivated implanted planar silicon (PIPS)
detector (Canberra BKLTPD-150-21-300AM) with an additional
thin protective aluminium layer of 200 lm. The protective layer re-
duces the noise signal from visible light substantially. The overall
energy resolution of the SBD is about 8 keV at a bias voltage of
60 V. The compact detector operates on high potential of 20 kV.
Signal and ground is fed via a flexible cable insulated with ceramic
beads and a high-voltage vacuum feed through to a preamplifier
and amplifier on a high-voltage platform outside the vacuum
chamber. The pulse height is proportional to the number of emit-
ted electrons per scattering event and allows one to study electron
number spectra (ENS) [31]. The output signal of the amplifier is fed
into an analog-digital-converter (ADC) and a data acquisition unit
via an insulation transformer. The SBD count rate amounts to
about 104 per second which is equivalent to a current of incident
projectiles of less then 10ꢂ15 A. Hence, we can exclude radiation
damage or beam-induced reconstruction of the surface triggered
by the incident projectile beam. The other advantages in the detec-
tion of electrons instead of the target current as performed in an
early variant of this method [28,32] are an efficient self-normaliza-
tion of data and an enhanced surface sensitivity (see also chapter
3).
The structure of the IBT curves shown in Fig. 3 results from a
detection scheme where the discriminator level of the SBD is set
to a pulse height interval equivalent to about 1–5 emitted elec-
trons. This signal is referred to the overall detector counts which
provides a stable normalization widely free from fluctuations of
The design of the detector is similar to former ones [28,29,31–
35], however, the setup is more compact and can thus be easier
positioned within an UHV chamber. The SBD detector is placed
Fig. 2. LEED pattern of the clean Ni(110) surface (left), the oxygen p(2 ꢀ 1) superstructure (middle) and the oxygen p(3 ꢀ 1)f superstructure (right) on Ni(110) at electron
energies of 130 eV (left and middle panel) and 110 eV (right panel).

M. Busch et al. / Surface Science 602 (2008) 2808–2815
2811
sequence, the signal for the [31] channel is very weak in the IBT
curve of the p(2 ꢀ 1) structure. The same holds for the [13] channel
at an azimuthal angle of h = 13.3° for both types of adsorption sites,
whereas the [32] channel at an azimuthal angle of 46.8° is only
blocked for adsorption on long-bridge sites.
For the broader [11] channel axial channeling is still present for
the p(2 ꢀ 1) and p(3 ꢀ 1)f superstructures with O adatoms on both
long-bridge and short-bridge adsorption sites. This holds for the
[10] channel and long-bridge as well as for the [01] channel and
short-bridge adsorption sites.
p(3x1)
1.0
1.0
p(2x1)
In the IBT curve for the p(3 ꢀ 1)f superstructure (cf. Fig. 3, upper
curve), the dips along the [31] and [32] directions are present
again, because no O adatoms are located within these channels.
A further feature for the p(3 ꢀ 1)f structure is the appearance of a
dip at an azimuthal angle of h = 76.7° which is attributed to the
[61] channel (d[61] = 0.57 Å). In the IBT curves for the p(2 ꢀ 1)
superstructure and clean Ni(110) this dip is not observed. The
presence of the [61] dip for the p(3 ꢀ 1) structure results from
the enhanced axial potential along [61] strings owing to a doubled
atom density.
We analyzed the experimental IBT curves in terms of computer
simulations based on a fourth order Runge-Kutta algorithm [37],
where projectile trajectories are calculated from Newton’s equa-
tion of motion in the collective potential of target atoms. We made
use of a Molière potential [38] with corrected Firsov screening
length as proposed by O’Connor and Biersack [39]. The number
of electrons emitted per projectile is deduced from a Monte Carlo
integration of the electron production rate. This rate is assumed
to be proportional to the electron density described by an expo-
nential decay towards vacuum. Each target atom is considered to
have an electron density ꢃexp(ꢂr), with r being the distance (in
a.u.) between projectile and target nucleus.
