Formation of a well-ordered ultrathin aluminum oxide film on Ni(1 1 1): Determination of its thickness, composition and structure

Article abstract of DOI:10.1016/j.susc.2007.09.050

A well-ordered ultrathin alumina film on top of Ni(1 1 1) has been obtained by an oxidation at 300 K of a thin Ni₃Al(1 1 1) layer epitaxially grown on Ni(1 1 1) and subsequent annealing at 1000 K. The formation of this film was studied by Rutherford backscattering spectrometry (RBS) under channeling conditions, nuclear reaction analysis (NRA), Auger electron spectroscopy (AES) and low energy electron diffraction (LEED). The absolute amounts of both types of atoms participating to the oxide film are, respectively, (2.3 ± 0.3) × 10¹⁵ Al/cm² and (3.3 ± 0.3) × 10¹⁵ O/cm². Hence it appears that this film, with a stoichiometry very close to Al₂O₃, has an oxygen content corresponding nearly to two compact planes of oxygen in bulk crystalline alumina (about 3.0 × 10¹⁵ O/cm²). A commensurate (5√3 × 5√3)R30° superstructure with a lattice parameter of 2.16 nm, can be deduced from the LEED pattern. This superstructure differs from the one observed by other authors for alumina films (of similar thickness and composition) formed on bulk-Ni₃Al(1 1 1), a surface with symmetries and interatomic distances comparable to that of Ni(1 1 1). This difference in structure is most probably connected to the absence, in our case, of any Al atom not strongly bound to O atoms at the interface between the ordered alumina film and the Ni substrate: the Al atoms not involved in the oxide film have diffused deeply in the bulk of the Ni substrate during the high temperature annealing stage needed for alumina ordering.

Full text of DOI:10.1016/j.susc.2007.09.050

Available online at www.sciencedirect.com
Surface Science 602 (2008) 67–76
www.elsevier.com/locate/susc
Formation of a well-ordered ultrathin aluminum oxide film on Ni(111):
Determination of its thickness, composition and structure
*
S e´ verine Le P e´ v e´ dic, Didier Schmaus , Camille Cohen
Institut des Nanosciences de Paris, Universit e´ Pierre et Marie Curie - Paris 6 and Universit e´ Denis Diderot - Paris 7, CNRS UMR 7588,
Campus Boucicaut, 140 rue de Lourmel, 75015 Paris, France
Received 28 July 2007; accepted for publication 27 September 2007
Available online 6 October 2007
Abstract
3
A well-ordered ultrathin alumina film on top of Ni(111) has been obtained by an oxidation at 300 K of a thin Ni Al(111) layer epi-
taxially grown on Ni(111) and subsequent annealing at 1000 K. The formation of this film was studied by Rutherford backscattering
spectrometry (RBS) under channeling conditions, nuclear reaction analysis (NRA), Auger electron spectroscopy (AES) and low energy
electron diffraction (LEED). The absolute amounts of both types of atoms participating to the oxide film are, respectively,
1
5
2
15
2
(
2 3
2.3 ± 0.3) · 10 Al/cm and (3.3 ± 0.3) · 10 O/cm . Hence it appears that this film, with a stoichiometry very close to Al O , has
1
5
2
an oxygen content corresponding nearly to two compact planes of oxygen in bulk crystalline alumina (about 3.0 · 10 O/cm ). A com-
p
p
mensurate (5 3 · 5 3)R30ꢀ superstructure with a lattice parameter of 2.16 nm, can be deduced from the LEED pattern. This
superstructure differs from the one observed by other authors for alumina films (of similar thickness and composition) formed on
3
bulk-Ni Al(111), a surface with symmetries and interatomic distances comparable to that of Ni(111). This difference in structure is most
probably connected to the absence, in our case, of any Al atom not strongly bound to O atoms at the interface between the ordered
alumina film and the Ni substrate: the Al atoms not involved in the oxide film have diffused deeply in the bulk of the Ni substrate during
the high temperature annealing stage needed for alumina ordering.
ꢁ
2007 Elsevier B.V. All rights reserved.
Keywords: Nickel; Ultrathin aluminum oxide; Epitaxy; Superstructure; Rutherford backscattering spectrometry (RBS); Nuclear reaction analysis (NRA);
Auger electron spectroscopy (AES); Low energy diffraction (LEED)
1
. Introduction
This is both due to the small thickness of the films and to
the interaction with the crystalline metallic substrate. These
features can be studied by methods needing charge trans-
port, like scanning tunneling microscopy (STM) or spec-
troscopy (STS) or other ones involving electrons or ion
beams, impossible to perform on massive insulating oxides.
One may be interested either in the oxide film itself, and by
the way its properties depend on its thickness, or in its po-
tential applications. In particular, such films can be used as
a support for metallic atoms or clusters [3]. A two-dimen-
sional (2D) ordering of the metallic deposits can be ob-
tained via different mechanisms, as illustrated by two
recent studies. In the first one, Au atoms deposited at
low temperature on a few monolayers thin MgO film
formed on Ag(001) present a 2D short range order, due
to their electrostatic repulsion, after a charge transfer from
Several interests, both fundamental and technological,
can be found in the study of thin oxide films formed on
metallic substrates, and particularly in well-ordered ultra-
thin oxide films on metal or metal alloy surfaces obtained,
on single crystals, by high temperature oxidation [1,2]. The
stoichiometry and some features of the atomic and elec-
tronic structures observed for such films may differ from
those existing in the near surface region of massive oxides.
*
Corresponding author. Address: Institut des Nanosciences de Paris,
Universit e´ Pierre et Marie Curie Paris 6, CNRS UMR 7588, Campus
Boucicaut, 140 rue de Lourmel, 75015 Paris, France. Tel.: +33 1 44 27 31
61; fax: +33 1 44 27 47 11.
E-mail address: didier.schmaus@insp.jussieu.fr (D. Schmaus).
0
039-6028/$ - see front matter ꢁ 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.susc.2007.09.050

68
S. Le P e´ v e´ dic et al. / Surface Science 602 (2008) 67–76
the substrate, through the oxide [4]. In the second one,
bimetallic Au–Pd clusters, nucleated on particular sites of
the superstructure of an ultrathin alumina film formed on
atomic structure has been recently proposed, following a
combination of extensive ab initio calculations and STM
observations ([8,9] and references therein). In the latter
Ni Al(111), present a good 2D long range order [5]. Such
case, where the surface symmetries of the Ni Al(111) sub-
3
3
a regular nucleation of supported clusters, which are often
observed to present an epitaxial relationship with the oxide
film, tends to narrow, during their growth, their distribu-
tions in size and shape. This can be useful for several appli-
cations, e.g., in catalysis or magnetic devices.
strate and of the compact O planes are both triangular, one
could expect a simpler structure. In fact this structure,
studied by several techniques, appears to be also complex
and is not yet fully resolved at the atomic level [10–16].
