Ultrathin alumina film on Cu-9at%Al(1 1 1)

Article abstract of DOI:10.1016/j.susc.2008.02.040

We have investigated the structure of the clean and the oxidized (1 1 1) surface of a Cu-Al alloy with 9 at% Al by scanning tunneling microscopy (STM), Auger electron spectroscopy (AES), low energy ion scattering (LEIS) and low energy electron diffraction (LEED). Annealing of the clean crystal at 680 °C leads to segregation of Al to the surface. The Al concentration at the annealed surface is 23 ± 2% and domains with a (√3 × √3)R30° superstructure are visible, as well as small Cu(1 1 1) areas and disordered patches. Oxidation at 680 °C leads to the formation of a well-ordered flat alumina film with two very similar oxide structures. One oxide structure has a nearly commensurate rectangular cell rotated by 30° with respect to a close-packed row of the substrate and grows in three different domains. The second structure has a commensurate cell consisting of four equivalent building blocks and has a rectangular centered symmetry. This structure is rotated by 18° with respect to a close-packed row of the substrate and grows in six different domains. The rectangular building blocks of these two oxide structures have a similar thickness, the same surface termination and the same number and arrangement of the atoms as the oxide film on NiAl(1 1 0) [G. Kresse, M. Schmid, E. Napetschnig, M. Shishkin, L. Ko?hler, P. Varga, Science 308 (2005) 1448]. In contrast to the oxide on NiAl(1 1 0), alumina on the Cu-Al alloy crystal does not show stress-induced domain boundaries and grows in large defect-free domains. Thus, Pd deposited on this oxide nucleates not only on domain boundaries and steps but also on the unperturbed oxide, forming (1 1 1)-oriented clusters.

Full text of DOI:10.1016/j.susc.2008.02.040

Surface Science 602 (2008) 1750–1756
Contents lists available at ScienceDirect
Surface Science
journal homepage: www.elsevier.com/locate/susc
Ultrathin alumina film on Cu–9at%Al(111)
*
E. Napetschnig, M. Schmid , P. Varga
Institut für Allgemeine Physik, Vienna University of Technology, Wiedner Hauptstr. 8-10, A-1040 Wien, Austria
a r t i c l e i n f o
a b s t r a c t
Article history:
We have investigated the structure of the clean and the oxidized (111) surface of a Cu–Al alloy with 9 at%
Al by scanning tunneling microscopy (STM), Auger electron spectroscopy (AES), low energy ion scattering
Received 27 November 2007
Accepted for publication 18 February 2008
Available online 17 March 2008
(LEIS) and low energy electron diffraction (LEED). Annealing of the clean crystal at 680 °C leads to segre-
gation of Al to the surface. The Al concentration at the annealed surface is 23 ± 2% and domains with a
p
p
(
3 ꢀ 3)R30° superstructure are visible, as well as small Cu(111) areas and disordered patches. Oxida-
Keywords:
tion at 680 °C leads to the formation of a well-ordered flat alumina film with two very similar oxide struc-
tures. One oxide structure has a nearly commensurate rectangular cell rotated by 30° with respect to a
close-packed row of the substrate and grows in three different domains. The second structure has a com-
mensurate cell consisting of four equivalent building blocks and has a rectangular centered symmetry.
This structure is rotated by 18° with respect to a close-packed row of the substrate and grows in six dif-
ferent domains. The rectangular building blocks of these two oxide structures have a similar thickness,
the same surface termination and the same number and arrangement of the atoms as the oxide film
on NiAl(110) [G. Kresse, M. Schmid, E. Napetschnig, M. Shishkin, L. Köhler, P. Varga, Science 308
Scanning tunneling microscopy (STM)
Auger electron spectroscopy (AES)
Low energy ion scattering (LEIS)
Copper
Aluminium
Aluminium oxide
Palladium
Metal–oxide interfaces
(2005) 1448]. In contrast to the oxide on NiAl(110), alumina on the Cu–Al alloy crystal does not show
stress-induced domain boundaries and grows in large defect-free domains. Thus, Pd deposited on this
oxide nucleates not only on domain boundaries and steps but also on the unperturbed oxide, forming
(111)-oriented clusters.
Ó 2008 Published by Elsevier B.V.