Ni(110)
1.0
0.8
0.6
0.4
0
15
30
45
60
75
90
Azimuthal Angle [deg]
Fig. 3. Normalized SBD counts versus azimuthal angle of incidence h for grazing
scattering of 15 keV hydrogen atoms from the clean Ni(110) surface and from the
oxygen p(2 ꢀ 1) and p(3 ꢀ 1)f superstructures on Ni(110) under a polar angle of
incidence of /in = 1.2°. Vertical lines indicate low-index directions [uv] in the
primitive rectangular surface lattice of Ni(110).
the projectile beam. The normalized intensity for random azimuth
is set to one.
c
1.0
The maximum in the electron number spectra (ENS) for random
azimuth is at about 3–4 electrons. For scattering along a low-index
direction, an enhancement of the mean number of emitted elec-
trons is accompanied by a decrease of events with low electron
numbers and results in minima (dips) of the normalized signal.
Scattering processes at the topmost layer produce lower electron
numbers than penetration into the subsurface layers, where larger
electron densities and longer trajectories lead to enhanced proba-
bilities for the emission of electrons. Thus, the change in the signal
for low numbers reflects the structure at the topmost surface layer.
Penetration of the topmost surface layer for a given normal projec-
tile energy E\ = E0 sin2 /in (E0 = projectile energy) is related to the
width of the axial channels d[uv] and the axial scattering potential.
This potential depends on the distance between atoms along
strings and d[uv]. For narrow axial channels the probability for pen-
etration of the surface layer and the change of the SBD signal are
reduced.
b
1.0
a
1.0
0.8
0.6
0.4
The IBT curve of the clean Ni(110) surface (cf. Fig. 3, lower
curve) shows prominent dips along [10], [01], and [11] as broadest
pffiffiffi
axial channels (d[10] = g = 3.52 Å, d ¼ g= 2 ¼ 2:49 Å, and d
¼
½01ꢁ
½11ꢁ
pffiffiffi
g= 3 ¼ 2:03 Å). The tiny peaks within the dips for [10] and [01]
are caused by strings of atoms in the second layer positioned in
the middle of these channels. The shape of the dip for the [11]
channel can be explained by projectiles which penetrate the top-
most layer and are focused and channeled within the bulk.
0
15
30
45
60
75
90
Azimuthal Angle [deg]
For the IBT curve of the p(2 ꢀ 1) structure (cf. Fig. 3, middle
curve) some dips at higher indexed directions are significantly re-
duced. As example we mention the [31] direction at an azimuthal
angle of h = 64.8°. In the structure models with long-bridge and
short-bridge adsorption sites, the O adatoms are positioned in
the middle of the [31] channel and blocking this channel. As a con-
Fig. 4. Calculated triangulation curves (solid) obtained by computer simulations
based on classical projectile trajectories for grazing scattering (/in = 1.2°) of 15 keV
hydrogen atoms from the clean Ni(110) surface (a) and from the oxygen p(3 ꢀ 1)f
superstructure with added-row (b) and saw-tooth reconstruction (c) and long-
bridge adsorption sites. Experimental data (dashed curves) are also shown (vertical
lines, cf. Fig. 3).

2812
M. Busch et al. / Surface Science 602 (2008) 2808–2815
Thermal vibrations of the lattice atoms are included in the sim-
ulations by Debye temperatures # for the clean Ni(110) surface
d
c
b
a
1.0
1.0
1.0
[40] for the parallel (# = 360 °C) and perpendicular (#\ = 225 °C)
k
motion. The thermal displacements of an individual atom are as-
sumed to be independent of the displacements of neighboring
atoms, since correlation effects on IBT curves are negligible [41].
In Figs. 4 and 5 we show calculated and experimental (cf. Fig. 3)
triangulation curves for grazing scattering (/in = 1.2°) of 15 keV
hydrogen atoms from the clean Ni(110) surface and from the oxy-
gen p(2 ꢀ 1) and p(3 ꢀ 1)f superstructures on Ni(110). The calcula-
tion of the triangulation curves for the p(3 ꢀ 1)f superstructure is
based on the added-row (cf. Fig. 1c) and saw-tooth reconstruction
with long-bridge adsorption sites. For the p(2 ꢀ 1) superstructure
we assumed oxygen adsorption on long-bridge sites in the
added-row (cf. Fig. 1b) and saw-tooth reconstruction and the unre-
constructed surface and additionally the oxygen adsorption on
short-bridge sites of the unreconstructed Ni lattice (cf. Fig. 1d).
