The structure of an ultrathin VOx (x ≈1) film grown epitaxially on TiO2 (110)

Article abstract of DOI:10.1016/S0039-6028(00)00543-4

The analysis by means of multiple scattering cluster (MSC) calculations of full 2π V 2p, Ti 2p and O 1s X-ray photoelectron diffraction (XPD) patterns from an ultrathin film (～4 ML) of VO_x (x≈1) deposited on the (1 × 1)-TiO₂ (110) (rutile) surface leads to a detailed determination of the structure and orientation of the overlayer and of its epitaxial relationship with the substrate. The comparison of XPD experimental data to theoretical simulations confirms the NaCl-like stacking of the overlayer which is suggested by the direct observation of the experimental 2π plots and reveals its (100) orientation. The [001] azimuth of the overlayer is aligned with the [112] direction on the substrate surface and the overlayer experiences an orthorhombic tensile strain (+ 7%) in order to match the substrate epitaxially, with a consequent 12-16% vertical interlayer spacing contraction. The atomic scale morphology of the (1 × 1)-TiO₂ (110) surface induces a strong buckling in the overlayer along the [110] substrate direction [0.5±0.1 A?].

Full text of DOI:10.1016/S0039-6028(00)00543-4

Surface Science 461 (2000) 118–128
www.elsevier.nl/locate/susc
The structure of an ultrathin VO (x#1) film grown
x
epitaxially on TiO (110)
2
Michela Della Negra, Mauro Sambi, Gaetano Granozzi *
via Loredan 4, 35131 Padova, Italy
Received 24 January 2000; accepted for publication 13 April 2000
Abstract
The analysis by means of multiple scattering cluster (MSC) calculations of full 2p V 2p, Ti 2p and O 1s X-ray
photoelectron diffraction ( XPD) patterns from an ultrathin film (~4 ML) of VO (x#1) deposited on the
x
(
1×1)-TiO (110) (rutile) surface leads to a detailed determination of the structure and orientation of the overlayer
2
and of its epitaxial relationship with the substrate. The comparison of XPD experimental data to theoretical
simulations confirms the NaCl-like stacking of the overlayer which is suggested by the direct observation of the
experimental 2p plots and reveals its (100) orientation. The [001] azimuth of the overlayer is aligned with the [ 1: 12]
direction on the substrate surface and the overlayer experiences an orthorhombic tensile strain (+7%) in order to
match the substrate epitaxially, with a consequent 12–16% vertical interlayer spacing contraction. The atomic scale
morphology of the (1×1)-TiO (110) surface induces a strong buckling in the overlayer along the [ 1: 10] substrate
2
˚
direction [0.5± 0.1 A]. © 2000 Elsevier Science B.V. All rights reserved.
Keywords: Epitaxy; Photoelectron diffraction; Surface structure, morphology, roughness, and topography; Titanium oxide;
Vanadium oxide
1
. Introduction
of polycrystalline vanadium oxide films by resort-
ing to simple and well-characterised single-crystal
model systems in a UHV environment, up to more
fundamental issues related to the properties of
strongly correlated electrons in narrow-d-band
oxides of reduced dimensionality. Within this
context, structural and electronic modifications
due to epitaxial matching to the underlying sub-
strate are particularly appealing to surface and
materials scientists [8].
Our group’s contribution to this field stretches
over several years [9–13]; a summary of our find-
ings has been given recently [13] and will not be
repeated here. Suffice it to say that a careful
control of the growth conditions allows one to
Since the first investigations of titania-supported
vanadium oxide ultrathin films carried out by
Heinrich and co-workers at the beginning of the
last decade [1], the interest in these systems has
been constantly renewed, as an ever-increasing
number of references in the literature demonstrates
[
2–5]. Reasons for such a concern have been
repeatedly stressed by the researchers in the field.
They range from the need to rationalise the cata-
lytic activity [6] and thermochromic properties [7]
*
Corresponding author. Fax +39 49 827 5161.
E-mail address: granozzi@ux1.unipd.it (G. Granozzi)
0
039-6028/00/$ - see front matter © 2000 Elsevier Science B.V. All rights reserved.