The most recent studies, performed, respectively, by low
temperature STM and spot profile analysis low energy elec-
tron diffraction (SPA-LEED) [15] and by dynamic scan-
ning force microscopy (SFM) [16], concluded to an
ultrathin alumina film made of two domains (rotated by
24ꢀ) of a ‘‘true’’ hexagonal superstructure, with a unit cell,
In particular, well-ordered ultrathin aluminum oxide
films formed on massive NiAl or Ni Al single crystals of
3
various orientations have been the subject of many studies
[
2]. These films are generally obtained by an oxidation at
ꢀ
5
an oxygen pressure of at most 10 mbar and around
000 K. This high temperature treatment is made possible
1
(lattice constant of about 4.15 nm), having
a
p
p
by the high melting temperatures of the NiAl or Ni Al or-
( 67 · 67)R47.784ꢀ relation to the Ni Al(111) lattice.
3
3
dered alloys (respectively, 1911 and 1658 K). We list here
the general features presented by these films, on which
most authors agree, even if their assumptions are not al-
ways fully demonstrated: (i) they are composed exclusively
of Al and O atoms; (ii) their stoichiometry is close to
Al O . In what follows we shall then refer to ‘‘alumina’’
The film is terminated by a distorted hexagonal lattice of
oxygen atoms (ions) rotated by 30ꢀ with respect to the sub-
strate lattice. In this surface plane the mean distance be-
˚
tween nearest neighbours oxygen atoms is 2.93 A, to be
˚
compared to 2.80 or 2.75 A, the corresponding distances
in compact oxygen planes, respectively, in c and a massive
alumina.
2
3
films even if it is not strictly Al O ; (iii) the surface of these
2
3
oxide films seems to be O terminated; (iv) as long as the
oxidation is achieved under low oxygen pressure the alu-
mina films formed have a self-limited thickness, always
about 5 A, whatever is the precise oxidation procedure,
with an overall oxygen content corresponding to about
two compact planes of oxygen in bulk alumina.
It would of course be interesting to obtain thin epitaxial
layers of Al oxide on another metallic substrate than a Ni–
Al alloy. A way to achieve this goal is to start from a thin
Ni–Al alloy formed on a Ni single crystal. Recent studies
have shown that such alloyed layers can be prepared with
good crystalline order after a proper annealing and dissolve
in the Ni substrate at higher temperatures [17–19]. One
may then think to oxidize the near surface region of the
thin alloyed layer at low temperatures, and to dissolve
the remaining alloy by high temperature annealing, obtain-
ing hence an Al oxide on top of the Ni substrate. To our
knowledge the only attempt for doing so is reported in
Ref. [17], describing the growth and oxidation of a
˚
One may remark that this self-limited thickness appears
to be too small for uses of such alumina film as insulating
barriers in magnetic tunnel junctions. In junctions having
Ni or Ni Fe (permalloy) as one of the ferromagnetic
8
0
20
electrodes the thickness of the alumina insulating barrier
is at least 1.4 nm [6,7].
Apart from their similarities in composition and thick-
ness it has to be noticed that all the various structures ob-
served on ultrathin alumina films formed on single crystals
of massive Ni–Al alloys differ from the numerous surface
structures, reconstructed or not, observed on a massive alu-
mina single crystal of any structure (a, j, c, d or h). This is
not surprising as the thicknesses of the films are markedly
smaller than the size, in a direction perpendicular to com-
pact O planes, of the unit mesh for any of the bulk struc-
tures of alumina (in corundum, this size is about
Ni Al(100) layer on Ni(100) (square symmetries). After
3
an annealing at 1200 K the so obtained oxide film was
˚
(8 ± 2) A thick, with a hexagonal structure (two-domains)
˚
having a lattice constant 2.87 A, that is the distance be-
tween two adjacent oxygen atoms in the (111) plane of
0
c alumina. This plane was found to be parallel to the
(100) surface of the Ni substrate.
In this paper we report on the preparation and charac-
terization of a well-ordered ultrathin aluminum oxide film
on Ni(111). This film has been obtained after the oxidation
1
.30 nm, corresponding to six (0001) oxygen planes). For
such a small thickness the structure and, probably to a
smaller extent, the precise overall stoichiometry of the alu-
mina film are obviously greatly influenced both by the bulk
characteristics of the Al–Ni alloy used and by the orienta-
tion of its surface.
around 300 K of a very thin Ni Al(111) layer epitaxially
grown on Ni(111) and a subsequent annealing at 1000 K.
The procedure used in order to obtain the epitaxial
3
Ni Al(111) layer, starting from ultrathin Al deposits on
3
a Ni(111) substrate, the characterization and the study of
the thermal stability of the alloy formed are reported in de-
tails in Refs. [18,19]. The surface symmetries of the Ni(111)
Up to now the most studied ultrathin aluminum oxide
films have been formed either on massive NiAl(110) or
on Ni Al(111) substrates. In the former case a complex
substrate and of the Ni Al(111) layer being triangular, we
3
3
structure has to be expected, due to the rectangular surface
symmetry of the substrate, to be compared to the triangu-
lar symmetry of the compact O planes. A model of this
could expect to obtain an aluminum oxide film of simpler
atomic structure than the one found in Ref. [17]. One of
our main aims was to compare the alumina films that we

S. Le P e´ v e´ dic et al. / Surface Science 602 (2008) 67–76
69
attempted to form on Ni(111) to those obtained on mas-
trace of impurities is detected by AES and the narrow spots
of the (1 · 1) LEED pattern are characteristic of a well-or-
dered unreconstructed surface.
The ultrathin aluminum oxide films formed on the
Ni(111) substrate were prepared in the following way:
sive Ni Al(111) [10–16]. Let us underline that the surfaces
3
of these two substrates differ mainly by their compositions,
˚
the lattice parameter of the L1 (fcc) Ni Al (3. 561 A) being
2
3
˚
only 1% larger than the one (3.524 A) of the (fcc) Ni. We
have also extended our study to the oxidation of alloyed
layers obtained from thicker Al deposits, which have a
composition and a crystallographic orientation differing
• Step 1: deposition of Al on the Ni(111) substrate at
130 K. The absolute amount deposited was systemati-
cally determined by RBS experiments, at the deposition
temperature, under channeling geometry and with graz-
ing detection angles around h ꢁ100ꢀ for reasons
depicted in Refs. [18,19]. In what follows the Al
amounts will often be expressed in Al ML, convention-
from Ni Al(111) (‘‘alloying transition’’) [18,19]. In this
3
way, we have also been able to examine up to what extent
the oxide film finally obtained depends on the thickness
and on the precise state of the Al–Ni alloy at the moment
of its oxidation.
The paper is organized as follows: after describing in
Section 2 the experimental conditions and the preparation
procedures, we present in Section 3 our experimental re-
sults concerning the thickness, composition and structure
of the oxide films formed. These results are discussed in
Section 4. Our conclusions are summarized in Section 5.
1
5
2
ally defined as 1 Al ML = 1.86 · 10 Al/cm , corre-
sponding to the areal density of atoms in a (111)
plane of Ni.
