Ultrathin ZrO2 films on Si-rich SiC(0 0 0 1)-(3 × 3): Growth and thermal stability

Article abstract of DOI:10.1016/j.susc.2007.04.026

The growth and thermal stability of ultrathin ZrO₂ films on the Si-rich SiC(0 0 0 1)-(3 × 3) surface have been explored using photoelectron spectroscopy (PES) and X-ray absorption spectroscopy (XAS). The films were grown in situ by chemical vapor deposition using the zirconium tetra tert-butoxide (ZTB) precursor. The O 1s XAS results show that growth at 400 °C yields tetragonal ZrO₂. An interface is formed between the ZrO₂ film and the SiC substrate. The interface contains Si in several chemically different states. This gives evidence for an interface that is much more complex than that formed upon oxidation with O₂. Si in a 4+ oxidation state is detected in the near surface region. This shows that intermixing of SiO₂ and ZrO₂ occurs, possibly under the formation of silicate. The alignment of the ZrO₂ and SiC band edges is discussed based on core level and valence PES spectra. Subsequent annealing of a deposited film was performed in order to study the thermal stability of the system. Annealing to 800 °C does not lead to decomposition of the tetragonal ZrO₂ (t-ZrO₂) but changes are observed within the interface region. After annealing to 1000 °C a laterally heterogeneous layer has formed. The decomposition of the film leads to regions with t-ZrO₂ remnants, metallic Zr silicide and Si aggregates.

Full text of DOI:10.1016/j.susc.2007.04.026

Surface Science 601 (2007) 2390–2400
www.elsevier.com/locate/susc
Ultrathin ZrO2 films on Si-rich SiC(0001)-(3 · 3):
Growth and thermal stability
a
b
a
b
c
P.G. Karlsson , L.I. Johansson , J.H. Richter , C. Virojanadara , J. Blomquist ,
c
a,
*
P. Uvdal , A. Sandell
a
Department of Physics, Uppsala University, Box 530, SE-75121 Uppsala, Sweden
Department of Physics and Measurement Technology, Linko¨ping University, SE-58183 Lund, Sweden
b
c
Chemical Physics, Lund University, Box 118, SE-22100 Lund, Sweden
Received 5 February 2007; accepted for publication 4 April 2007
Available online 11 April 2007
Abstract
The growth and thermal stability of ultrathin ZrO2 films on the Si-rich SiC(0001)-(3 · 3) surface have been explored using photoelec-
tron spectroscopy (PES) and X-ray absorption spectroscopy (XAS). The films were grown in situ by chemical vapor deposition using the
zirconium tetra tert-butoxide (ZTB) precursor. The O 1s XAS results show that growth at 400 ꢀC yields tetragonal ZrO2. An interface is
formed between the ZrO2 film and the SiC substrate. The interface contains Si in several chemically different states. This gives evidence
for an interface that is much more complex than that formed upon oxidation with O2. Si in a 4+ oxidation state is detected in the near
surface region. This shows that intermixing of SiO2 and ZrO2 occurs, possibly under the formation of silicate. The alignment of the ZrO2
and SiC band edges is discussed based on core level and valence PES spectra. Subsequent annealing of a deposited film was performed in
order to study the thermal stability of the system. Annealing to 800 ꢀC does not lead to decomposition of the tetragonal ZrO2 (t-ZrO2)
but changes are observed within the interface region. After annealing to 1000 ꢀC a laterally heterogeneous layer has formed. The decom-
position of the film leads to regions with t-ZrO2 remnants, metallic Zr silicide and Si aggregates.
ꢁ 2007 Elsevier B.V. All rights reserved.
Keywords: Zirconium dioxide; Silicon carbide; Chemical vapor deposition; Semiconductor–insulator interfaces; Synchrotron radiation photoelectron
spectroscopy; X-ray absorption spectroscopy
1. Introduction
information on interface states and there are consequently
a number of PES studies of the SiO2/SiC interface [4–8].
Less attention has been paid to the deposition of a dis-
similar oxide on SiC. An important set back for the use
of SiO2/SiC in SiC based MOS devices is the formation
of electronically active interface states [2,3]. A high density
of interface trap states near the conduction band limits the
channel mobility. The use of an alternative oxide as gate
insulator may be a way to circumvent this problem. The
Al2O3/6H-SiC interface formed by atomic layer deposition
shows for example a significantly reduced density of inter-
face states [9].
Silicon carbide is a material suitable for high tempera-
ture, high power, high voltage and high frequency devices
and sensors. Substantial efforts have been made in order
to integrate SiC in MOSFET technology [1–3]. This work
commonly entails the formation of a gate-insulating SiO2
layer on top of the SiC substrate. The electronic properties
are critically dependent on the interface properties. Photo-
electron spectroscopy (PES) is a very useful tool to obtain
ZrO2 is an oxide with many attractive properties such as
high thermal stability, low thermal conductivity and high
dielectric constant. Pure and modified ZrO2 is used for
*
Corresponding author.
E-mail address: anders.sandell@fysik.uu.se (A. Sandell).
0039-6028/$ - see front matter ꢁ 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.susc.2007.04.026