PII: S0039-6028 ( 00 ) 00543-4

M. Della Negra et al. / Surface Science 461 (2000) 118–128
119
drive the synthesis of the epitaxial layer towards a
determinate stoichiometry within the wide interval
spanned by the V–O phase diagram. In particular,
our most recent efforts have been directed at
obtaining VO ultrathin layers on TiO (110), with
both cations and anions and led us to conclude
that the overlayer was characterised by a rocksalt
structure, in tune with the bulk structure of VO ,
x
for 0.8<x<1.2. The C symmetry of the curves
2v
has been taken as a preliminary indication of a
(110)-oriented overlayer surface, but the question
was left open for further investigation.
x
2
x being as close as possible to unity (the d phase
in the bulk V–O phase diagram [14]), by resorting
to a synthesis route that implies a stepwise and
controlled oxidation of metal deposits. The oxida-
tion has been performed by means of UHV annea-
ling cycles, in order to promote oxygen diffusion
from the bulk of the substrate to the surface.
Special attention has been placed on specifying the
parameters which determine the quality of the
product and on establishing how to control them
in order to obtain some monolayers of the desired
compound. In fact, it has quickly become evident
that even small variations of the deposition rate,
deposition time, annealing temperature and time
displace the reaction towards different phases [13].
Actually, small modifications can yield lower,
largely non-stoichiometric oxides (e.g., mixtures of
b and c phases [14]) or compounds with a too
high oxidation state, such as V O or even defective
The present paper deals with the structural
characterisation of the VO ultrathin film. For this
x
purpose, full 2p photoelectron diffraction patterns
of V 2p and O 1s photoemission lines from the
VO overlayer and of the Ti 2p core level from the
x
substrate have been measured, and the experimen-
tal data have been analysed by means of multiple
scattering cluster (MSC) simulations. While the
results of this study confirm the NaCl-like stacking
of the epitaxial overlayer, new light has been shed
on its structure, orientation and epitaxial relation-
ship with the substrate. We show that a (100)-
oriented, highly strained and buckled cubic
arrangement is the model which best fits the experi-
mental observations on the basis of an R-factor
analysis.
2
3
VO .
2
After a careful optimisation of the growth
parameters, we finally succeeded in obtaining an
2. Experimental procedure
approximately 4 ML thick VO (x#1) ultrathin
All experiments described were performed
under UHV conditions, at a base pressure of
10−10 mbar in the preparation chamber. The
x
film. A negative binding energy gradient for the
V 2p
photoemission core level from the
3/2
substrate/overlayer interface toward the surface
indicates that interfacial vanadium is characterised
by a (+3) formal oxidation state, which decreases
to (+2) within a couple of monolayers [13]. No
extensive intermixing at the interface has been
detected, and the growth mode was shown to
follow a layer-by-layer mechanism to a good
approximation. From a structural point of view,
local order in the overlayer is supported by X-ray
photoelectron diffraction ( XPD) [15] modulations,
while long-range translational order is excluded
on the basis of the disappearance of the sharp low
energy electron diffraction (LEED) pattern of the
substrate as soon as the first monolayer has been
deposited. In addition, the strong similarity of
O 1s and V 2p XPD modulations (in a series of
TiO (110) crystal (rutile phase, Atomergic) was
2
cleaned by cycles of Ar+ sputtering at 2 keV,
followed by annealing at 823 K in 2×10−6 mbar
of O . The cleaning cycles were repeated until a
2
sharp (1×1) LEED pattern was observed. A VG
Microtech Rear View LEED-RVL900 instrument
was used to check the long-range order of the
surface.
99.95% pure V (Atomergic) depositions were
performed at room temperature (RT) by means
of an electron beam evaporator (Caburn MDC,
model EB90). The deposition rate was about 0.1 V
equivalent ML min−1, as estimated by XPS cal-
ibrations in our previous studies [9,10,13]. The
experiments discussed herein consist of cycles of
UHV evaporations of metallic vanadium on the
3/2
azimuthal curves at selected polar angles) indicated
the analogous structural environment ‘seen’ by
TiO (110) surface at RT, at steps of 5 min
2
(~0.5 ML), with each step followed by a UHV

1
20
M. Della Negra et al. / Surface Science 461 (2000) 118–128
annealing, lasting from 1 to 10 min, at temper-
atures between 130 and 230°C [13], in order to
promote the oxidation due to bulk-to-surface
oxygen migration. Depositions up to 36 min have
been carried out, which correspond roughly to a
total amount of 4 ML of deposited vanadium
oxide.
to each other). The scattering order has been set
to 4, which was sufficient to achieve convergence
and provided the best agreement with the experi-
mental data. An inner potential of 15 V was used
and the angular broadening of the photoelectron
emission direction, to match the spectrometer finite
acceptance angle (± 4°), was also included in the
simulations.