• Step 2: annealing at 750 K during 20 min. We have
shown in previous studies concerning Ni–Al alloying
processes that this annealing leads to the formation of
a thin ordered alloyed layer whose composition and
structure depend on the deposited Al amount [18,19].
For very thin Al deposits we have concluded to the for-
mation, after annealing, of a crystallographically and
2
. Experimental
The experiments have been performed in an ultra-high
ꢀ
11
vacuum chamber (base pressure ꢁfew 10
mbar), con-
chemically well-ordered uniform layer of Ni Al(111)
3
nected through a differential pumping section to a
in epitaxy with the Ni substrate. We have found that
the situation is different when the deposited Al amount
2
.5 MV Van de Graaff ion accelerator, allowing to perform
Rutherford backscattering spectrometry (RBS) under
channeling conditions and nuclear reaction analysis
exceeds a critical value N , which may depend on the
C
crystalline quality of the Ni substrate surface, and
1
5
2
(
NRA). This chamber is equipped with a three axes goni-
corresponds, in our case, to N = 8.5 · 10 Al/cm
C
ometer, a four-grid LEED-AES system, a sputtering ion
gun, gas introduction lines and evaporators. More details
about the evaporators, the fixed and movable surface bar-
rier detectors used in RBS in order to count and analyze in
energy the backscattered ions, and also about the reference
sample used for RBS measurements, which contains
(4.5 Al ML). For these thicker deposits we observe the
formation of an ordered and relaxed layer of NiAl(110)
on top of a thin interfacial Ni Al epitaxial layer contain-
3
ing the critical Al amount.
• Step 3: oxidation at RT by backfilling the UHV cham-
ꢀ
6
ber with O2 gas (p(O ) = 1.3 · 10 mbar, during
2
15
2
(
5.63 ± 0.15) · 10 Bi/cm implanted in Si, are available
ꢁ1000 s, that is with an exposure ꢁ1000 L).
in Ref. [18]. The fixed NRA surface barrier detector,
covered with a thin (10 lm) Al leaf in order to stop the
unwanted backscattered ions, is placed at h = 150ꢀ off
the incident ion beam. The reference sample for the mea-
• Step 4: annealing at 1000 K during 20 min.
Here we shall put the emphasis on our results concern-
ing the oxide film obtained, after step 4, consecutive to
16
17
surement of oxygen contents by NRA via the O(d,p) O
an oxidation of a well-ordered Ni Al(111) layer on
3
reaction is a thin film of Ta O , containing (970 ± 20) ·
Ni(111) formed from very thin Al deposits. We shall also
refer to results obtained after oxidation around 300 K of
either a poorly ordered Al layer, deposited at 130 K on
Ni (in this case the step 2 of the above procedure is ex-
cluded) or a well-ordered NiAl(110) layer obtained from
Al deposits exceeding the critical thickness.
2
5
15 16
2
1
0
O/cm , on top of a Si substrate ([20] and references
therein).
The Ni single crystal is a disk (11 mm in diameter, 3 mm
thick), cut from a rod slowly grown by the Czochralsky
method, oriented by XRD, and then mechanically and elec-
trochemically polished. Its mosa ¨ı c spread, checked by
channeling, is found to be of the order of 2 mrad [18].
The temperature of the Ni single crystal can be varied be-
tween 120 and 1100 K. It is measured by a Chromel-Alu-
mel (type K) thermocouple inserted in a hole made by
electroerosion in the side of the crystal. The cleanliness of
the sample surface is checked by Auger electron spectros-
copy (AES), and its surface crystalline order by low energy
electron diffraction (LEED). After several cycles of sputter-
ing at RT with 600 eV Ar ions and annealing at 900 K no
3. Results
We present in this section the results of the measure-
ments performed at each step of the preparation procedure
described above. They concern the surface composition
and the surface structure as well as the absolute quantities
of Ni, Al and O atoms involved in, respectively, the Al
deposited, the alloyed and the oxidized layers. For this pur-
pose we have used AES and LEED for characterizing the

70
S. Le P e´ v e´ dic et al. / Surface Science 602 (2008) 67–76
0
films surface, while RBS-channeling and NRA were used
to characterize the ‘‘bulk’’ of the films.
appearance of the Al peak, which is characteristic of Al in
its metallic state. One concludes to the complete aluminum
oxidation, all the more so that in spectrum 1(c) an Auger
peak appears at 40 eV, which can be attributed to the for-
3.1. Surface composition of the aluminum oxide layer (by
AES)
3
+
mation of Al . The appearance of this Auger peak was al-
ready observed in Refs. [12] and [17] during the formation
of thin alumina films. The AES spectrum (not shown), reg-
istered before the annealing procedure at 1000 K (i.e., just
after oxidation at 300 K), is identical to the one shown in
1(c). This demonstrates that the surface composition is
not affected by the annealing procedure. One may also ob-
serve in Fig. 1c the quasi-total disappearance of the Ni sig-
nal. This feature is a strong indication that the very thin
oxide film formed is continuous, covering entirely the sam-
ple surface.
Another remarkable result is that the same surface com-
position, characterized again by an AES spectrum nearly
identical to the one shown in 1(c), is also observed for
the other oxide films that we have formed, by oxidizing
either an Al deposit or an NiAl (110) layer.
We first focus on the surface of the oxide layer obtained
after oxidation and annealing at high temperature of an
ultrathin Ni Al(111) layer formed on Ni(111). Fig. 1
3
shows the evolution of some low energy Auger peaks, at
different steps of the preparation procedure. A peak at
0
6
1 eV (Ni ), corresponding to an Ni MVV transition, is ob-
served on spectrum 1(a), recorded on clean Ni(111). Only a
0
peak at 68 eV (Al ), corresponding to an Al LVV transition
appears on a spectrum (not shown here) recorded after low
temperature deposition of 2 Al ML, as it can be seen in
0
0
Fig. 6a of Ref. [18]. Both an Ni and an Al peaks are vis-
ible on spectrum 1(b) with relative amplitudes characteris-
tic of the surface of a Ni Al(111) alloy [18], which was
3
formed after annealing the 2 ML thick Al deposit at
7
50 K during 20 min. The spectrum 1(c), registered after
oxidation at 300 K of the previous Ni Al(111) layer and
after annealing at 1000 K during 20 min, is markedly differ-
ent from the previous ones: one can readily observe the dis-
3.2. Overall composition of the aluminum oxide layer (by
RBS and NRA)
3
The Al contents of the various layers involved in this
study have been determined by RBS under channeling
conditions.