1
. Introduction
Thin alumina films grown on various metal substrates are of
electron diffraction (LEED) [11,12]. At low annealing temperatures
the LEED pattern indicates a (1 ꢀ 1) structure which evolves to a
p
p
diffuse ( 3 ꢀ 3)R30° pattern at elevated temperatures. While
the sample is at high temperatures, LEED shows a (1 ꢀ 1) pattern
again. The temperature reported for the second reversible transi-
tion is between 227 °C [13] and 327 °C [11].
interest for different technological applications. They are used in
catalysis, for gas sensors, as coatings and in microelectronics. In re-
search, a very common application is the use of thin alumina films
on a metal substrate as a support for metal particles serving as
model catalysts. Using such ultra thin oxide films eliminates charg-
ing problems when applying measurement methods involving
charged particles. For well-defined and reproducible results, the
alumina film has to be well ordered. Since more than two decades,
many groups have attempted to find suitable substrates and reci-
pes for growing such ultra-thin alumina films [1–7].
The high Al concentration at the surface, reported as 22% Al at
the surface [12], should make it possible to grow a stoichiometric
alumina layer. It was argued that the relatively small misfit of 6–
8% between the close-packed O planes of a-Al O or c-Al O and
2 3 2 3
the Cu–9at%Al(111) lattice should facilitate epitaxial growth [8].
Recent studies [8–11] have reported the formation of 0.4–3.5 nm
p
p
thick c-Al
ture on Cu–9at%Al(111) after oxygen adsorption at temperatures
between 577 and 725 °C, using rather high O doses (1200–
2
O
3
-like alumina films with a (7/ 3 ꢀ 7/ 3)R30° struc-
The Cu–Al(111) alloy with 9 at% of Al is one of the substrates
where growth of a well ordered Al
2
O
3
film was reported [8–10].
2
ꢁ
6
Concerning the clean (not oxidized) alloy, Auger electron spectros-
copy (AES) measurements show an increase of the Al concentration
at the surface when the crystal is annealed in ultrahigh vacuum
4000 L; 1 L = 1.33 ꢀ 10 mbar s). In these studies, AES, XPS
(X-ray photoelectron spectroscopy), SEM (scanning electron
microscopy), LEED and RHEED (reflection high energy electron dif-
fraction) were used. As STM (scanning tunneling microscopy) mea-
surements were missing for the clean surface as well as for the
alumina-covered crystal, in our work we have focused on STM
measurements.
(
UHV) [11,12]. The different Al concentrations at the surface in-
duce different superstructures, which are visible in low-energy
We compare our findings on the alumina film grown on Cu–
9at%Al(111) with the well-known thin alumina film on NiAl(110).
*
Corresponding author. Tel.: +43 1 58801 13452; fax: +43 1 58801 13499.
E-mail address: schmid@iap.tuwien.ac.at (M. Schmid).
0
039-6028/$ - see front matter Ó 2008 Published by Elsevier B.V.
doi:10.1016/j.susc.2008.02.040

E. Napetschnig et al. / Surface Science 602 (2008) 1750–1756
1751
For the alumina film on NiAl(110), the oxide film structure was
determined recently by combining STM images and density func-
tional theory calculations [14]. The alumina film on NiAl(110)
sists of 28 atoms per unit cell. The Al
s
layer is only 0.4 Å lower than
the topmost layer and has 24 atoms per unit cell, which are all in
the center of either a triangle or a square of the O atoms. The Al
layer has a slightly distorted hexagonal structure. The O atoms of
the third alumina layer, named O (‘‘i” for ‘‘interface”), replicate
the Al lattice; each O atom is located below an Al atom [14].
Fig. 1b shows the atomic structure of this layer together with the
atomic positions of the interfacial Al layer. The 16 Al atoms are ar-
ranged in pentagon-heptagon pairs and can be probed by STM at
larger tip sample distances [14]. The Al atoms are located prefer-
s
s
2 3
has a structure and a stoichiometry that differs from all Al O bulk
phases. It has a nearly rectangular unit cell (a = 18.0 Å, b = 10.6 Å;
bold lines in Fig. 1) in two reflection domains. The unit cell has
twofold rotational symmetry with glide planes parallel to the
two sides of the unit cell, i.e., p2gg plane-group symmetry. The film
consists of two aluminium and two oxygen layers and the surface
is oxygen terminated. Fig. 1a shows the two topmost layers, named
i
s
i
s
i
i
i
O
O
s
and Al
s
(‘‘s” for surface). Both layers have been imaged by STM,
at low temperatures [14]. The O
entially above Ni rows of the substrate and avoid its Al rows. This
row matching leads to a compressive stress and a slight distortion
of the rectangular oxide cell.