XPS and XPD data were collected with a modi-
fied VG Escalab MKII photoelectron spectrome-
ter, using the Al Ka line of a 300 W Al X-ray
lamp. The XPS scale was calibrated by assuming
the BE of the Ag 3d line at 368.3 eV with respect
3. Experimental results
5/2
to the Fermi level. The sample was mounted on a
two-axis goniometer which allowed the sweeping
of the electron emission direction with an angular
resolution of ± 1° both in polar angle (h, defined
with respect to the normal to the surface) and in
azimuthal angle (w, defined with respect to the
Fig. 1 displays full 2p h/w plots for (a) V 2p,
(b) O 1s and (c) Ti 2p photoemission lines,
obtained on a 4 ML thick VO overlayer deposited
on the TiO (110) substrate. 2p plots for O 1s and
2
Ti 2p emission from the clean and stoichiomet-
3/2
ric (1×1)-TiO (110) surface are shown for refer-
ence (Fig. 1d and e, respectively). All plots are
2
[
1: 10] direction on the substrate surface).
The chemical composition of the deposited layer
reported as x functions, defined as x=(I−I )/I ,
0
0
has been monitored, before and after each annea-
ling cycle, by recording the Ti 2p, V 2p, and O 1s
XPS features at h=0° and h=70° from the sample
normal. The vanadium oxidation state was then
established on the basis of the position and shape
of the V 2p feature (BE=513.6 eV), as explained
where I is the measured absolute intensity and I
0
a smooth background obtained by a polynomial
fit to the azimuthal average of each data set [18].
As already noted in our previous work [13],
vanadium and oxygen modulations are very similar
to each other, while being different from either
Ti 2p or O 1s 2p plots for the clean substrate.
3
/2
elsewhere [13].
3/2
Full 2p angle-scanned XPD plots were obtained
This direct observation strongly suggests that
vanadium and oxygen in the overlayer have ana-
logous structural environments and that the over-
layer structure is different from that of the
substrate. We can exclude the presence of the
metallic vanadium phase on the basis of (a) the
V 2p binding energy (~512.5 eV for the metal
[19–25], 513.6 eV in our case), (b) the V 2p XPD
2p plot symmetry (fourfold for the body-centred
cubic (bcc) metal exposing the (100) surface on
for Ti 2p (E =1027 eV for Ti 2p ), O 1s
k
3/2
(
E =956 eV) and V 2p (E =973 eV for V 2p )
k
k
3/2
core photoemission lines after a conventional
linear background subtraction. Since the V 2p
1/2
line lies in the region where the O 1s Ka satellite
3,4
is present, a procedure was set up [9,10] in order
to obtain only the V 2p photoelectron diffraction
contribution to XPD curves.
The analysis of XPD experimental data has
been accomplished by means of the multiple-scat-
tering calculation of diffraction (MSCD) package
TiO (110) [26], twofold in our case) and (c) the
2
angle-resolved XPS measurements [13], which
[
16], with clusters of at most 160 atoms (see
show that oxygen is present in the ultrathin over-
layer, while titanium is confined to the substrate.