Al3+
0
Ni Al0
We shall first present the case of a poorly ordered Al
1
5
2
layer (13.9 · 10 Al/cm , that is 7.5 Al ML) deposited at
30 K, which has been oxidized around 260 K, during the
1
(a)
temperature raise from 130 towards 300 K, and then an-
nealed at 1000 K. The oxidation temperature was chosen
sufficiently low to slow down the Al–Ni alloying kinetics
and then to avoid the presence of any Ni atom at the sur-
face to be oxidized. This has been checked by AES just
prior to the oxidation, and the corresponding spectrum
showed indeed that the surface was purely Al. The conse-
quences of the oxidation and annealing procedures are
illustrated by the two RBS spectra presented in Fig. 2, with
the incident beam along the [110] axis of the Ni(111) sub-
strate. The first spectrum has been recorded at 300 K a few
hours after the oxidation at 260 K. During this time, the
sample was kept at room temperature. The second spec-
trum, also recorded at 300 K, was taken after the annealing
stage at 1000 K. In the two spectra, the Ni background is
low, thanks both to the alignment of the incident beam
along the [110] axis of the substrate and to the grazing
detection direction chosen. The Al peaks are then clearly
visible. The corresponding Al amounts are obtained by
measuring the integrals of these peaks and comparing them
to the integral of the Bi peak in RBS spectra registered on
our RBS reference sample with known Bi content. In the
first spectrum, the Al content was measured to be
(
b)
(c)
4
0
61 68
Energy (eV)
0
Fig. 1. AES spectra registered at 300 K showing the evolution of the Ni
0
3+
(
4
at 61 eV) and the Al (at 68 eV) peaks and the appearance of an Al (at
0 eV) peak along various steps of the preparation procedure: (a) clean
Ni(111); (b) thin film of Ni Al(111) on Ni(111) obtained after annealing
at 750 K an 1.8 ML Al deposit on Ni(111); (c) ultrathin alumina film on
Ni(111) obtained after oxidation at 300 K and annealing at 1000 K of this
15
2
13.5 · 10 Al/cm (7.3 Al ML), which is identical, within
3
our statistical uncertainties, to the amount measured just
after deposition. The integral of the Al peak in the second
spectrum corresponds to a much lower Al amount, equal to
3
Ni Al(111) film.

S. Le P e´ v e´ dic et al. / Surface Science 602 (2008) 67–76
71
1000
trum is not involved in the formation of an oxide film.
On the contrary, after annealing at 1000 K, the Ni peak
integral measured on the second spectrum is only slightly
Al O /Al/Ni(111)
2
3
Al O /Ni(111)
2
3
800
600
400
200
0
higher (by 6%) than I , showing that only a very limited
SP
Nickel
amount of Ni–Al alloy remains at the interface. This is a
good indication that most of the small remaining Al
amount is involved in the formation of a thin superficial
aluminum oxide. It is also an indication that this oxide
does not introduce a strong disordering at the interface.
The Al content of the oxide layer has also been deter-
mined in another case, that is for an Al deposit containing
Aluminium
Oxygen
200
less than the critical Al amount N , which gives rise, after
C
alloying and ordering at 750 K, to a thin Ni Al layer. This
3
case, in which one oxidizes a well-ordered Ni Al(111)
3
150
250
300
350
400
layer, is illustrated in Fig. 3. This figure shows two RBS
aligned spectra, with the incident beam along the [110] axis
of the Ni(111) substrate, recorded, respectively, after step 2
(annealing at 750 K of the Al deposit) and step 4 (oxidation
at 300 K and annealing at 1000 K of the alloy formed at
Channel
4
+
Fig. 2. RBS spectra, recorded at RT, with a 1.5 MeV He beam incident
along the [110] axis of a Ni(111) crystal. : poorly ordered thin film of Al
on Ni(111) with 13.5 · 10 Al/cm , (7.3 Al ML) after oxidation at 260 K;
ultrathin alumina film on Ni(111) with an Al content of
15
2
:
1
5
2
2
.5 · 10 Al/cm (1.3 Al ML) obtained after oxidation at 260 K and
step 2). As the Ni Al alloy formed after step 2 is well-or-
3
annealing at 1000 K of this Al film on Ni(111). Detection angle: h = 94.9ꢀ.
Energy calibration: 2.70 keV/channel. Same number of incident ions for
both spectra.
dered and in full epitaxy with the Ni substrate the Ni peak
integral measured after this stage is nearly equal to ISP. The
Al contents deduced from the Al peaks appearing in Fig. 3
1
5
2
were, respectively: 3.3 · 10 Al/cm (1.8 Al ML) for the al-
1
5
2
15
(
2.5 ± 0.3) · 10 Al/cm (1.3 Al ML). This demonstrates
loyed layer formed after step 2, and (2.3 ± 0.3) · 10 Al/
2
that most of the Al atoms diffuse towards the bulk of the
Ni substrate during the annealing at 1000 K that follows
the oxidation step. In fact, for determining with the lowest
possible uncertainties the Al amounts we have also taken
into account other RBS spectra (not presented here) re-
corded with other detection angles and spectra with the
incident beam along the [111] axis of the Ni substrate
cm (1.2 Al ML) for the ultrathin oxide layer formed after
step 4. The Al content of the alloyed layer formed after step
2 is found identical to the deposited Al amount
(1.8 Al ML), whereas the Al content of the oxide layer,
after annealing at 1000 K happens to be smaller and very
close to the final value obtained in the case presented in
1
5
2
Fig. 2, that is (2.5 ± 0.3) · 10 Al/cm (1.3 Al ML). In
[
(
18,19]. The ± 12% overall relative uncertainty on the value
1
5
2
2.5 ± 0.3) · 10 Al/cm
(1.3 Al ML) measured after
2
000
Ni Al(111) / Ni(111)
annealing at 1000 K has been estimated from counting sta-
tistics, from errors due to background subtraction and
from the precision on the absolute amount of Bi in the
RBS reference sample.
Other interesting observations can be made when com-
paring the contribution of backscattering events on near
surface Ni atoms in the two spectra of Fig. 2. In both cases
this contribution gives rise to a peak in the high-energy re-
gion of the Ni signal. The integral of this peak is signifi-
cantly higher in the first spectrum (before annealing at
3
Al O / Ni(111)
1800
2
3
1
1
1
600
400
200
Nickel
1000
8
6
00
00
Aluminium
Oxygen
200
1
000 K) than in the second one (after annealing). In fact
the peak in the first spectrum has an integral much higher
by 54%) than the one observed on a spectrum registered
400
2
00
0
(
on the bare Ni substrate (not shown), and that we shall call
ISP. The integral ISP corresponds to the contribution of the
first atomic layers of the Ni substrate (the so-called ‘‘chan-
neling surface peak’’) ([20] and references therein). The
high value of the Ni peak integral measured on the first
spectrum thus demonstrates that during the few hours
spent at room temperature after oxidation at 260 K, a sig-
nificant amount of disordered Ni–Al alloy has been formed
1
50
250
300
350
400
Channel
4
+
Fig. 3. RBS spectra, recorded at RT, with a 1.5 MeV He beam incident
along the [110] axis of a Ni(111) crystal. : thin epitaxial film of
1
5
2
Ni
1.8 Al ML), obtained after annealing at 750 K an 1.8 Al ML deposit on
Ni(111); : ultrathin alumina film on Ni(111), with an Al content of
3
Al(111) on Ni(111), with an Al content of 3.3 · 10 Al/cm
(
1
5
2
2
.3 · 10 Al/cm (1.2 Al ML), obtained after oxidation at 300 K and
annealing at 1000 K of this Ni Al(111) film on Ni(111). Detection angle:
3
(
consistently with previous observations [18,19]) and thus
h = 94.9ꢀ. Energy calibration: 2.73 keV/channel. Same number of incident
ions for both spectra.
