Superstructures on thin Al2O3 grown on Ni3Al(0 0 1)

Article abstract of DOI:10.1016/j.susc.2006.02.079

Thin ordered layers of Al₂O₃ grown on a metallic Ni₃Al(0 0 1) substrate were studied using UHV STM and LEED. The oxidation of the substrate at 800 K and 1100 K forms different structural phases of Al₂O₃. Oxidation of Ni₃Al(0 0 1) at 800 K leads to a simple surface pattern, while oxidation at 1100 K results in a variety of surface patterns revealing superimposed quasi-periodic superstructures larger than expected from a simple moiré model. Atomic resolution achieved with STM on the Al₂O₃ surface allows us to determine all possible mutual orientations of the crystalline lattices. The spectroscopic STM dI/dV plots indicated different electronic properties at various points on the oxide superstructure. We conjecture that they can be attributed to ordering of structural patches distributed in a regular way.

Full text of DOI:10.1016/j.susc.2006.02.079

Surface Science 600 (2006) 4375–4379
www.elsevier.com/locate/susc
Superstructures on thin Al O grown on Ni Al(001)
2
3
3
*
O. Kurnosikov , C.F.J. Flipse, H.J.M. Swagten, B. Koopmans, W.J.M. de Jonge
Department of Applied Physics, Center for NanoMaterials and COBRA Research Institute, Eindhoven University of Technology,
P.O. Box 513, 5600MB Eindhoven, The Netherlands
Available online 22 May 2006
Abstract
Thin ordered layers of Al O grown on a metallic Ni Al(001) substrate were studied using UHV STM and LEED. The oxidation of
2
3
3
2 3 3
the substrate at 800 K and 1100 K forms different structural phases of Al O . Oxidation of Ni Al(001) at 800 K leads to a simple surface
pattern, while oxidation at 1100 K results in a variety of surface patterns revealing superimposed quasi-periodic superstructures larger
than expected from a simple moir e´ model. Atomic resolution achieved with STM on the Al O surface allows us to determine all possible
2
3
mutual orientations of the crystalline lattices. The spectroscopic STM dI/dV plots indicated different electronic properties at various
points on the oxide superstructure. We conjecture that they can be attributed to ordering of structural patches distributed in a regular
way.
ꢀ
2006 Elsevier B.V. All rights reserved.
Keywords: Scanning tunneling microscopy; Low energy electron diffraction; Aluminum oxide; Thin film structure; Nickel compounds; Lowindex single
crystal surfaces; Surface structure, morphology, roughness, and topography; Defects
1
. Introduction
LEED patterns [8]. Because of several symmetrically equiv-
alent mutual orientations of crystalline lattices of the sub-
strate and the aluminium oxide layer, the superstructure
forms several types of domains with the dimension of sev-
eral tens of nanometers [9].
Very thin layers of dielectrics are widely used not only
for electrical insulation, but also as a tunnel barrier, for
example in magnetic tunnel junctions [1]. Ordered epitaxial
insulating layers therefore can be a good choice due to their
homogeneous thickness and well-defined interface atomic
structure.
Although the LEED analysis allows one to determine
the unit cell of aluminium oxide accommodated to the
substrate, the structural details of the Al O layers remain
2
3
It has already been shown that the controllable oxida-
tion of single crystalline surfaces of some binary alloys con-
undetermined. In some cases the aluminium oxide is asso-
ciated with several known bulk phases of Al O as c-alu-
2
3
taining aluminium such as NiAl, Ni Al, and CoAl provides
mina [10], h-alumina [11] or j-alumina [12]. This implies,
in all cases, hexagonal packing of oxygen at the surface,
while different distribution of aluminium atoms within
the tetrahedral or octahedral coordination in between
the oxygen atoms determines a variety of phases [13].
The thickness of the aluminium oxide is reported to be
about 0.5 nm that corresponds to only two layers of oxy-
gen [12,14]. The analysis of the atomic structures even with
atomically resolved STM images are quite complicated
and some additional modelling is usually necessary [15].
However, recently it has been evidenced that the crystal-
line lattice of Al O at the surface can be significantly
3
almost perfect atomically flat layers of aluminium oxide of
sub-nanometer thickness [2–4]. Ordered Al O layers can
2
3
be obtained on different orientations of the substrate:
111), (001), and (110) [5–7]. In all the cases the oxidised
(
surface reveals periodical or quasi-periodical surface super-
structures observed directly in STM measurements, and
results in a complicated distribution of multiple spots in
*
Corresponding author. Tel.: +31 247 4307; fax: +31 247 5724.
E-mail address: o.kurnosikov@tue.nl (O. Kurnosikov).
2
3
0
039-6028/$ - see front matter ꢀ 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.susc.2006.02.079