Fig. 4 shows fair agreement between experimental and calcu-
lated IBT curves for the clean Ni(110) surface and the p(3 ꢀ 1)f
superstructure for both the added-row and saw-tooth reconstruc-
tion. This holds for the occurrence of the same number of dips at
the same azimuthal angles of incidence and the relative ratios of
the dip signal in the experimental and simulated IBT curves. Devi-
ations in the widths of dips for some channels are ascribed to sim-
plifications in the description on the electron production and
emission processes. Since the simulated IBT curves for the added-
row and saw-tooth reconstruction are similar, we cannot distin-
guish between these two structure models for the p(3 ꢀ 1)f
superstructure.
1.0
0.9
0.8
0.7
0.6
0.5
For the p(2 ꢀ 1) superstructure, however, we reveal a clear dif-
ference between the experiment and simulated IBT curves for a
reconstructed surface with added Ni rows in the topmost layer
and long-bridge adsorption sites. A comparison is shown in Fig.
5a for the added-row and in Fig. 5b for the saw-tooth reconstruc-
tion. The experimental dips for the [11] channel at an azimuthal
angle of h = 35.3° and for the [12] channel at h = 19.5° are not pres-
ent in the simulated IBT curve for the added-row reconstruction.
The dip for the [11] channel in the simulated IBT curve for the
saw-tooth reconstruction (cf. Fig. 5b) is less pronounced compared
to the experiment. Furthermore, the dip for the [10] channel is sig-
nificantly reduced compared to the experimental IBT curve for
both the added-row and the saw-tooth reconstruction. On the
other hand, the simulated IBT curves for these two structure mod-
els show a pronounced dip for the [41] at h = 70.5° channel and a
slightly smaller dip for the [43] channel at h = 43.3°, which is not
in accord with the experimental data.
For an unreconstructed p(2 ꢀ 1) superstructure with long-
bridge adsorption sites additional dips are present in the simulated
IBT curve (cf. Fig. 5c); in particular for [11] and [12] channels. For
the reconstructed and the unreconstructed p(2 ꢀ 1) superstructure
with long-bridge adsorption sites the O adatoms are exactly posi-
tioned in the middle of the [11] and [12] channels. But for the
unreconstructed surface, channeling along these channels is still
possible, which results in dips in the simulation (cf. Fig. 5c). Main
reason is the higher atomic density along strings of Ni atoms for
the [11] and [12] channels in the reconstructed surface compared
to the unreconstructed surface. For illustration, we have listed the
modifications of the widths and atomic string densities of several
channels in the topmost layer of the p(2 ꢀ 1) superstructure for
the different structure models in Table 1. For the reconstructed
superstructures the width and the atomic string density along
the [11] and [12] channels, which are formed by strings of Ni
atoms and of O atoms only, are reduced by a factor of two com-
pared to the clean Ni(110) surface. The string density of Ni atoms
along the [11] and [12] channels remains unchanged within the
unreconstructed p(2 ꢀ 1) superstructure. Only the string density
0
15
30
45
60
75
90
Azimuthal Angle [deg]
Fig. 5. Calculated triangulation curves (solid) obtained by computer simulations
based on classical projectile trajectories for grazing scattering (/in = 1.2°) of 15 keV
hydrogen atoms from the oxygen p(2 ꢀ 1) superstructure with added-row (a) and
saw-tooth reconstruction (b) with long-bridge adsorption sites and from the
unreconstructed p(2 ꢀ 1) superstructure with long-bridge (c) and short-bridge
adsorption sites (d). Experimental data (dashed curves) are also shown (vertical
lines, cf. Fig. 3).
of the O atoms along these channels is reduced. The presence of
dips for the [31], [32], [23], and [13] channels in the simulation
for the unreconstructed p(2 ꢀ 1) superstructure with long-bridge
adsorption sites is not in accord to the experiment. It is possible
to change the dip heights in the simulated IBT curves by variations
of the polar angle of incidence /in, the normal projectile energy E\,
and the potential between projectiles and target atoms. However,
the intensity ratios of the dips are not changed. The ratio of the
dip intensities of the [31] and [41] channels in the simulation
for the unreconstructed p(2 ꢀ 1) superstructure with long-bridge
adsorption sites cannot be reconciled with the experiment (cf.