that a large part of the Al amount measured in this spec-

72
S. Le P e´ v e´ dic et al. / Surface Science 602 (2008) 67–76
both cases all the Al atoms in excess from that common va-
lue have diffused deeply in the bulk of the Ni(111) sub-
strate during the annealing stage at 1000 K. The Ni peak
observed on the spectrum of Fig. 3 recorded after stage 4
1000 K (second spectrum), one can observe small peaks
attributed to O atoms at or near to the surface. In Fig. 3
such a peak is hardly visible on the spectrum recorded be-
fore oxidation, one may then conclude, semi-quantita-
tively, that most of the oxygen cannot be attributed to a
remaining contamination of the sample but is incorporated
during oxidation (step 3). As the O peak remains on spec-
trum 2 registered after step 4, we may also conclude, again
semi-quantitatively that the O atoms fixed during the oxi-
dation step around RT remain in the superficial region
after annealing at 1000 K.
has also an integral close to I , demonstrating again that
SP
the disorder at the interface between the aluminum oxide
film and the Ni substrate is weak.
The third case studied corresponds to the oxidation of a
well-structured Al–Ni alloyed film but now obtained from
an Al deposit (13.4 Al ML) exceeding the critical amount
N . The alloy formed after annealing at 750 K is then
C
NiAl(110)/Ni Al(111) [18,19]. In this case also our analy-
In fact, as can be seen in Figs. 2 and 3, the amount of O
atoms involved in the aluminum oxide layer is hard to eval-
uate with accuracy from the very small O peaks appearing
on RBS spectra, even when using, as here, channeling con-
ditions and a grazing detection geometry. This is a conse-
quence of the very low backscattering cross section on O
atoms, markedly smaller than on Al atoms. Thus, peak
areas corresponding to small O amounts become hard to
evaluate with precision on RBS spectra.
3
sis of the RBS spectra (not shown here) demonstrates that
the overall Al content of the alloy formed after step 2 is
identical (within our experimental uncertainty) to the one
of the Al deposit and that the Al content is markedly smal-
ler after step 4. However, we have observed that the RBS
spectra recorded after annealing at 1000 K differ from the
ones presented in Figs. 2 and 3, exhibiting much higher
backgrounds under the Al peaks. Such a high background
makes much less precise the evaluation of the remaining
Al content after stage 4, which was found to be
For this reason, we have determined the amount of oxy-
1
6
gen in the surface layers by measuring by NRA their
O
1
5
2
16
17
(
4.1 ± 1.2) · 10 Al/cm (2.2 Al ML), a value significantly
content, via the O(d,p ) O nuclear reaction with inci-
1
higher than in the two previous cases described above. Fur-
thermore, the high background observed is due to higher
backscattering yields on subsurface Ni atoms under chan-
neling conditions, which is indicative of a poorer crystalline
quality of the Ni substrate in the interfacial region than in
the cases illustrated by Figs. 2 and 3. This disordering also
induces a huge Ni peak at high energy on the spectrum reg-
istered after stage 4, whose integral is much higher than ISP.
We now focus on the oxygen contents of the layers
formed in the three cases presented above. On the two
RBS spectra presented in Fig. 2, recorded, respectively,
after oxidation (first spectrum) and after annealing at
dent deuterons of 850 keV. The detected particles are high
energy protons, and we analyze those, named p , corre-
1
sponding to the largest cross section, which is associated
1
7
to the first excited state of the residual nucleus O (for
more details see Ref. [20]). The calibration of the oxygen
content associated to a p peak area is obtained from a
1
spectrum registered on the reference sample with known
oxygen amount. The p peak areas in the spectra registered
1
on the Ni(111) crystal at various steps of our procedure
then provide readily the oxygen contents after each step.
Typical NRA spectra, obtained, respectively, for the ref-
erence sample and for the ultrathin oxide film formed by
16O(d,p
) O
17
1
6 17
O(d,p ) O
1
1
1
1
1
400
200
000
40
35
30
25
20
15
10
5
0
8
6
4
2
00
00
00
00
0
12
13
C(d,p) C
1
6
17
O(d,p ) O
0
0
100
200
300
400
0
100
200
300
400
Channel
Channel
1
6
17
2
+
Fig. 4. NRA spectra, showing, as a function of their detection energy the products of the O(d,p) O nuclear reaction, for 850 keV
H
(deuterons)
1
5 16
2
incident on: (a) a Ta
000 K annealing, containing 3.3 · 10 O/cm . Detection angle: h = 150ꢀ. Energy calibration: 6.1 keV/channel. The number of incident ions is eight times
larger for the spectrum (b) than for the spectrum (a).
2
O
5
/Si reference sample containing 970 · 10
O/cm and (b) the ultrathin aluminum oxide on Ni(111), after 300 K oxidation and
15
2
1

S. Le P e´ v e´ dic et al. / Surface Science 602 (2008) 67–76
73
oxidation of Ni Al(111)/Ni(111), are presented in Fig. 4,
three systems oxidized in this study, i.e. an epitaxial layer
3
the number of incident deuterons (beam dose) being eight
times larger in the latter case. These spectra exhibit several
peaks corresponding to various nuclear reaction products,
of Ni Al(111) on Ni(111), a poorly ordered layer of Al
deposited on Ni(111) and an epitaxial layer of NiAl(110)
3
on Ni Al(111)/Ni(111). One may observe that, at least in
3
the p peak mentioned above being the one appearing at
the first two cases, where no evidence was found of the pres-
ence of any Ni atom in the oxide film, neither by AES nor by
RBS, similar compositions and thicknesses of the aluminum
oxide films are obtained, their differences lying within our re-
ported experimental uncertainties. For the O contents, the
extreme values consistent with these uncertainties are,
1
low energy. The wide peak seen in Fig. 4b, typically be-
tween channels 230 and 250, its low energy edge being
1
6
17
superimposed to the O(d,p ) O peak already visible in
0
Fig. 4a, is an artefact which appears progressively and
non linearly all along the recording duration: it is due to
nuclear reactions between incident deuterons and already
1
5
15
2
respectively, 2.1 · 10 and 3.6 · 10 O/cm . If one consid-
1
3
2
implanted deuterons. A very low (ꢁ10 C/cm ) carbon
ers that the O content of a compact oxygen plane in massive
1
5
2
contamination on the oxide film surface can be inferred
alumina is very close to 1.5 · 10 O/cm (respectively,
1
2
13
15 15 2
from the very small peak, due to the C(d,p) C reaction,
appearing around channel 300 in Fig. 4b. The oxygen
amount measured before oxidation, i.e. as well on the bare
substrate, as after Al deposition or after formation of the
ordered Ni–Al alloyed layer, was always found to be in
1.53 · 10 and 1.48 · 10 O/cm for a and c-alumina) all
our measured O contents exclude the possibility for the oxide
2
films for having, at least on an average over the ꢁ1 mm
beam spot area, a number of similar compact O planes differ-
ent from two. In fact, when averaging over the three systems
that we have studied, one obtains, for the O content, a mean
value of 2.9 · 10 O/cm , very close to the O content of
two compact planes in bulk massive alumina (about
1
5
2
the (0.7 ± 0.2) · 10 O/cm interval. This amount does
not correspond to oxygen incorporation in the bulk of
the Ni substrate which was certified to be very small (about
1
5
2
1
5
2
1
5 ppm). It can probably also not be attributed to impuri-
3.0 · 10 O/cm ).