4376
O. Kurnosikov et al. / Surface Science 600 (2006) 4375–4379
distorted, and that the simple hexagonal unit cell of the
oxygen lattice belonging to the high symmetry class is
not applicable [12]. Another recent ab initio calculation
of the structure of thin aluminium oxide [16] shows that
additionally to the distortion of the crystalline lattice,
the stoichiometry of the film is not Al O but Al O . It
2. Experimental
The experiment was carried out in an ultrahigh vacuum
(UHV) chamber at a base pressure of 10
with sample preparation facilities, LEED analysis and an
‘Omicron’ low-temperature STM. A clean substrate of
À11
mbar equipped
2
3
11 13
means that the surface contains many structural packing
defects. However, the distribution of the structural defects
and the distortions across the surface was not analysed in
detail.
Ni Al(001) was exposed to 100–1000 L of oxygen at differ-
3
ent temperatures of the substrate within the range of 800–
1100 K. The STM measurements were taken at 77 K or
4 K. We used tungsten STM tips, which were additionally
processed in UHV by electron bombardment. We used a
very low tunnel current value of 20–200 pA. At high bias
voltage of about 3 V, we suppose tunneling into the states
within the oxide layer occurs, while low bias voltage in the
range 10–100 mV provides the possibility of tunnelling
through the oxide into the interface between the oxide layer
and the metallic substrate.
In this article we argue that the model of the rigid lattice
of aluminium oxide grown on Ni Al(001) is not completely
3
correct, the surface superstructure is more complicated,
and is showing several phases than expected. We will prove
that, depending on the regime of oxidation, the surface
atomic structure of the Al O layer suffers 1-D or 2-D
2
3
reconstructions at low and high temperatures of oxidation,
respectively. Atomic resolution, achieved with a scanning
tunnelling microscope (STM) on the Al O surface, allows
3. Results and discussion
2
3
us to interpret low-energy electron diffraction (LEED)
observations and to determine all possible mutual orienta-
tions of the crystalline lattices.
Oxidation at 800 K and 1100 K forms two different
phases of Al O revealing completely different LEED pat-
2
3
2
Fig. 1. (a) STM image of a 40 · 40 nm area of the low-temperature phase of Al
2
O
3
scanned at V
t
= À0.8 V, and I
t
= 660 pA. The surface is covered with
2
domains with a stripe structure in few directions. To show the domains, a dotted line along the domain borders is added to the image; (b) 3.7 · 3.7 nm
STM current image obtained at V
across the surface. The stripes form an angle of 60ꢁ with respect to the [110] direction of the Ni
= 0.22 V, I = 1.37 nA, and low feedback gain, revealing some details of the stripe-like structure; (d) STS measurements (1) on the chain in between the
stripe, and (2) in the middle of the stripe structure.
t t
= 64 mV, I = 1.88 nA, and low feedback gain, resolving both the atomic structure of the substrate and the stripes
2
3
Al substrate; (c) STM image of 5 · 5 nm obtained at
V
t
t