s
at room temperature and Al
s
s
layer shows triangular and square atomic arrangements and con-
Fig. 1. The atomic arrangement of the alumina film on NiAl(110). (a) The surface – a square and triangular arrangement of oxygen atoms is marked, and the thin lines mark
rows of Al atoms leading to a strong (6,2) LEED spot. (b) The third (oxygen) and fourth (aluminium) layer. Ni rows of the underlying NiAl(110) surface are indicated by
vertical lines.
s

1
752
E. Napetschnig et al. / Surface Science 602 (2008) 1750–1756
The alumina film on NiAl(110) exhibits two types of domain
‘‘A” in Fig. 2b) as well as areas of pure Cu(111) (marked ‘‘B”),
p
p
boundaries [15–17], reflection domain boundaries between differ-
ent orientations of the oxide and line defects separating areas of
the same orientation of the oxide. The latter are often referred to
as anti-phase domain boundaries and caused by the stress due to
row matching with the substrate [17]. The domain boundaries ap-
pear bright in STM images at higher positive bias voltages because
of unoccupied states localized at the domain boundaries [17,18].
Both, antiphase and reflection domain boundaries can act as nucle-
ation centers for metal clusters on the alumina film on NiAl(110)
mainly at the domain boundaries of the ( 3 ꢀ 3)R30° structure.
Additionally, there are also a few small areas (marked ‘‘C”) where
p
p
the Al atoms are locally denser than in the ( 3 ꢀ 3)R30°
structure.
We have drawn the (1 ꢀ 1) lattice over the atomically resolved
STM images with chemical contrast (Fig 2c) and measured the
heights on the (1 ꢀ 1) positions, resulting in a bimodal distribution
(Fig. 2d). The histogram can be fitted by a superposition of two
Gaussian distributions. The ratio of the concentrations of Cu and
Al in the surface is equal to ratio of the areas under the two Gaus-
sians. The quantification results in a surface concentration of
[
7]. For example, Pd clusters nucleate at defects on the alumina
surface, i.e., at domain boundaries and step edges [19,20].
2
3 ± 2% Al, in excellent agreement with the previous AES result
of 22% obtained after annealing at a somewhat lower temperature
12].
2
. Experimental
[
+
The Cu–9at%Al(111) sample was prepared by 2 keV Ar sputter-
ing and annealing at 680 °C for 10 min in a UHV chamber with a
3
.2. Oxidized Cu–9%Al(111) crystal
ꢁ10
base pressure in the low 10
mbar range. The temperature was
After sufficient cleaning, oxidation of the crystal at 680 °C led to
measured by a thermocouple, which was mounted on the non-
transferable part of the sample holder and thus showed a different
temperature (600 °C). The true sample temperature was deter-
mined with a disappearing filament pyrometer, calibrated with
the thermocouple to eliminate the influence of light attenuation
by the viewport. We have checked the cleanness of the sample
after sputtering and annealing with AES. After several cycles of
sputtering and annealing, no impurities like sulphur or carbon
were detected. Unless noted otherwise, the alumina film was pre-
pared by dosing of 100 L oxygen with an oxygen partial pressure of
a surface covered by a well-ordered flat oxide film with large do-
mains and thus very good long range ordering (Fig. 3). To estimate
the thickness and surface composition of this oxide film we com-
pared it with the alumina film on NiAl(110), which has a well-
known thickness, stoichiometry and surface composition [14,15].
For the estimation of the thickness we calculated the expected
AES spectrum with the AES database SESSA [22] for the two crys-
tals covered by a thin alumina film. The results were corrected
by assuming that the sensitivity of the CMA is proportional to
the energy. Considering that we do not know the exact energy-
dependent analyzer sensitivity, that we use Auger peak-to-peak
heights instead of intensities and also considering possible system-
atic errors of the electron transport calculation, we should not di-
rectly compare the experimental and calculated results.