ties at the sample surface: the surface contamination is very
low, as demonstrated by the weak amount of carbon mea-
sured. The oxygen amount found prior to the oxidation
step 3 then probably corresponds to some surface preoxi-
dation of the Ni or of the Ni–Al alloy surfaces. After the
end of our procedure (step 4), one can then assume that
this oxygen amount is part of the oxide formed. The mea-
sured oxygen amount after annealing at 1000 K was nearly
identical, within our experimental uncertainties, to the one
found after oxidation around RT (step 3), evidencing a
negligible oxygen loss during annealing. The oxygen con-
tent of the ultrathin oxide film was determined to be
In Table 1 are also reported the Al atomic fractions xAl
Q /(Q + Q ) deduced from the reported values of Q
=
Al
Al
O
Al
and Q . For the two first cases the extreme values consistent
O
with our uncertainties are, respectively, xAl = 0.36 and
xAl = 0.52, with an average value xAl = 0.44. Thus, the stoi-
chiometry of these oxide films, when averaged over the film
thickness, is very close to that of Al O (x = 0.40). For this
2
3
Al
reason, in what follows we shall refer to them as ‘‘alumina’’
films. However our measured values of xAl are maybe indic-
ative of a small over-stoichiometry in Al, as compared to
Al O . The same Al atomic fraction (x = 0.43) was found
2
3
Al
for a thin layer of aluminum oxide, formed on a massive
NiAl(110) single crystal [8]. We shall come back in the fol-
lowing sections to the third case, for which a larger Al atomic
fraction was obtained, with an also larger uncertainty.
The stoichiometry that we observe after annealing at
1000 K, close to Al O in the first two cases, confirms the
1
5
2
(
(
3.3 ± 0.3) · 10 O/cm . The large relative uncertainty
9%) on this small value, apart from the 2% precision on
the reference sample, mainly comes from low counting sta-
tistics in the p peak corresponding to the oxide film (very
1
small O content). Trying to substantially improve the
counting statistics for the oxide film would lead to unreal-
istic recording times, already of the order of half-an-hour
for the NRA spectrum presented in Fig. 4b.
2
3
assumption that during the annealing step a large fraction
of the Al atoms (in fact all or nearly all those not bonded to
O atoms) diffuse towards the bulk of the Ni substrate. This
is consistent with our previously reported results on the dis-
solution of alloy layers above 800 K [18,19]. In our exper-
iments, the formed alumina is thus likely to be in direct
contact with the Ni(111) surface.
To summarize our results, we report in Table 1 the
amounts of Al and O (QAl and Q respectively, determined
O
by RBS and NRA) involved in the oxide layer after oxidation
around room temperature and annealing at 1000 K for the
Table 1
Overall composition (aluminum amount (QAl), oxygen amount (Q
O
) and atomic fraction of Al (xAl)) for aluminum oxide films obtained after an oxidation
Al(111) film on Ni(111) formed (see text) from a deposit of
.8 Al ML (1 Al ML = 1.86 · 10 Al/cm ); – column 2: a poorly ordered deposit of 7.5 Al ML on Ni(111); – column 3: an epitaxial film of NiAl(110) on
Ni Al(111)/Ni(111) formed (see text) from a deposit of 13.4 Al ML
around room temperature and an annealing at 1000 K of: – column 1: an epitaxial Ni
3
15
2
1
3
Oxidation of Ni Al/Ni(111)
deposit: 1.8 Al ML
3
Oxidation of Al/Ni(111)
deposit: 7.5 Al ML
Oxidation of NiAl/ Ni
deposit: 13.4 Al ML
3
Al/Ni(111)
1
5
2
Q
Al (10 Al/cm )
2.3 ± 0.3
3.3 ± 0.3
2.5 ± 0.3
2.9 ± 0.3
4.1 ± 1.2
2.6 ± 0.5
1
5
2
Q
O
(10 O/cm )
x
Al
0.41 ± 0.05
0.46 ± 0.06
0.61 ± 0.12

74
S. Le P e´ v e´ dic et al. / Surface Science 602 (2008) 67–76
In summary, we observe that, at least in the first two
pared to the ones appearing on Fig. 5a and b obtained, at
the same electron kinetic energy (82 eV), respectively, for
cases we studied, the ultrathin oxide films obtained after
annealing at 1000 K are composed exclusively of Al and
O atoms, with a stoichiometry close to Al O with an O
the clean Ni(111) and for a thin Ni Al(111) film formed
3
on Ni(111). When comparing Fig. 5a and b one has to
keep in mind (see Section 1) that the lattice parameter for
2
3
content corresponding to two oxygen compact planes in
massive alumina.
Ni Al is only 1% larger than the one for Ni. Then the
3
We have performed oxidations at temperatures higher
than RT, up to 750 K, and the results obtained on the
thickness and composition of the oxide films formed re-
main fully consistent with the conclusions given above.
Let us recall that very similar compositions and thicknesses
were also inferred by various authors ([2] and references
therein) when oxidizing massive Ni–Al alloys at rather high
temperatures, around 1000 K.
LEED pattern appearing on Fig. 5b can be, at the first
glance, considered as characteristic of a p(2 · 2) superstruc-
ture on Ni(111), despite the fact that it has been proved
[18] that it corresponds to a thin but three-dimensional
Ni Al(111) film. Let us now comment Fig. 5c on which
3
one can observe the appearance of numerous sharp spots
that can be related to a superstructure with a large lattice
parameter. As some of these spots (see the white arrows
in Fig. 5c) can clearly be attributed to the Ni(111) sub-
strate (see Fig. 5a) this superstructure is commensurate
with the substrate surface. The weakness of the spots
attributed to the substrate confirms that the very thin oxide
is continuous. From the analysis of the pattern LP1 in
Fig. 5c, the structure of the oxide layer was determined
3.3. Surface structure of the aluminum oxide layer (by
LEED)
The surface structure of the alumina films obtained after
oxidation and annealing at 1000 K of thin films (Al, Ni Al
3
p
p
to be simply (5 3 · 5 3)R30ꢀ, with only one domain.
or NiAl) on Ni(111) was studied by LEED. In any case we
observed numerous spots attributed to the presence of a
structured oxide layer. Despite the strong similarities in
composition and thickness of these oxide films (see the pre-
vious section) it appeared that the observed structures were
very different from one case to another.