O. Kurnosikov et al. / Surface Science 600 (2006) 4375–4379
4377
terns. None of the spots of the LEED pattern of the low
temperature phase obtained at 800 K can be attributed to
the spots in the LEED pattern of the high temperature
phase obtained at 1100 K, which means that a complete
phase transition has occured.
Because it is in principle impossible to deduce the real
structure of the aluminium oxide layer only from LEED
analysis, STM measurements of the surface of both phases
has been done.
that can result in only two directions of the moir e´ stripes,
while we have observed domains with four orientations of
the stripes. Additional details of the STM images with high
resolution elucidate this situation.
Firstly, we can determine the mutual orientation of the
stripes and crystalline lattice of the substrate. The STM
images sometimes reveal the crystalline lattice of the Ni Al
3
substrate on the background of the 1.25 nm-nanostripes
(Fig. 1(b)). For this particular image, an angle of 60ꢁ between
the stripes of the superstructure and the [110] direction of the
substrate is clearly visible. Because of the symmetry of the
lattices, we determined four possible angles of ± 30ꢁ and
± 60ꢁ in the stripe directions with respect to [110].
3
.1. Low-temperature phase
On a large scale, the STM images of the low-tempera-
ture phase of Al O reveal regular stripe superstructures
Secondly, the atomic order in the film keeps the orienta-
tion of the stripes. Fig. 1(c) shows that the atoms or the
atomic sites with enhanced electron state density run along
the edges of the stripes. The distance d between the neigh-
bouring bright spots within one chain in Fig. 1(c) is
d = 0.285 nm. This coincides well with the expected dis-
tance of 0.291 nm between oxygen atoms in Al O , and lies
2
3
running in four different directions. The stripes with a peri-
odicity of about 1.25 nm form ꢀ30 nm large domains
(
Fig. 1(a)). Regarding the interpretation of LEED pro-
posed by Bardi et al. [14], one should anticipate moire´
stripes across the surface. However, detailed analysis, pre-
sented below, shows that the stripes observed in our exper-
iment have no relation with the moir e´ pattern. Bardi’s
treatment [14] implies only two structural equivalent mu-
tual orientations of Ni Al and Al O crystalline lattices
2
3
within the tolerance of STM measurements. It means that
the observed stripes relate directly to the atomic structure
of the aluminium oxide.
3
2
3
Fig. 2. Different domains of (a) phase-I at V
t
t t t t t
= 3.48 V, I = 98 pA, (b) phase-II at V = 3.48 V, I = 95 pA, (c) phase-III at V = 3.55 V, I = 100 pA and
2
(d) phase-IV at V
t t
= 3.69 V, I = 196 pA. For the all scans (a–d) the tunnelling was into empty electronic states of the sample. The scan area is 50 · 50 nm .
The orientation of the crystalline lattice of the substrate is the same for all four images, and it is indicated only in (d).

4378
O. Kurnosikov et al. / Surface Science 600 (2006) 4375–4379
Thirdly, the width of the stripes is not always the same.
middle of the stripe reveals a gap, the conductivity on the
stripe border shows additional states within the gap.
Although the stripes are mainly 1.25 nm wide (W1), some-
times, 1.0 nm wide strips could be found (see W2 at the
right of Fig. 1(c)). The difference between W1 and W2 cor-
responds exactly to the distance of 0.252 nm between the
3.2. High-temperature phase
atomic rows of Ni Al in the [110] direction, and also coin-
We have found four types of surface structures (Fig. 2),
which are present simultaneously on the surface, i.e. at the
same sample treatment conditions. The images presented in
Fig. 2 have been obtained at approximately the same tun-
nelling regimes. The superstructures not only consist of
stripes, but also buckled lines, and more complex features
such as aligned rhombuses.
Surprisingly, the characteristic dimension of these sur-
face nanostructures is much larger than when originating
from a simple surface reconstruction. The distance between
the stripes and lines within one domain varies from 3 nm in
phase-I up to 7 nm in phase-III. Another surprising fact is
that the shape of the structure appearing on the surface is
not a simple distribution of lines or spots. One can see
some complicated structures within or between the stripes
(Fig. 2(a,b,c)).
3
cides with the expected atomic row distance within Al O .
2
3
This observation confirms that the aluminium oxide film is
generally ordered according to Bardi’s interpretation [16].
However, the incommensurability does not contribute in
the surface superstructures as observed in the experiment.
The only the reason for these superstructures is the intrinsic
inhomogeneity of the aluminium oxide film.
In our case the observed stripe-like superstructure relates
to a quasi-periodical 1-D distribution of defects in the alu-
minium oxide film, and can be well described by a 5 · 1
reconstruction of the crystalline lattice of the aluminum
oxide layer. The white spots on the STM image (Fig. 1(c))
correspond to the sites with enhanced electron density that
can be expected for vacancies in bulk aluminium oxide.
Additionally, we did STS measurement on the stripe and
in the chain of bright spots in between the stripes. Typical
STS plots are presented in Fig. 1(d). While the point on the
The lines in the superstructures run in well-defined direc-
tions with respect to the crystalline lattice of the substrate.
2
Fig. 3. (a) 20 · 20 nm STM image of phase-II obtained at V
t
= 3.4 V, I
t
= 50 pA shows tiny features of the superstructure across the stripes; (b) high-
2
resolution 10 · 10 nm STM current image shows the origin of the features; (c) the same image covered by rhombic patches with non-perturbed lattice. The
shift of atomic rows is indicated by the dotted lines; (d) dI/dV plot measured on the ordered patches (1) and between them (2). Insert shows the typical
dI/dV of the phase-III on the black area (3), on the structured features (4) and the white stripes (5).