Nevertheless, assuming that the oxide films on both substrates
are equal, and further assuming equal influence of instrumental
factors on the Ni (848 eV) and Cu (920 eV) substrate lines, the
experimental intensity ratio between the oxygen line (510 eV)
and the substrate lines should be proportional to the calculated
one. Table 1 shows that the proportionality factor between mea-
sured and calculated oxygen/substrate ratios is only 6% lower for
the oxide on Cu–9at%Al(111) than for the oxide on NiAl(110). This
deviation is within the experimental uncertainty as well as within
the uncertainty of the calculation (both are estimated around 10%).
These results allow us to rule out a significantly different thickness
of the oxide, especially a different number of layers: removing the
ꢁ
7
1
.3 ꢀ 10 mbar at 680 °C. STM, AES, LEED and low-energy ion
scattering (LEIS) measurements were performed at room tempera-
ture in the analysis chamber, which has a base pressure below
5
ꢁ11
ꢀ 10
mbar. The STM measurements were performed using a
customized commercial STM (Omicron l-STM) with an electro-
chemically etched W tip. All STM images were obtained in constant
+
current mode. LEIS measurements were carried out with 1 keV He
ions at a scattering angle of 90° and a current density of about
2
3
nA/mm . The LEIS spectra were obtained with a hemispherical
energy analyzer operated with fixed retardation ratio. AES spectra
were taken with a cylindrical mirror analyzer (CMA) with a con-
centric electron gun and 3 keV electron energy; the Auger peak-
to-peak heights (APPH) of the differentiated spectra were used
for quantitative analysis. Pd was deposited from a rod using a
water-cooled electron-beam evaporator (Focus EFM3). During
deposition a retarding voltage was applied to the orifice of the
evaporator to suppress high-energy metal ions from the evapora-
tor, as such ions otherwise create nucleation centers on the surface
i i
interfacial layers Al and O would result in a calculated O/Cu inten-
sity ratio of 0.29; duplicating these layers gives a ratio of 0.92.
To determine the composition of the uppermost layer of the
oxide film on Cu–9%Al(111), we compared the ratios of the areas
of the oxygen to the Al peak in the LEIS spectra of the oxide films
on the two substrates. For the oxide on Cu–9%Al(111), the O/Al
peak intensity ratio was 0.72, which is 6% below that of the oxide
on NiAl(110). Considering the experimental error, this points to a
similar or possibly the same surface composition of the oxides on
Cu–9%Al(111) and NiAl(110).
[
21]. The deposition rate, calibrated with a quartz crystal microbal-
ance, was 0.07 ML/min. We define 1 ML Pd here as the amount of
atoms in one Pd(111) layer.
3
. Results
3
.1. Pure Cu–9%Al(111) crystal
We have investigated the clean Cu–9%Al(111) surface after
The LEED pattern of the oxidized surface of the Cu–9%Al(111)
crystal is shown in Fig. 4. It is a complex pattern with a multitude
of reflections, different from the diffraction pattern reported in
Refs. [8,10,11,13]. The LEED pattern shown is slightly distorted
due to residual magnetic fields and a non-perfect geometry. Never-
theless, we can use the pattern to derive the unit cells, a procedure
aided by the STM images (see below). The first structure (Fig. 4a) is
a rectangular one with three rotational domains, nearly commen-
surate with the substrate. The unit cell dimensions, derived from
the LEED pattern, are 18.2 Å ꢀ 10.6 Å. The long side of the unit cell
is rotated by 30° with respect to a close-packed row of the sub-
annealing at 680 °C, the same temperature as used for oxidation.
p
p
The LEED image shows a diffuse ( 3 ꢀ 3)R30° pattern (Fig. 2a).