On the basis of similar arguments, we can discard
the V O phase as well, since the corundum struc-
Section 4) for photoelectrons emitted from V 2p
and O 1s levels. Atomic potentials for the calcula-
tion of scattering phase shifts and radial matrix
elements have been generated within a muffin-tin
model using the MUFPOT program [17] for vana-
dium and oxygen atoms. The number of phase
shifts was 18 for oxygen and 19 for vanadium at
both relevant kinetic energies (which are very close
2
3
ture is not consistent with the symmetry of experi-
mental data and with similar modulation for
vanadium and oxygen, and the BE of V 2p in
3
/2
V O would be appreciably higher (515.9 eV [3,4])
2
3
than the value detected on our ultrathin film. In

M. Della Negra et al. / Surface Science 461 (2000) 118–128
121
Fig. 1. Full 2p h/w plots for (a) V 2p, (b) 0 1s and (c) Ti 2p photoemission lines, obtained on a 4 ML thick VO overlayer deposited
on the TiO (110) substrate. 2p plots for (d) O 1s and (e) Ti 2p emission from the clean and stoichiometric (1×1)-TiO (110)
2
3/2
2
surface are shown for reference. (f ) Main substrate azimuthal directions and definition of polar (h) and azimuthal (w) angles.

1
22
M. Della Negra et al. / Surface Science 461 (2000) 118–128
summary, there is quite good evidence that the
ultrathin film has a composition close to VO and
a rock-salt structure, which implies that both ions
experience the same sixfold coordination due to
near neighbouring counterions.
this reason, the structural optimisation of the
ultrathin film by means of theoretical simulations
has been performed on the V 2p XPD pattern only.
As far as the substrate is concerned, the Ti 2p
p plot after the deposition (Fig. 1c) shows the
4. Theoretical simulations and discussion
2
same features as the clean substrate (Fig. 1e),
although the anisotropy is attenuated by the over-
layer. This observation, considered together with
the radically different shape of the V 2p 2p plot
from the overlayer, suggests that an eventual
intermixing of the cationic species at the interface
involves a few atomic percent of the cations which
contribute to the XPD signal and is not detectable
within the sensitivity of the technique. In fact, a
substantial amount of substitutional Ti cations in
the rocksalt lattice of the overlayer would produce
an average rocksalt/rutile Ti 2p XPD pattern
whose features could not be explained as a simple
isotropic attenuation of the pattern from the clean
substrate. Conversely, a significant migration of V
cations from the overlayer to the topmost layers
of the substrate would prevent a description of the
V structural environment in terms of a pure (albeit
distorted) NaCl lattice, while in the following
paragraphs we demonstrate that precisely such a
structural assumption leads to an R-factor mini-
mum in the comparison of experimental data to
simulations. In conclusion, there are good argu-
ments in favour of a reasonably sharp interface
between the two oxides.
In this context, oxygen has a rather particular
status, since it is obviously present both in the
overlayer and in the substrate and, consequently,
the XPD O 1s features are an average of both
substrate and overlayer contributions to the O 1s
photoemission line, especially for take-off angles
close to the sample normal, where the sampling
depth is at its maximum. The fact that the substrate
XPD O 1s modulations are not completely attenu-
ated by the overlayer is indicated by the detectabil-
ity of the still anisotropic Ti 2p signal (Fig. 1c),
which has a kinetic energy (and thus an inelastic
attenuation length) only slightly higher than the
O 1s core level. As a consequence, the V 2p XPD
pattern is a more reliable probe of the local
ultrathin film structure than the O 1s 2p plot – for
According to the above-stated considerations
and to data symmetry, we formulated a set of
plausible hypotheses about the VO overlayer struc-
ture and checked their agreement with experimen-
tal data. For a quantitative analysis we adopted a
modified reliability factor R [27]:
a
∑ (x
i,sim
−Ax
)2
i
i,exp
i,exp
R =min
a
∑
[x2 +(Ax
)2]
A
i i,sim
where x
and x
i,exp
are the simulated and experi-
i,sim
mental x values, respectively and A is a scaling
factor which takes care of the fact that the simu-
lated anisotropy is larger than the experimental
value. The various structural models which have
been investigated lead consistently to a rather
constant value of A=4.9± 0.4, which means that
the experimental anisotropy amounts to just 19–
22% of the simulated values. Reasons for the
reduced amplitude of diffraction modulations have
to be found as usual in the structural disorder of
the overlayer. In fact, the cluster-based method
cannot simulate efficiently the long-range defects
of the real system (dislocations, grain boundaries,
etc.) which are certainly present owing to the lack
of a long-range translational periodicity in the
overlayer. In addition, a strong deviation from an
ideal crystal lattice should be expected in particular
in the case of VO , since even a perfectly epitaxial
x
ultrathin film would still have a high density of
point defects which are not considered in the
calculations. VO in its bulk phase is known for
x
having a high concentration of both cationic and
anionic vacancies (16% of the lattice sites) [14]
even for the 1:1 stoichiometric ratio (x=1).