Fig. 5c).
We find a better correspondence between experiment and sim-
ulation for an unreconstructed p(2 ꢀ 1) superstructure with short-
bridge adsorption sites (cf. Fig. 5d). In the simulation for this struc-
ture the dips appear at the correct angles with adequate intensity.
In comparison to structure models with long-bridge adsorption
sites, the widths of the channels [10] and [01] do not change for
the unreconstructed p(2 ꢀ 1) superstructure with short-bridge
adsorption sites. Hence, the dip ratio of these channels also agrees
with the experimental IBT curve. Furthermore, the width of the
[41] channel is significantly smaller compared to his width within
the added-row and saw-tooth reconstruction, whereas the [43]
channel is blocked in the unreconstructed p(2 ꢀ 1) superstructure

M. Busch et al. / Surface Science 602 (2008) 2808–2815
2813
Table 1
Widths and atomic string densities of channels in the topmost layer of the p(2 ꢀ 1) superstructure for different structure models
Channel
Long-bridge
Added-row
Width
Short-bridge
Unreconstructed
Width
Saw-tooth
Width
Unreconstructed
Width
Density
Density
Density
Density
[01]
[14]
[13]
[12]
[11]
[21]
[31]
[41]
[10]
2
2
2
2
1
2a
1/2
1
1/2
1
1/2
1
1/2
1
1
1/2
1/2
1/2
1/2
1
1/2
2
1/2
1/2
1/2
1/2
1/2
1
1/2
2
1/2
1/2
1/2
1/2
1/2
1
1/2
2
1/2
1/2
1/2
1/2
1/2
1
1/2
2
1/2
1/2
1/2
1/2
1/2
1
1/2
1
1/2
1/2a
1/2a
1/2a
1/2a
2a
4/3
1/2a
4/3
1/2a
1
1/2a
2a
1/2a
1
1/2a
1
5/4
The given factors indicate the enlargement and reduction of the widths and atomic string densities with respect to the unreconstructed topmost layer of clean Ni(110).
Thereby, it is assumed that the oxygen adatoms contribute to the formation of atomic strings.
a
At these channels, only the density of one of both atomic strings of a channel will be changed, while the density of the other atomic string remains unchanged.
with short-bridge adsorption sites (cf. Fig. 1d). Therefore, the sim-
ulated IBT curve for this structure show only a small dip for the
[41] channel but no dip for the [43] channel.
[01]
[11]
300 L
[10]
In conclusion from these findings we favour short-bridge
adsorption sites for the p(2 ꢀ 1) superstructure compared to the
adsorption of O atoms on long-bridge sites. In particular, we rule
out the added-row and the saw-tooth reconstruction for the
p(2 ꢀ 1) adsorbate layer. In view of other studies which favoured
the added-row reconstruction, we have to conclude that under
our conditions of surface preparation an unreconstructed
p(2 ꢀ 1) superstructure is formed.
120 L
80 L
50 L
3.2. NiO(100) formation by exposure to molecular and atomic oxygen
The formation of ultrathin NiO(100) layers by exposure to
molecular (cf. Fig. 6) and atomic oxygen (cf. Fig. 7) was also inves-
tigated here using IBT. The exposure was gradually increased up to
a dose of 15 L for atomic and 300 L for molecular oxygen. In both
cases the structural changes from the primitive rectangular surface
lattice of Ni(110) to the square lattice of NiO(100) (cf. Fig. 8) is evi-
dent by a pronounced dip at an azimuthal angle of 45°, the direc-
tion of the [11] channel of the primitive square lattice of
NiO(100). For the exposure to molecular oxygen this dip appears
at an oxygen dose above 30 L and represents the initial state for
the formation of a NiO(100) layer [21,24,27,42–44]. For further
exposure to molecular oxygen, the IBT curves shown in Fig. 6 are
only slightly changed and show always the dips for the primitive
square lattice. Diffuse LEED spots indicate that the (1 ꢀ 1) pattern
of the NiO(100) layer is not perfectly ordered.