Since cooling in our system is not fast enough to freeze the high-
temperature (1 ꢀ 1) phase, this observation is in agreement with
Refs. [11,13]. We have obtained atomically resolved STM images
with chemical contrast (Fig. 2b and c). Since the atomic radius of
Al is larger than that of Cu, and Al is the more reactive metal, we
attribute the bright species to Al, in agreement with the surface
composition known from the literature [12] (see below). The
p
p
images show areas with a ( 3 ꢀ 3)R30° superstructure (marked

E. Napetschnig et al. / Surface Science 602 (2008) 1750–1756
1753
a
b
DB
C
B
A
c
d
5
4
3
2
1
0
0
0
0
0
0
0
5
10 15 20 25 30 35 40
Height [pm]
Fig. 2. The clean Cu–9%Al(111) crystal after annealing at 680 °C. (a) LEED pattern taken at 90 eV with the reciprocal cell of the (p
3 ꢀ
p
3)R30° superstructure marked. (b)
p
p
Atomically resolved STM image (100 Å ꢀ 80 Å; 2 mV/10.5 nA) with chemical contrast (Al atoms appear bright and Cu atoms dark) showing areas with an Al ( 3 ꢀ 3)R30°
p
p
superstructure (A), pure Cu(111) areas (B) and Al-rich defects (C). A domain boundary (DB) of the Al ( 3 ꢀ 3)R30° superstructure is marked. (c) Atomically resolved STM
image (100 Å ꢀ 100 Å; 1 mV/39.2 nA) with the (1 ꢀ 1) lattice marked by white points. (d) Histogram of the heights at the lattice positions (grey). The black line is a fit by a
sum of two Gaussian distributions.
strate. Also the second oxide structure has a rectangular cell
(Fig. 4b), with a size of 18.3 Å ꢀ 10.6 Å, rotated by ±18° with re-
spect to a close-packed row of the substrate and therefore forming
six rotational domains. Two by two of these rectangular building
blocks form a larger commensurate cell with centered symmetry.
Both superstructures show weak or no intensity of the odd LEED
spots on the axes (squares in Fig. 4; the intensity visible in some
squares can be attributed to different spots of another domain).
This indicates p2gg symmetry of the rectangular cells, which is
the plane group of the surface oxide on NiAl(110). Both super-
structures on Cu–9%Al(111) show strong (6,2) or (6,ꢁ2) spots.
These are also strong spots for the LEED pattern of the surface
oxide on NiAl(110) (see, e.g., Fig. 10 of Ref. [7]). Considering that
Al is a stronger scatterer for electrons than oxygen, these spots
s
are mainly caused by the Al rows marked in Fig. 1a. The other
superstructure spots, even if they are strong at one electron energy,
get weaker at most other energies (e.g., the (3,1) or (3,ꢁ1) spots,
which are quite strong in Fig. 4).
Atomically resolved STM images taken at very low tunneling
voltages in the mV range show that the unit cell of the first oxide
structure (Fig. 5a) and one building block of the second structure
(Fig. 5b) have the same amount of atoms and an identical arrange-
ment of the atoms. The 28 visible atoms are arranged in squares
and triangles with p2gg symmetry. Not only the number of atoms
and the symmetry but also the atomic structure are the same as for
Fig. 3. STM image of the oxidized surface (3000 Å ꢀ 3000 Å; 2.8 V/0.4 nA). Domain
boundaries in the oxide are visible as bright lines.
s
the surface oxygen (O ) layer of the ultrathin alumina film on

1
754
E. Napetschnig et al. / Surface Science 602 (2008) 1750–1756
Table 1
Measured and calculated [22] Auger intensity rations for the alumina films on Cu–9%Al(111) and NiAl(110)
Substrate
Auger lines (eV)
Experimental APPH ratio
Calculated intensity ratio
Ratio experimental/calculated
Cu–9%Al(111)
NiAl(110)
O(510)/Cu(920)
O(510)/Ni(848)
0.75
1.29
0.60
0.97
1.25
1.33
Calculated intensities are corrected for the transmission of the analyzer, which is assumed to be proportional to the energy. For both substrates, calculated ratios assume the
film composition, density and thickness determined for NiAl(110).
NiAl(110) [14] imaged by STM at similar tunneling conditions
Fig. 5c).
Fig. 5d–f shows STM images taken at larger tip sample dis-
tances, where we know from the studies of the oxide on NiAl(110)
that the Al layer is imaged [14,17]. Although we have not obtained
9%Al(111) show 16 atoms, arranged in heptagon and pentagon
pairs exactly as for the Al of the alumina film on NiAl(110).