Another indication of the high level of structural
disorder in the overlayer is the optimised cluster
size used throughout our simulations: clusters of
at most 160 atoms gave consistently the best fits
to experimental 2p plots for all the structural
models considered, while larger clusters produced

M. Della Negra et al. / Surface Science 461 (2000) 118–128
123
simulated patterns where the onset of Kikuchi
bands [15] formation was clearly seen, at variance
with the experimental data, which show no features
related to medium-to-long range scattering (see
Fig. 1a and b).
The main experimental features (Fig. 1a) that
should be reproduced by theoretical simulations
are: (a) the along-normal FS peak (h=0°, w=0°);
two features along the [ 1: 10] azimuth of the sub-
cells. The overlayer with this orientation could
then be either largely incoherent with the substrate
and with bulk-like lattice parameters, or it could
experience a slight tensile strain in order to match
the substrate epitaxially (pseudomorphism would
hold only along the [001] azimuth of the substrate).
The simulated 2p plot for V 2p emission result-
ing from the incoherent bulk-like (110) arrange-
ment is shown in Fig. 3b (the experimental data
are reported in Fig. 3a for direct comparison).
Although there is a good agreement in peak posi-
tions (less so for peak intensities) along the [ 1: 10]
strate (w=0°) at (b) h=48° and (c) h=68°; (d) a
peak at h=56°, w=42°; (e) a peak at h=74°, w=
69° and (f ) a peak along the [001] azimuth of the
substrate (w=90°) at h=60°. Several structural
hypotheses have been put forward in order to
rationalise the experimental observations.
azimuthal direction of the substrate (w=0°), the
FS event (d), found at h=56°, w=42° in the
experimental pattern (see Fig. 1a), is shifted to h=
60°, w=36° in the simulation. In addition, the FS
4
.1. VO (110)/TiO (110)
peak (f ), found at h=60°, w=90° in the experimen-
tal pattern (Fig. 1a), is shifted to h=45°, w=90°
in the simulation. In other words, while the experi-
mental plot is characterised by a nearly fourfold
symmetry of the main FS peaks, the symmetry of
the calculated pattern is more distinctly twofold.
The difference in azimuthal orientation of the main
FS events between experiment and simulation is
not adjustable by small distortions of the overlayer,
and the resulting R-factor is consequently rather
high: R=0.46. The strained epitaxial overlayer
produces an even higher R-factor (R=0.52), thus
demonstrating that this structural hypothesis is
unsatisfactory.
2
Given the binary symmetry of the V 2p 2p plot,
the most natural initial guess is a (110)-oriented
NaCl-like ultrathin film (see Fig. 2a). Since the
VO lattice parameter is 4.09 A [14,28–31], if the
˚
[
[
1: 10] direction of the overlayer is matched to the
001] azimuth of the substrate (Fig. 2b) (giving a
structure indicated as VO(110)/TiO (110),
2
[
1: 10] //[001] , o=overlayer, s=substrate), there
o
s
is a −2.3% lattice mismatch along the short side
of the substrate unit cell. Along the longer side,
the mismatch is reasonably small (−4.9%) if a
coincidence lattice is assumed where three over-
layer unit cells are matched to two substrate unit
Incidentally, it should be noted that the
Fig. 2. Top view of (a) VO (110) and (b) TiO (110). Main azimuthal directions, surface unit cells and unit cell parameters are shown.
2

1
24
M. Della Negra et al. / Surface Science 461 (2000) 118–128
Fig. 3. (a) Experimental V 2p 2p plot; (b) MSC-SW simulation of V 2p emission from VO(110)/TiO (110), [ 1: 10] //[001] (see text).