For exposure to atomic oxygen, the transition from the primi-
tive rectangular lattice of Ni(110) to the square lattice of
NiO(100) proceeds in the IBT curves already at 1 L. This value is
by more than one order of magnitude lower compared to oxidation
using molecular oxygen. This is attributed to the feature that stick-
ing and dissociation are no longer rate-limiting steps in the process
of oxide layer formation. For a further exposure to atomic oxygen,
the curves displayed in Fig. 7 are not changed and show always the
dips of the primitive square lattice of NiO(100).
40 L
30 L
20 L
10 L
1.0
1.0
Ni(110)
1.0
0.8
0.6
0.4
We simulated IBT curves for grazing scattering of 50 keV hydro-
gen atoms from clean Ni(110) and clean NiO(100) surfaces. The
results of the simulations are shown in Fig. 9. For the clean
Ni(110) surface we find also for a projectile energy of 50 keV good
agreement with the data, and for the clean NiO(100) surface the
simulation is in accord with the experiment for a dose of 5 L atomic
oxygen. For NiO(100) formation with molecular oxygen, the smal-
ler dip intensities for the [11] channel compared to the [10] and
[01] channels are in poorer agreement. We thus conclude that
0
15
30
45
60
75
90
Azimuthal Angle [deg]
Fig. 6. Normalized SBD counts versus azimuthal angle of incidence h for grazing
scattering of 50 keV hydrogen atoms (/in = 1.2°) with increasing exposure of the
clean Ni(110) surface to molecular oxygen up to 300 L. Vertical lines indicate low-
index directions [uv] and [uv] in the primitive rectangular (lower part) and square
(upper part) surface lattice of Ni(110) and NiO(100), respectively.

2814
M. Busch et al. / Surface Science 602 (2008) 2808–2815
0
15
30
45
60
75
90
[11]
15 L
10 L
5 L
[01]
[10]
1.0
NiO(100)
2 L
1.0
0.8
0.6
0.4
1.5 L
Ni(110)
1 L
0.5 L
1.0
1.0
0
15
30
45
60
75
90
Azimuthal Angle [deg]
Ni(110)
1.0
0.8
0.6
0.4
Fig. 9. Calculated triangulation curves (solid) obtained by computer simulations
based on classical projectile trajectories for grazing scattering of 50 keV hydrogen
atoms from the clean Ni(110) and NiO(100) surfaces. The experimentally deduced
IBT curves (dashed, for NiO(100) the IBT curve after a dosis of atomic oxygen of 5 L
was taken from Fig. 7) are also shown (vertical lines, cf. Fig. 6).
exposure of Ni(110) to molecular oxygen leads to a NiO(100) film
with a higher roughness than for exposure to atomic oxygen.
4. Conclusions
0
15
30
45
60
75
90
The atomic arrangement of well ordered oxygen p(2 ꢀ 1) and
p(3 ꢀ 1)f adsorbate superstructures and the formation of ultrathin
NiO(100) layers on a Ni(110) single crystal was investigated by
grazing scattering of fast hydrogen atoms. We have used ion beam
triangulation (IBT) based on the detection of the number of emitted
electrons and obtained direct informations on the positions of oxy-
gen adsorbates and the structure of nickel oxide layer. We ana-
lyzed the experimental curves via computer simulations based
on structure models for the oxygen adsorbates.
Azimuthal Angle [deg]
Fig. 7. Normalized SBD counts versus azimuthal angle of incidence h for grazing
scattering of 50 keV hydrogen atoms (/in = 1.2°) with increasing exposure of the
clean Ni(110) surface to atomic oxygen up to 15 L (vertical lines, cf. Fig. 6).
[01]
[13]
[12]
[11]
From the fair agreement between the experimental data and
simulations, we provide support for a reconstruction like added-
row or saw-tooth with long-bridge adsorption sites for the
p(3 ꢀ 1)f superstructure on Ni(110) formed at 460 K [1,12,14,18–
22]. For this superstructure Ni-rows are added on the topmost
layer and the O atoms are adsorbed on the long-bridge sites of
the Ni lattice. The best accordance was obtained assuming a posi-
tion of O adatoms above the topmost layer of Ni atoms of about
0.3 Å.