(
i
Coming back to the large-scale images of the oxide film on the
Cu–Al alloy crystal taken at positive bias voltage (e.g., Fig. 3), we
can now determine that the bright lines are rotational domain
boundaries in the wider sense, i.e., boundaries between two differ-
ent rotations of essentially the same kind of structure. Having
understood the structure, we have to expect three domains of
the non-commensurate structure shown in Figs. 4a and 5a (three
rotations of the rectangle; no mirror domains because of the
p2gg symmetry) and six domains of the commensurate structure
shown in Figs. 4b and 5b. Domain boundaries can occur between
any of these nine domains. In our STM images, we did not observe
any lines caused by antiphase (translational) domain boundaries
like those forming the majority of the line defects in the oxide film
on NiAl(110) [15–18].
i
images of comparable quality as for NiAl(110) (Fig. 5f) of this layer,
the STM images (Fig. 5d and e) indicate that both structures on Cu–
We finally note that the ultrathin oxide described here is not
the same oxide structure as observed by Yoshitake et al. [8–
1
1,13]. Using a slightly lower pressure and temperature
ꢁ8
(
6.5 ꢀ 10 mbar, approx. 580 °C), but a significantly higher oxy-
gen dose (1000 L) than for preparation of the ultrathin films de-
scribed above, we have obtained a LEED pattern dominated by
p
p
the (7/ 3 ꢀ 7/ 3)R30° structure reported in these papers, but
with an additional weaker contribution of the 30°-rotated struc-
ture described above. AES showed an O/Cu ratio twice as high as
for the ultrathin oxide described above, indicating a thicker oxide.
3.3. Palladium clusters on the alumina film
Fig. 6a shows an STM image after deposition of 1 ML Pd at room
temperature. The cluster density of the Pd clusters on the alumina
1
6
2
film on the Cu–Al alloy is 0.18 ꢀ 10 clusters/m , about 1/6th of
the density on the alumina film on NiAl(110). The clusters on alu-
mina/Cu–9%Al(111) are larger, however, with typical side lengths
between 10 and 20 nm, in contrast to 5–10 nm on the oxidized
NiAl(110) crystal. On both oxides, the Pd clusters have a truncated
triangular shape with flat facets on top, indicating (111) orienta-
tion [19,20]. In contrast to the film on NiAl(110), there are very
few high Pd clusters with a steep slope and a round top. Fig. 6b
shows the height distribution of the Pd clusters. It has distinct
maxima that are separated by the Pd(111) interlayer distance,
confirming the (111) orientation of the clusters. The heights given
in this histogram are the values measured from the STM image,
uncorrected for the voltage dependent apparent thickness of the
oxide film (discussed in Ref. [20]). We estimate that the true clus-
ter height is about 2 Å lower, thus the maximum of the true height
distribution is at 18 Å, only one or two layers higher than on the
oxide on NiAl(110).
The nucleation sites are different from the ones on the oxide of
NiAl(110). Apart from Pd clusters nucleating on the domain
boundaries and the step edges there are also Pd clusters at posi-
tions where no domain boundaries are visible.
Fig. 4. LEED pattern of the oxidized surface taken at 90 eV. The reciprocal lattices of
the rectangular oxide structures marked are rotated by (a) 30° and (b) 18° with
respect to the close-packed directions of the substrate. In the right side symmetry-
equivalent rotational domains are also indicated. The integer-order maxima of the
substrate are circled in (a), and small squares show positions with vanishing inte-
nsity in a structure with p2gg symmetry and one domain.
4
. Discussion
Our STM results show that the alumina film on the Cu–
9%Al(111) crystal has exactly the same surface structure as the

E. Napetschnig et al. / Surface Science 602 (2008) 1750–1756
1755
Fig. 5. STM images of (a–c) the surface oxygen layer with atomic resolution and (d–f) the interface Al layer visible at somewhat higher tunneling resistance. The left images
a,d) show the first (incommensurate) oxide structure and the center images (b,e) show the second (commensurate) oxide on Cu–9%Al(111). The oxide on NiAl(110) is
(
shown for comparison (right). In all images, the dots mark the atomic positions of the respective layers calculated for alumina on NiAl(110) [14]. In (b,e) the primitive unit
cell is drawn in black and the four rectangular building blocks are marked in white.
2
.25Å
3
2
1
0
0
0
0
1
0
15 20 25 30 35 40 45
Height (Å)
2
Fig. 6. (a) STM image (2500 ꢀ 2500 Å , 2.8 V/0.2 nA) (b) and cluster height distribution after deposition of 1 ML Pd on the alumina film on Cu–9%Al(111) at room
temperature.
film on NiAl(110) [14]. As this structure is unique among all alu-
mina structures found so far, we conclude that also the thickness
and composition of the oxide is the same as on NiAl(110), and
our STM images also indicate the same structure at the interface.