2
o
s
(
[
c) MSC-SW simulation of V 2p emission from two non-equipopulated, mutually perpendicular domains, consisting of 40%
1: 10] //[ 1: 10] and 60% [ 1: 10] //[001] alignments.
o
s
o
s
VO(110)/TiO (110), [ 1: 10] //[ 1: 10] orientation
domains would produce a perfectly fourfold sym-
metry.) In fact, an R-factor analysis as a function
of the weighting factors leads to a minimum (R=
0.34) when 40% of [ 1: 10] //[ 1: 10] is summed to
2
o
s
(
with the overlayer rotated by 90° with respect to
the VO(110)/TiO (110), [ 1: 10] //[001] structure
2
o
s
discussed above), where no simple matching of the
overlayer to the substrate is possible, gives an even
better R-factor, R=0.40, but of course it still
causes a too pronounced twofold symmetry in the
o
s
60% of [ 1: 10] //[001] . The resulting 2p plot is
o
s
shown in Fig. 3c. However, in this case a modest
improvement in the R-factor is gained at the
expense of a rather unlikely physical assumption,
namely that the overlayer experiences the corruga-
tion potential of the substrate surface to a limited
extent, while it is known from the literature [1]
and from our previous work [9–13] that the
overlayer/substrate interaction is strong when the
vanadium oxide overlayer is obtained by successive
cycles of metal deposition and annealing.
2p plot.
In order to modify the symmetry of the main
FS events from twofold to nearly fourfold while
still retaining the (110) orientation of the over-
layer, one has to assume the presence of two non-
equipopulated, mutually perpendicular domains,
i.e., a weighted average of the [ 1: 10] //[ 1: 10] and
o
s
[
1: 10] //[001] alignments. (Equally populated
o
s

M. Della Negra et al. / Surface Science 461 (2000) 118–128
125
This contradiction can be avoided only by rely-
ing on a new structural hypothesis. We show in
the next section that there is a simpler and more
efficient way of reproducing theoretically the exper-
imental data which, in addition, leads to our best-
fit R-factor value.
matches the [ 1: 12] direction (w=42.3°) on the
underlying TiO (110). An additional in-plane
2
strain, which increases the lattice parameter of the
overlayer by 2.1% along the [001] and by 11.2%
along the [ 1: 10] substrate directions, finally leads
to a perfect matching of two (now rectangular)
surface unit cells of the VO ultrathin film with one
unit cell of the substrate (see Fig. 4). (An alterna-
tive and completely equivalent description refers
the strain to the crystal axes rather than to the
surface axes of the overlayer: now the in-plane
strain is described as a centred-rectangular distor-
tion of the square NaCl (100) unit cell, with a 7%
expansion of the lattice parameter along the [010]
and [001] overlayer directions and a reduction of
the angle between the two unit vectors from 90°
to 84.6°, Fig. 4, left.)
The in-plane tensile strain in turn leads to a
contraction of interplanar distances, which leads
to an orthorhombically strained 3D overlayer unit
cell and increases the polar angle at which the FS
event (d) is detected from its bulk-like value (h=
45°) to the value actually found in our experimen-
4
.2. VO (100)/TiO (110)
2
If we suppose that the FS features (d) at h=
6°, w=42° (Fig. 1a and b) in both V 2p and O 1s
p plots are due to scattering along close-packed
5
2
atomic chains, we find that the latter should be
azimuthally aligned along the [ 1: 12] direction of
the substrate, which is found precisely at w=42.3°
(
It corresponds to the diagonal of a substrate half
unit cell, see Fig. 4). Moreover, we note that a
bulk NaCl-like crystal exposing the (100) plane
has both cationic and anionic chains (at h=45°)
along the [001] azimuth on the surface. Therefore,
when the [011] direction of the overlayer is aligned
with the [ 1: 10] azimuth of the substrate (w=0°),
the [001] azimuth of the overlayer (w=45°) nearly
Fig. 4. Top view of a strained single monolayer of VO (100) pseudomorphic to TiO (110). Two alternative surface unit cells are
2
shown: rectangular with lattice parameters a and b, a≠b (centre bottom) and centred-rectangular with lattice parameters a∞ and b∞,
a∞=b∞ (left). The substrate unit cell is shown as a grey rectangle (right). The matching of overlayer [001] and [011] directions to
substrate [ 1: 12] and [ 1: 10] azimuths, respectively, is shown.