The primitive mesh of the superstructure (lattice constant:
2
.16 nm) together with the one of the Ni(111) surface (lat-
˚
tice constant 2.49 A) are depicted in Fig. 5d. It is worth-
while to notice that half of the spots appearing in Fig. 5b
do not exist on Fig. 5c: this demonstrates in this case, after
the 1000 K annealing, the absence of any remaining inter-
By far, the best-structured oxide film was obtained after
the oxidation and the annealing at 1000 K of an epitaxial
facial (111) ordered plane of Ni Al, under the oxide film.
3
The LEED patterns of the oxide films obtained after
stage 4 (annealing at 1000 K of the oxides formed around
RT) either on a poorly ordered Al film (7.5 Al ML)
ultrathin Ni Al(111) film on Ni(111) formed from a
3
1
.8 Al ML deposit. The corresponding LEED pattern (that
we shall call LP1) is presented in Fig. 5c, and can be com-
3
Fig. 5. LEED patterns (primary energy: 82 eV) at RT of: (a) a clean Ni(111); (b) a thin film of Ni Al(111) on Ni(111); (c) an ultrathin film of alumina on
p
p
Ni(111) formed from a 1.8 Al ML deposit (LP1); the white arrows indicate the spots attributed to the Ni substrate; (d) in real space: the (5 3 · 5 3)R30ꢀ
superstructure, corresponding to Fig. 5c, of the alumina film (large white unit cell with a parameter of 2.16 nm) superimposed to the Ni(111) surface lattice
˚
(small black unit cell with a parameter of 2.49 A).

S. Le P e´ v e´ dic et al. / Surface Science 602 (2008) 67–76
75
deposited on Ni(111) or on an epitaxial film NiAl(110)/
corresponding to LP2 and LP3 appears very similar to that
observed and analyzed in Ref. [11], after oxidation of a
Ni Al(111)/Ni(111) formed from a 13.4 Al ML deposit
3
(
let us call them, respectively, LP2 and LP3), are very dif-
massive Ni Al(111) substrate.
3
ferent from the pattern LP1 shown in Fig. 5c. The complex
LEED pattern LP3 is nearly identical, with fainter spots, to
the one presented in Fig. 5 of Ref. [11], for the aluminum
oxide formed after oxygen exposure at 1000 K of a
In order to evaluate the amount of Al involved in the
interfacial alloy, we have made the assumption that all the
oxide films formed have the Al O stoichiometry and then
ascribed the Al ‘‘excesses’’ (giving rise to the over-stoichiom-
etries in Al reported in Table 1) to the interfacial alloy. When
2
3
Ni Al(111) massive crystal. In particular one clearly ob-
3
serves in LP3 spots characteristic of a Ni Al(111) surface
subtracting to the Al amounts (Q ), measured by RBS,
3
Al
(
1
see Fig. 5b). This implies that, after the annealing at
000 K, one or a few (111) ordered planes of Ni Al re-
two thirds of the O amounts (Q ), measured by NRA,
O
1
5
2
one finds Al ‘‘excesses’’ of, respectively, 0.1 · 10 Al/cm
3
15 2
mained at the interface between the oxide film and the Ni
substrate. The pattern LP2 also presents strong analogies
with the one presented in Ref. [11], with the difference that
spots specifically attributed to Ni Al(111) are hardly visi-
ble, whereas those corresponding to Ni(111) are sharp
and intense (see Fig. 5a).
(0.05 Al ML), 0.6 · 10 Al/cm
(0.32 Al ML), and
1
5
2
2.4 · 10 Al/cm (1.3 Al ML) for the oxide films described
in columns 1, 2 and 3 of Table 1. As expected, the Al excesses
increase with the initially deposited Al quantity, which was,
respectively, 1.8, 7.5 and 13.4 Al ML. If we now compare
these Al excesses to the Al content (0.46 · 10 Al/cm )
0.25 Al ML) of a bulk-like (111) plane of Ni Al (all these
3
1
5
2
(
3
4
. Discussion
planes are identical.), they correspond, respectively, to 0.2,
.3 and 5.2 numbers of such planes. These last numbers seem
1
When trying to interpret the fact that we observed differ-
to be consistent, at least semi-quantitatively, with the LEED
patterns LP1, LP2 and LP3 where, respectively, no spots (ꢁ0
plane), hardly visible spots (ꢁ1 plane), and definite spots (ꢁ5
ent structures for ultrathin oxide films comprising similar
quantities of Al and O, on the same Ni(111) substrate,
one could be tempted to attribute these differences to the
precise composition and structure of the layer that has been
planes) can be attributed to an ordered Ni Al(111) interfa-
3
cial layer.
oxidized (Ni Al(111) for LP1, Al for LP2 and NiAl(110)
Let us now comment what we think is the only true equi-
librium structure of a well-ordered aluminum oxide film on a
3
for LP3). In fact, this hypothesis is highly questionable,
and, in what follows, we attempt to provide another expla-
nation that we find much more likely and that was briefly
evocated at the end of Section 3.
p
p
Ni(111) substrate, that is the one domain (5 3 · 5 3)R30ꢀ
structure, with a lattice constant of 2.16 nm, (corresponding
toLP1) represented in Fig. 5d. This structure differs from any
one previously observed at the surface of massive alumina or
for thin alumina films formed on any face of a massive Ni–Al
alloy single crystal [2]. In particular, this structure, simpler
and with a smaller lattice constant, differs from that obtained
for alumina films formed at the surface of massive
We must first underline that, contrarily to LP2 and LP3
(
not shown here but already commented in the preceding
section), the LEED pattern LP1 presented in Fig. 5c was
observed for several, but always small, Al deposited
amounts (always less than the critical amount N ). We
C
p
p
then think that LP1 can be associated to the equilibrium
structure corresponding to a well-ordered aluminum oxide
film in direct contact with the Ni substrate, with no Al
atoms in excess (not involved in the oxide film), which is
not the case for the two other patterns.
Ni Al(111), i.e. a two-domains ( 67 · 67) R47.784ꢀ, with
3
a lattice constant of 4.15 nm [15,16]. One could attribute the
difference between the oxide structures to the lattice mis-
match between the two massive substrates. However this
mismatch is small (the lattice parameter is only 1% larger
The patterns LP2 and LP3 were obtained starting from
in massive Ni Al than in massive Ni – see Section 1). More-
3
deposited Al quantities significantly higher than N . The
over, as already indicated, the patterns LP2 and LP3, which
C
Al amounts not implied in the oxide film after the oxida-
tion stage 3 are thus more important than in the cases lead-
ing to the pattern LP1. We then think that LP2 and LP3
correspond to ‘‘transient’’ states in which the annealing
at 1000 K (during 20 min in the three cases) has not fully
dissolved the Al atoms in excess into the Ni substrate. In
these two latter cases, full dissolution of these larger Al
amounts would probably have required higher temperature
or longer annealing times. As already underlined in Section
correspond to alumina formed on very thin Ni Al layers,
most probably strained to adapt to the Ni lattice parameter,
are very similar to those observed for alumina formed on
3
massive Ni Al. Thus, one may then rather think that the
3
much more complex structure that forms on Ni Al(111) is
3
related to the presence, in this case, of Al atoms beneath
p
p
the oxide, whereas the (5 3 · 5 3)R30ꢀ structure described
here corresponds to a direct contact of the alumina film with
the Ni(111) substrate.