Also our AES and LEIS values indicate essentially the same thick-
ness and composition of the ultrathin oxides on the two substrates
and rule out any significant differences such as a different number
of layers. An even more sensitive indicator of the oxide composi-
tion is the appearance in STM images at positive voltages near
the conduction band edge, where deviations from the charge–neu-
tral stoichiometry lead to band bending, which is easily detected
by STM [17]. With the exception of the domain boundaries, we
did not find any indications for such deviations from charge neu-
trality. As the oxides on Cu–9%Al(111) and NiAl(110) have the
same structure, we must conclude that this structure, with its tight
intertwining of oxygen triangles and squares at the surface is
highly favorable, also in absence of the interfacial constraint im-
posed by the row matching on the NiAl(110) surface. This notion
is supported by our experience that the preparation of the ultrathin
alumina film on Ni
3
Al(111), which is structurally similar, but with
the O triangles and squares in more separate areas [23], has a
s
much narrower window of preparation parameters than the oxides
on NiAl(110) and Cu–9%Al(111), indicating a lower stability of the
oxide on Ni Al(111).
3

1
756
E. Napetschnig et al. / Surface Science 602 (2008) 1750–1756
The main differences between the oxide films on Cu–9%Al(111)
mains. The only defects visible by STM are the domain boundaries
between the nine different orientations of the alumina structure on
the substrate. We can conclude that the oxide structure found on
NiAl(110) seems to be the most stable structure for a thin alumina
film on different substrates and that Cu–9%Al(111) is an ideal sub-
strate to grow an alumina film with a very good long-range order-
ing. A different density of domain boundaries on Cu–9%Al(111)
and NiAl(110) makes it possible to obtain different metal cluster
densities on the same oxide, which is interesting for studies of
model catalysts.
and NiAl(110) are the slightly different unit cell dimensions. Com-
pared to the slightly sheared rectangle on NiAl(110), the lattice of
the oxide structures on the Cu–Al alloy crystal is rectangular and
expanded by 1–2% in the direction of the longer side of the unit
cell. This can be explained by the row matching between the oxide
and the NiAl(110) substrate, which exerts compressive stress on
the oxide along one diagonal of the rectangular cell [17]. On Cu–
9
%Al(111), one of the unit cells is incommensurate with the sub-
strate in all directions, thus it can freely assume its preferred lattice
dimensions. This tells us that the potential energy surface of the
Cu–9%Al(111) substrate is less corrugated than on NiAl(110). This
fact may be explained on the one hand by the much weaker ten-
dency towards chemical ordering in Cu–Al alloys compared to
NiAl, making it possible to avoid the unfavorable configuration
Acknowledgements
This work was supported by the Austrian Fonds zur Förderung
der wissenschaftlichen Forschung within the joint Research Project
S9003, ‘‘Nanoscience on Surfaces”. We thank Dr. Michiko Yoshitake
from National Institute for Materials Science (NIMS), Japan, for pro-
viding the two Cu–9%Al(111) crystals.
[
i
14] of having a substrate Al atom below an Al atom of the sub-
strate. On the other hand the energy gain of a heteroatomic Al–
Cu bond between oxide and substrate is much lower than for an
Al–Ni bond, as indicated by enthalpies of alloy formation [24].
The absence of row matching and, thus, compressive stress in the
oxide on Cu–9%Al(111) is also the reason why there are no anti-
phase domain boundaries in this film, leading to larger areas free
of domain boundaries.
The lower density of domain boundaries leads to a lower den-
sity of Pd clusters nucleating on Cu–9%Al(111). On alumina/
NiAl(110), Pd clusters nucleate exclusively on domain boundaries,
while the distance between the domain boundaries on Cu–
References
[
[
[
1] C.T. Campbell, Surf. Sci. Rep. 27 (1997) 1.
2] C.R. Henry, Surf. Sci. Rep. 31 (1998) 231.