1
26
M. Della Negra et al. / Surface Science 461 (2000) 118–128
Fig. 5. (a) Experimental V 2p 2p plot; (b) best-fit MSC-SW simulation of V 2p emission from VO(100)/TiO (110) without buckling
2
(
(
see text); (c) MSC-SW simulation of V 2p emission from VO(100)/TiO (110) with buckling along the [011] direction of the overlayer;
2
d) MSC-SW simulation of O 1s emission from VO(100)/TiO (110) with buckling along the [011] direction of the overlayer (see text).
2
tal 2p plots (h=56°). An R-factor analysis of this
structure has been performed as a function of the
vertical interplanar spacing in the overlayer. The
best-fit value is now R=0.29, which corresponds
to a large contraction (16.6%) of interplanar dis-
tances. The experimental V 2p 2p plot is compared
with the resulting simulation in Fig. 5 (a and b,
respectively). This arrangement gives the best R-

M. Della Negra et al. / Surface Science 461 (2000) 118–128
127
to be placed above two fivefold-coordinated Ti
atoms, and one V atom should sit in the centre of
the rectangular oxygen cell thus defined, on top of
an in-plane O atom of the substrate. In order to
keep the bond lengths between these three over-
layer atoms (2 O+1 V) and the corresponding
substrate surface atoms at realistic values, they
should be placed at least at approximately 1.7–
˚
.0 A from the substrate surface plane. Since the
2
bridging oxygen atoms are protruding from this
˚
˚
plane by only ~1.3 A, this matching introduces a
buckling of 0.4–0.7 A in the overlayer along its
[
011] direction, as shown schematically in Fig. 6.
The R-factor analysis applied to this model
leads to the overall best-fit value R=0.24, which
˚
corresponds to a buckling of 0.5± 0.1 A in the
overlayer. The best-fit average vertical interlayer
contraction in the VO ultrathin film necessary to
reproduce the main FS events (d) of Fig. 1a is still
16.6%, as in the unbuckled overlayer, although the
simulation appears to be rather insensitive to this
parameter, which can be reduced to 12% with only
a marginal increase of the R-factor value (R=
Fig. 6. Matching of VO (100) to TiO (110). Surface bridging
oxygen atoms of the substrate are shown as white circles. The
main azimuthal directions on both the overlayer and the sub-
strate are indicated. The buckling of the overlayer along its
2
[
011] azimuth is shown.
0.26). The best-fit simulated 2p plot for V 2p
emission is reported in Fig. 5c. Fig. 5d shows the
corresponding simulation for O 1s emission from
the overlayer. As expected, the latter bears a strong
resemblance to Fig. 5c.
factor value so far; it is a strained single-domain
structure, which therefore relies on a strong
overlayer/support interaction, and which also gives
a much better visual fit to the experimental data
than the previously explored arrangements.
A further improvement of the best-fit R-factor
depends on a closer inspection of the interfacial
structure at the atomic level. It is well known that
the (1×1) surface termination of a clean stoichio-
Acknowledgements
This work has been partially funded by
‘
Progetto Finalizzato Materiali Speciali per
Tecnologie Avanzate II’ of the CNR, Rome and
by Ministero della Ricerca Scientifica
Tecnologica (MURST) through the National
Program: ‘Crescita, struttura reattivit a` di
metric TiO (110) surface consists of alternate
2
rows of sixfold- and fivefold-coordinated Ti cat-
e
ions parallel to the [001] direction [32] (see Figs. 2b
and 6). The sixfold-coordinated fully saturated
cations support a row of bridging O atoms pro-
e
˚
superfici di materiali e di film superficiali’.
truding by 1.3 A from the surface for a bulk
termination, while fivefold-coordinated, unsatu-
rated cations are almost coplanar with four O
atoms of the surface layer.
References
If two VO surface unit cells are matched to one
unit cell of the TiO substrate, two bridging oxygen
atoms of the substrate correspond exactly to two
corners of the overlayer unit cell. To complete one
overlayer unit cell, two additional O atoms have
2
[1] Z. Zhang, V.E. Heinrich, Surf. Sci. 277 (1992) 263.
[
2] H. Poelman, K. Devriendt, L. Fiermans, O. Dewaele, G.
Heynderickx, G.F. Froment, Surf. Sci. 377–379 (1997)
819.