3
, one can observe, mainly on LP3, but also more faintly
on LP2, evidence for the presence of ordered Ni Al(111)
5. Conclusion
3
at the interface. This proves the absence of full dissolution
but also indicates that the Al excess at the interface is cer-
The composition (close to Al O ) and thickness (with an
2
3
tainly lower than N , critical value for the ‘‘alloying tran-
sition’’ [18,19]. It is then natural that the oxide structure
overall oxygen content corresponding closely to two com-
pact planes of oxygen in bulk crystalline alumina) of the
C

76
S. Le P e´ v e´ dic et al. / Surface Science 602 (2008) 67–76
various ultrathin aluminum oxide films formed in our
work, after annealing at 1000 K, are found to be, within
our experimental precision, independent of the thickness
of the initial Al deposit and of the temperature at which
the oxidation was achieved (up to 750 K). They were also
observed to be the same when oxidation was carried either
directly on the disordered Al deposit or on an ordered al-
loyed layer formed from this deposit. These results are sim-
ilar to those already reported [2] when forming the oxide,
either by an oxidation followed by an annealing around
different and by far more simple than that of the oxide
which forms on massive Ni Al(111). This difference
3
could appear somewhat surprising since distances be-
tween nearest neighbour atoms in (111) planes of Ni
and Ni Al differ very little. The very complex structure
3
that forms on Ni Al(111) is then most probably related
3
to the presence and lattice location of Al atoms at the
interface between the substrate and the oxide.
Acknowledgement
1
000 K or by a direct oxidation around 1000 K, on various
surfaces of NiAl or Ni Al single crystals. As our results are
We thank Geoffroy Pr e´ vot (INSP-Paris-France) for
fruitful discussions.
3
the first ones based on quantitative measurements of the Al
and O contents, performed by RBS and NRA, they then
confirm and add credit to the assumptions made in the lit-
erature on the thickness and composition of the oxide films
formed on massive Ni–Al alloys. The existence, for oxida-
tions achieved under a low oxygen pressure, of a constant
and self-limited thickness of aluminum oxide found to
form in a broad temperature range and from very different
starting points is indeed intriguing.
References
[1] H.-J. Freund, H. Kuhlenbeck, V. Staemmler, Rep. Prog. Phys. 59
(
1996) 283.
[
[
[
2] R. Franchy, Surf. Sci. Rep. 38 (2000) 195.
3] M. Ba u¨ mer, H.-J. Freund, Prog. Surf. Sci. 61 (1999) 127.
4] M. Sterrer, T. Risse, M. Heyde, H.-P. Rust, H.-J. Freund, Phys. Rev.
Lett. 98 (2007) 206103.
[
[
5] G. Hamm, C. Becker, C.R. Henry, Nanotechnology 17 (2006) 1943.
6] Y. Suezawa, F. Takahashi, Y. Gondo, Jpn. J. Appl. Phys. 31 (1992)
L1415.
In the course of our measurements, we have been able to
form some alumina films on thin Ni Al(111) layers remain-
3
ing on the Ni(111) substrate. The LEED patterns corre-
sponding to these transient situations (where the alloy
remaining after oxidation had not been fully dissolved in
the substrate during the post-oxidation annealing) are simi-
lar to those observed for oxides formed on massive
[
7] R. Guerrero, F.G. Aliev, Y. Tsekovnyak, T.S. Santos, J.S. Moodera,
Phys. Rev. Lett. 97 (2006) 266602.
[8] G. Kresse, M. Schmid, E. Napetschnig, M. Shishkin, L. K o¨ hler, P.
Varga, Science 308 (2005) 1440.
[9] M. Schmid, M. Shishkin, G. Kresse, E. Napetschnig, P. Varga, M.
Kulawik, N. Nilius, H.-P. Rust, H.-J. Freund, Phys. Rev. Lett. 97
Ni Al(111) [15,16]. The structure and orientation of the
3
(
2006) 046101.
oxide formed is thus the same whatever the thickness (down
[10] U. Bardi, A. Atrei, G. Rovida, Surf. Sci. 268 (1992) 87.
to a few (111) planes) of the Ni Al layer on which it lies.
[11] C. Becker, J. Kandler, H. Raaf, R. Linke, T. Pelster, M. Dr a¨ ger, M.
Tanemura, K. Wandelt, J. Vac. Sci. Technol. A 16 (1998) 1000.
3
But the most interesting result that we have observed
concerns the formation of a well-ordered aluminum
oxide film in direct contact with a Ni(111) substrate,
which, to our knowledge, had never been reported up
to now. In order to form such an oxide we have dissoci-
ated the oxidation and ordering steps. A post-oxidation
annealing at 1000 K has been achieved, which allows
both the dissolution of the Ni–Al alloy excess not in-
volved on the oxide already formed in a preceding step
around RT, and the ordering of the aluminum oxide
on top of the Ni crystal. As already stated, this oxide
has about the same thickness and the same stoichiome-
try, very close to Al O , than those formed on massive
[
12] S.G. Addepalli, B. Ekstrom, N.P. Magtoto, J.-S. Lin, J.A. Kelber,
Surf. Sci. 442 (1999) 385.
[13] A. Rosenhahn, J. Schneider, C. Becker, K. Wandelt, J. Vac. Sci.
Technol. A 18 (2000) 1923.
[14] T. Maroutian, S. Degen, C. Becker, K. Wandelt, R. Berndt, Phys.
Rev. B 68 (2003) 155414.
[
15] S. Degen, A. Krupski, M. Kralj, A. Langner, C. Becker, M.
Sokolowski, K. Wandelt, Surf. Sci. 576 (2005) L57.
16] G. Hamm, C. Barth, C. Becker, K. Wandelt, C.R. Henry, Phys. Rev.
Lett. 97 (2006) 126106.
[
[17] A. Wehner, Y. Jeliazova, R. Franchy, Surf. Sci. 531 (2003) 287.
[
18] S. Le P e´ v e´ dic, D. Schmaus, C. Cohen, Surf. Sci. 600 (2006) 565.
[
19] S. Le P e´ v e´ dic, D. Schmaus, C. Cohen, Surf. Sci. 601 (2007) 395.
[20] D. Schmaus, I.C. Vickridge, in: P. Marcus, F. Mansfeld (Eds.),
Analytical Methods for Corrosion Science and Engineering, CRC,
Taylor & Francis group, Boca Raton, London, New York, Singa-
pore, 2006, p. 103.
2
3
Ni–Al
alloys.
However,
the
one
domain
p
p
(
5 3 · 5 3)R30ꢀ structure that we observe is markedly