[21]
[31]
For the oxygen p(2 ꢀ 1) superstructure formed at 460 K we de-
rive from IBT clear signature for the unreconstructed adsorbate
structure with O atoms on the short-bridge sites of the Ni lattice.
This finding is not in accord with a considerable number of studies
by different experimental methods [12–14,18–22]. On the other
hand, Masuda et al. [23] observed also an unreconstructed
p(2 ꢀ 1) superstructure with oxygen at short-bridge adsorption
sites at room temperature. At an elevated temperature of 540 K,
thermal ordering leads to the established reconstruction accompa-
nied by a sharpening of the LEED spots. In our experiments,
[10]
Fig. 8. Structural model of the NiO(100) surface. Gray bars indicate channels along
low-index directions [uv] in the primitive square surface lattice of NiO(100).
Widths of gray bars indicate widths of axial channels. Gray circles: nickel atoms,
black circles: oxygen atoms. Grid lines denote NiO(100) lattice.

M. Busch et al. / Surface Science 602 (2008) 2808–2815
2815
[8] Y. Kuk, F.M. Chua, P.J. Silverman, J.A. Meyer, Phys. Rev. B 41 (1990) 12393;
however, we find clear-cut evidence that the unreconstructed
phase persists at elevated temperatures. A possible reason for this
behaviour could be based on the morphology of our clean Ni(110)
surface which was prepared in a lengthy procedure of cycles of
grazing sputtering with 50 keV Ar+ ions and subsequent annealing
to 850 K for 10 min. With a carefully oriented crystal surface (mis-
cut about 0.1°) smooth terraces formed by topmost surface atoms
with a width exceeding 1000 Å can be achieved. In case atom
transport from step edges plays a dominant role in the added-
row reconstruction, it might be conceivable that for a large terrace
width this mechanism is suppressed and the unreconstructed
phase will persist. In order to support this interpretation, detailed
studies on the p(2 ꢀ 1) superstructure as function of target tem-
perature should be performed.
For the formation of an ultrathin NiO(100) layer by exposure to
atomic instead of molecular oxygen, the dose can be reduced by
more than one order of magnitude and leads to a NiO(100) layer
with a smaller roughness. For the exposure to atomic oxygen the
structural change from the primitive rectangular lattice of
Ni(110) to the square lattice of NiO(100) proceeds at a dose of
1 L only. This is in line with previous studies where we have found
a similar behaviour for the formation of an FeO(111) layer on a
Fe(110) crystal [28].
Y. Kuk, F.M. Chua, P.J. Silverman, J.A. Meyer, Phys. Rev. Lett. 63 (1989) 386.
[9] M. Taniguchi, K. Tanaka, T. Hashizume, T. Sakurai, Surf. Sci. 262 (1992) L123.
[10] F. Jensen, F. Besenbacher, E. Laegsgaard, I. Stensgaard, Phys. Rev. B 42 (1990)
9206.
[11] M. Schuster, C. Varelas, Surf. Sci. 134 (1983) 195.
[12] H. Niehus, G. Comsa, Surf. Sci. 151 (1985) L171.
[13] D.J. O’Connor, Surf. Sci. 173 (1986) 593.
[14] T. Engel, K.H. Rieder, Surf. Sci. 148 (1984) 321.
[15] K. Baberschke, U. Döbler, L. Wenzel, D. Arvanitis, Phys. Rev. B 33 (1986) 5910.
[16] A.M. Baró, G. Binnig, H. Rohrer, Ch. Gerber, E.StollA. Baratoff, E. Salvan, Phys.
Rev. Lett. 52 (1984) 1304.
[17] J.A. Stroscio, M. Persson, W. Ho, Phys. Rev. B 33 (1986) 6758.
[18] R.G. Smeenk, R.M. Tromp, F.W. Saris, Surf. Sci. 107 (1981) 429.
[19] G. Kleinle, J. Wintterlin, G. Ertl, R.J. Behm, F. Jona, W. Moritz, Surf. Sci. 225
(1990) 171.