O. Kurnosikov et al. / Surface Science 600 (2006) 4375–4379
4379
The stripes within phase-I always run in [100] or [010]
directions; there are only two kinds of domains of this
phase. The stripes of phase-II, phase-III, and phase-IV
form an angle of about 30ꢁ with the [100] or [010] direc-
tions. Therefore there are four differently oriented domains
of each of these phases. About 90 percent of the surface is
covered by phase-I and phase-II (Fig. 2(a) and (b)) with
approximately an equal contribution, while phase–III and
phase-IV are quite rare: it covers only a few percents of
the surface.
In the high resolution STM images, one can see a well-
ordered internal structure with a characteristic size of
about 1 nm (Fig. 3(a–c), only phase-II is shown). Detailed
analysis shows that in these particular points, the crystal-
line lattice is strictly distorted. The dotted lines in
Fig. 3(c) follow the atomic rows. The atomic rows are
not straight, but they have regular breaks. Thus, the crys-
talline lattice of the thin aluminium oxide is strictly dis-
torted. It presents regularly distributed defects of packing
with a well-determined density, but not just some statistical
point defects.
This analysis allows as to represent the crystalline lattice
of aluminium oxide as a mosaic-like structure built up from
rhombic-shaped blocks having non-distorted crystalline
lattice. The rhombic patches are shifted from each other
on several Angstroms (Fig. 3(c)). The rhombic blocks form
their own superlattice. The points of connection of four
blocks result in a higher contrast in the STM images (see
encircled points in Fig. 3(a, b)).
We have found approximately similar rhombic blocks
both in phase-I and phase-II. However the orientation of
atoms and connection of the blocks differs, resulting in dif-
ferent superstructures.
in the electronic state density is visible only at bias voltages
of above 3 V (see the inset in Fig. 3(d)).
4. Conclusions
We have shown that well-ordered Al O layers are, in
2
3
principle, full of structural defects. However, the defects,
which are responsible for surface superstructures, are well
ordered. The low-temperature oxidation regime leads to
simple lines of dislocations that can be described by the
5 · 1 reconstruction. The high-temperature oxidation re-
gime leads to a 2-D quasi-periodical distortion of the
Al O layers, forming well-ordered nanopatches of rhom-
2
3
bic blocks and a distorted lattice in between them.
References
[1] J.S. Moodera, L.R. Kinder, T.M. Wong, R. Meservey, Phus. Rev.
Lett. 74 (1995) 3273.
[
[
2] R. Franchy, Surf. Sci. Rep. 38 (2000) 195.
3] A. Rosenhahn, J. Schneider, C. Becker, K. Wandelt, Appl. Surf. Sci.
1
42 (1999) 169.
[4] V. Rose, V. Podgursky, I. Costina, R. Franchy, H. Ibach, Surf. Sci.
77 (2005) 139.
5
[
[
[
5] R. Franchy, J. Masuch, P. Gassmann, Appl. Surf. Sci. 93 (1996) 317.
6] R.-P. Blum, D. Ahlbehrendt, H. Niehus, Surf. Sci. 366 (1996) 107.
7] J. Libuda, F. Winkelmann, M. B a¨ umer, H.-J. Freund, Th. Bertrams,
H. Neddermeyer, K. M u¨ ller, Surf. Sci. 318 (1994) 61.
[8] Th. Bertrams, A. Brodde, H. Neddermeyer, J. Vac. Sci. Technol. B 12
1994) 2122.
9] M. B a¨ umer, H.-J. Freund, Prog. Surf. Sci. 61 (1999) 127.
(
[
[
[
10] S.G. Addepalli, B. Ekstrom, N.P. Matoto, J.-S. Lin, J.A. Kelber,
Surf. Sci. 442 (1999) 385.
11] P. Gassmann, R. Franchy, H. Ibach, Surf. Sci. 319 (1994) 95.
[12] A. Stierle, F. Renner, R. Streitel, H. Dosch, W. Drube, B.C. Cowie,
Science 303 (2004) 1652.
[
[
[
13] D.R. Jennison, A. Bogicevic, Surf. Sci. 464 (2000) 108.
14] U. Bardi, A. Atrei, G. Rovida, Surf. Sci. 268 (1992) 87.
15] R.M. Jaeger, H. Kuhlenbeck, H.-J. Freund, M. Wuttig, W. Hoff-
mann, R. Franchy, H. Ibach, Surf. Sci. 259 (1991) 235.
All the phases show approximately the same dI/dV
behaviour. That is why only the example of the STS mea-
surements for phase-III is presented in Fig. 3(d). The STS
measurements on the patches and between them reveal a
wide energy gap of about 7 eV. A remarkable difference
[16] G. Kresse, M. Schmid, E. Napetschnig, M. Shishkin, L. Kohler, P.
Varga, Science 308 (2005) 1440.