3] R.M. Jaeger, H. Kuhlenbeck, H.-J. Freund, M. Wuttig, W. Hoffmann, R. Franchy,
H. Ibach, Surf. Sci. 259 (1991) 235.
[
[
4] C. Becker, A. Rosenhahn, A. Wiltner, K. von Bergmann, J. Schneider, P. Pervan,
M. Milun, M. Kralj, K. Wandelt, New J. Phys. 4 (2002) 75.
5] S. Degen, C. Becker, K. Wandelt, Faraday Discuss. 125 (2004) 343.
[6] R. Franchy, Surf. Sci. Rep. 38 (2000) 195.
[
[
[
7] M. Bäumer, H.-J. Freund, Prog. Surf. Sci. 61 (1999) 127.
8] S. Nemsak, M. Yoshitake, K. Masek, Surf. Sci. 600 (2006) 4357.
9] Y. Yamauchi, M. Yoshitake, W. Song, Jpn. J. Appl. Phys. 42 (2003) 4721.
9
%Al(111) is so large that Pd clusters nucleate also between them.
We cannot determine whether Pd nucleation between the domain
boundaries is homogeneous nucleation or induced by point de-
fects, but STM images of the pure oxide show no indication of such
defects. The orientation and overall shape of the Pd clusters is the
same on both oxides, but the clusters on alumina/Cu–9at%Al(111)
are about two times as large in their in-plane dimensions, though
only 1–2 ML (ꢂ10–20%) higher . This provides another confirma-
tion of our previous conclusion [20] that the height/width ratio is
not determined by energy minimization but rather by kinetic lim-
itations of interlayer mass transport.
[10] M. Yoshitake, S. Bera, Y. Yamauchi, Surf. Interface Anal. 35 (2003) 824.
[11] M. Yoshitake, S. Bera, Y. Yamauchi, W. Song, J. Vac. Sci. Technol. A 21 (2003)
1290.
[
[
12] R.J. Baird, T.J. Potter, J. Vac. Sci. Technol. A 3 (1985) 1371.
13] W. Song, M. Yoshitake, S. Bera, Y. Yamauchi, Jpn. J. App. Phys. 42 (2003) 4716.
[14] G. Kresse, M. Schmid, E. Napetschnig, M. Shishkin, L. Köhler, P. Varga, Science
08 (2005) 1448.
3
[
15] J. Libuda, F. Winkelmann, M. Bäumer, H.-J. Freund, Th. Bertrams, H.
Neddermeyer, K. Müller, Surf. Sci. 318 (1994) 61.
[16] M. Kulawik, N. Nilius, H.-P. Rust, H.-J. Freund, Phys. Rev. Lett. 91 (2003)
56101.
2
[
17] M. Schmid, M. Shishkin, G. Kresse, E. Napetschnig, P. Varga, M. Kulawik, N.
Nilius, H.-P. Rust, H.-J. Freund, Phys. Rev. Lett. 97 (2006) 046101.
[
[
18] N. Nilius, M. Kulawik, H.-P. Rust, H.-J. Freund, Phys. Rev. B 69 (2004) 121401.
19] K.H. Hansen, T. Worren, S. Stempel, E. Lꢀgsgaard, M. Bäumer, H.-J. Freund, F.
Besenbacher, I. Stensgaard, Phys. Rev. Lett. 83 (1999) 4120.
5
. Conclusions
When a clean Cu–9at%Al(111) crystal is exposed to oxygen at
80 °C, a well ordered flat thin alumina film grows, which has
[20] E. Napetschnig, M. Schmid, P. Varga, Surf. Sci. 601 (2007) 3233.
[21] C. Nagl, O. Haller, E. Platzgummer, M. Schmid, P. Varga, Surf. Sci. 321 (1994)
6
237.
the same structure and properties as the alumina film on
NiAl(110), except for the absence of compressive stress and the
antiphase domain boundaries caused by the stress. Thus, the oxide
film on Cu–9%Al(111) grows in large, probably defect-free do-
[
[
22] W. Smekal, W.S.M. Werner, C.J. Powell, Surf. Interface Anal. 37 (2005) 1059.
23] M. Schmid, G. Kresse, A. Buchsbaum, E. Napetschnig, S. Gritschneder, M.
Reichling, P. Varga, Phys. Rev. Lett. 99 (2007) 196104.
[
24] B. Predel, Landolt-Börnstein, New Series IV/5a, Springer, 1991.