[20] B. Voigtländer, S. Lehwald, H. Ibach, Surf. Sci. 225 (1990) 162.
[21] L. Eierdal, F. Besenbacher, E. Laegsgaard, I. Stensgaard, Surf. Sci. 312 (1994) 31.
[22] K. Suzuki, A. Yamane, R. Ozawa, Y. Gunji, K. Higashiyama, H. Fukutani, Surf. Sci.
365 (1996) 248.
[23] S. Masuda, M. Nishijima, Y. Sakisaka, M. Onchi, Phys. Rev. B 25 (1982) 863.
[24] C.R. Brundle, J.Q. Broughton, in: D.A. King, D.P. Woodruff (Eds.), The Chemical
Physics of Solid Surfaces and Heterogenous Catalysis, vol. 3a, North Holland,
Amsterdam, 1990.
[25] P.H. Holloway, J. Vac. Sci. Technol. 18 (1981) 653.
[26] P.H. Holloway, J.B. Hudson, Surf. Sci. 43 (1979) 123.
[27] P.R. Norton, P.E. Bindner, T.E. Jackman, Surf. Sci. 175 (1986) 313.
[28] M. Busch, M. Gruyters, H. Winter, Surf. Sci. 600 (2006) 2778.
[29] M. Busch, M. Gruyters, H. Winter, Surf. Sci. 582 (2005) 31.
[30] H. Winter, K. Maass, S. Lederer, HP. Winter, F. Aumayr, Phys. Rev. B 69 (2004)
054110.
[31] F. Aumayr, G. Lakits, HP. Winter, Appl. Surf. Sci. 63 (1991) 17.
[32] R. Pfandzelter, T. Bernhard, H. Winter, Phys. Rev. Lett. 90 (2003) 036102.
[33] G. Lakits, F. Aumayr, HP. Winter, Rev. Sci. Instrum. 60 (1989) 3151.
[34] T. Bernhard, M. Baron, M. Gruyters, H. Winter, Phys. Rev. Lett. 95 (2005)
087601.
[35] M. Baron, T. Bernhard, M. Gruyters, H. Winter, Surf. Sci. 600 (2006) 3924.
[36] D.S. Gemmell, Rev. Mod. Phys. 46 (1974) 129.
[37] W.H. Press, B.P. Flannery, S.A. Teukolsky, W.T. Vetterling, Numerical Recipes,
Cambridge University Press, Cambridge, 2002.
Acknowledgements
We thank K. Maass for his assistance in the preparation of the
experiments, S. Wethekam and J. Seifert for their help in perform-
ing the computer simulations.
References
[38] G. Molière, Z. Naturf. 2A (1947) 133.
[39] D.J. O’Connor, J.P. Biersack, Nucl. Instrum. Meth. Phys. Res. B 15 (1986) 14.
[40] D.P. Jackson, Surf. Sci. 43 (1974) 431.
[1] F. Besenbacher, J.K. Norskov, Prog. Surf. Sci. 44 (1993) 5.
[2] L.H. Germer, A.U. MacRae, J. Appl. Phys. 33 (1962) 2923.
[3] A.U. MacRae, Surf. Sci. 1 (1964) 319.
[4] L.H. Germer, J.W. May, R.J. Szostak, Surf. Sci. 8 (1967) 430.
[5] J.W. May, L.H. Germer, Surf. Sci. 11 (1968) 443.
[6] F. Jensen, F. Besenbacher, E. Laegsgaard, I. Stensgaard, Phys. Rev. B 41 (1990)
10233.
[41] S. Wethekam, H. Winter, Nucl. Instrum. Meth. Phys. Res. B 258 (2007) 48.
[42] C. Benndorf, B. Egert, C. Nöbl, H. Seidel, F. Thieme, Surf. Sci. 92 (1980) 636.
[43] P.H. Holloway, R.A. Outlaw, Surf. Sci. 111 (1981) 300.
[44] T. Komeda, Y. Sakisaka, M. Onchi, H. Kato, S. Masuda, K. Yagi, Surf. Sci. 188
(1987) 45.
[7] D.J. Coulman, J. Wintterlin, R.J. Behm, G. Ertl, Phys. Rev. Lett. 64 (1990) 1761.