1
28
M. Della Negra et al. / Surface Science 461 (2000) 118–128
[
[
3] J. Biener, M. B a¨ umer, R.J. Madix, Surf. Sci. 432 (1999)
78.
4] R.J. Madix, J. Biener, M. B a¨ umer, A. Dinger, Faraday
Discuss. 114 (1999) 67.
5] Q. Guo, S. Lee, D.W. Goodman, Surf. Sci. 437 (1999) 38.
6] I.E. Wachs, B.M. Weckhuysen, Appl. Catalysis A: General
[17] J.B. Pendry, J. Phys. C 13 (1980) 937.
1
[18] H. Li, B.P. Tonner, Phys. Rev. B 37 (1988) 3959.
[19] G.A. Sawatzky, D. Post, Phys. Rev. B 20 (1979) 1546.
[20] A. Lebugle, U. Axelsson, R. Nyholm, N. Martensson,
Phys. Scr. 23 (1981) 825.
[21] C.D. Wagner, W.M. Riggs, L.E. Davis, J.F. Moulder, G.E.
Muilenberg, Handbook of X-Ray Photoelectron Spectro-
scopy, Perkin-Elmer Corporation, Physical Electronics
Division, Eden Prairie, MN, 1979.
[22] R.J. Colton, A.M. Guzman, J.W. Rabalais, J. Appl. Phys.
49 (1978) 409.
[23] V.I. Nefedov, Y.V. Salyn, A.A. Chertkov, L.N. Padurets,
Z.N. Khim, Zh. Neorg. Khim. 19 (1974) 1443.
[24] C.E. Myers, H.F. Franzen, J.W. Anderegg, Inorg. Chem.
24 (1985) 1822.
[
[
1
57 (1997) 67.
[
[
7] C.B. Greenberg, Thin Solid Films 251 (1994) 81.
8] D.K. Fork, J.M. Phillips, R. Ramesh, R.M. Wolf (Eds.),
Epitaxial Oxide Thin Films and Heterostructures, MRS
Symposium Proceedings Vol. 341 Materials Research Soci-
ety, Pittsburgh, PA, 1994.
[
9] M. Sambi, G. Sangiovanni, G. Granozzi, F. Parmigiani,
Phys. Rev. B 54 (1996) 13 464.
[
[
[
10] M. Sambi, G. Sangiovanni, G. Granozzi, F. Parmigiani,
Phys. Rev. B 55 (1997) 7850.
11] P.J. Møller, Zh. Li, Th. Egebjerg, M. Sambi, G. Granozzi,
Surf. Sci. 402–404 (1998) 239.
12] M. Sambi, M. Della Negra, G. Granozzi, Z.S. Li, J. Hoff-
mann Jørgensen, P.J. Møller, Appl. Surf. Sci. 142 (1999)
35.
13] M. Della Negra, M. Sambi, G. Granozzi, Surf. Sci. 436
1999) 227.
14] J. Stringer, J. Less-Common Met. 8 (1965) 1.
15] C.S. Fadley, in: R.Z. Bachrach (Ed.), Synchrotron Radia-
tion Research: Advances in Surface Science, Plenum, New
York, 1992.
16] Y. Chen, M.A. Van Hove,http://electron.lbl.gov/
mscdpack/
[25] J. Kasperkiewics, J.A. Kovacich, D. Lichtman, J. Electron
Spectrosc. Relat. Phenom. 32 (1983) 128.
[26] M. Sambi, E. Pin, G. Sangiovanni, L. Zaratin, G. Gra-
nozzi, F. Parmigiani, Surf. Sci. 349 (1996) L169.
[27] A.M. Bradshaw, D.P. Woodruff, in: W. Eberhardt (Ed.),
Application of Synchrotron Radiation, Springer, Berlin,
1995., Chapter 4.
[28] N. Sch o¨ nberg, Acta Chem. Scand. 8 (1954) 221.
[29] G. Andersson, Acta Chem. Scand. 8 (1954) 1599.
[30] S. Westman, C. Nordmark, Acta Chem. Scand. 14
(1960) 465.
1
[
(
[
[
[31] G. Stenstr o¨ m, S. Westman, Acta Chem. Scand. 22 (1968)
1712.
[32] V.E. Henrich, Rep. Prog. Phys. 48 (1985) 1481.
[