P.G. Karlsson et al. / Surface Science 601 (2007) 2390–2400
2391
example as thermal barrier coating on gas turbine blades
[10] and it is considered as a gate oxide material in Si based
MOSFETs [11]. It has been demonstrated that MOS de-
vices fabricated using ZrO2 as dielectric show promise
[12,13].
From the above follows that the combination of the two
compounds SiC and ZrO2 can be most useful for devices
and materials that need to function at high temperatures.
For example, ZrO2 coated SiC is used as catalyst support
and in turbine parts [14,15]. ZrO2 may also prove to be
an alternative as gate oxide in SiC based MOSFET devices.
Consequently, a careful characterization of the ZrO2/SiC
system is well motivated.
The preparation of a well-defined substrate surface
greatly facilitates the realization of a good oxide/SiC inter-
face. The Si-rich SiC(0001)-(3 · 3) surface reconstruction
is well defined and characterized in detail [4–8,16–20]. It
is prepared by Si deposition at elevated temperature fol-
lowed by additional annealing at higher temperature. From
STM and LEED holography data a structural model has
been derived in which the (3 · 3) surface is composed of
Si-adatoms, Si trimers and a Si adlayer on top of the out-
ermost Si–C bilayer [19,20]. Recent photoemission results
reveal the presence of three surface components in the Si
2p spectrum, consistent with the structure model [8]. Exten-
sive oxidation of the Si-rich (3 · 3) surface using O2 at
800 ꢀC results in complete oxidation of the surface Si atoms
[7,8]. Only one suboxide, Si1+, is observed besides the Si4+
species of fully developed SiO2 [7,8].
Upon new injection of the storage ring spectra have been
re-measured. The XAS spectra were calibrated by using
first- and second-order light from the monochromator.
The investigations were carried out on an n-doped 8ꢀ off
cut 4H-SiC(0001) sample on which a (3 · 3) surface recon-
struction was prepared before the ZrO2 deposition. The
(3 · 3) surface reconstruction was prepared the common
way by Si deposition at a substrate temperature of
ꢀ800 ꢀC and subsequent annealing to ꢀ1000 ꢀC. This re-
sulted in a well-ordered (3 · 3) surface as confirmed by
low energy electron diffraction (LEED).
ZrO2 was deposited at a sample temperature of 400 ꢀC
using zirconium tetra tert-butoxide (ZTB) as single source
precursor. The ZTB liquid was purified by multiple freeze–
pump–thaw cycles. The sample was exposed to ZTB via a
dosing tube (10 mm in diameter) positioned a few mm from
the sample surface. The high vacuum parts of the ZTB dos-
ing system were thoroughly baked out and subsequently
passivated by exposing them to ZTB. The pressure during
deposition was 2 · 10À8 mbar, read as the background
pressure in the chamber.
The film was grown in intervals between which measure-
ments were performed. After a total exposure time of 1700 s
the resulting film was subsequently annealed to higher tem-
peratures: 600, 800 and 1000 ꢀC. The duration of each
annealing was 60 s and the pressure never exceeded 3 ·
10À9 mbar during annealing. The changes observed after
annealing to 600 ꢀC were negligible.
In this paper, we present an investigation of the growth
and thermal stability of ultrathin ZrO2 on the Si-rich
SiC(0001)-(3 · 3). The growth is undertaken by chemical
vapor deposition under ultrahigh vacuum conditions. The
growth is characterized in a stepwise fashion by employing
synchrotron radiation excited photoelectron spectroscopy
(PES) and X-ray absorption spectroscopy (XAS). Subse-
quent annealing of the deposited film was performed in or-
der to study the thermal stability of the system.
3. Results and discussion
3.1. Film growth
3.1.1. Overlayer thickness and local structure
Fig. 1 shows the film thickness as function of ZTB expo-
sure time. The film thickness is estimated from the attenu-
ation of the component in the Si 2p PES signal attributed
2. Experimental
50
40
30
20
10
0
The spectra were recorded at beamline D1011 at the
Swedish National Synchrotron Radiation Facility MAX-
II [21]. The end station comprises a Scienta 200 mm radius
hemispherical electron energy analyzer for PES measure-
ments and a multichannel plate for recording of XAS spec-
tra by collection of Auger electrons. The base pressure was
in the low 10À10 mbar or better. The sample holder was
made of tantalum and was etched in HCl prior to mount-
ing. The tantalum clips holding the sample were annealed
in vacuum. The PES spectra were all recorded at normal
emission and the binding energy (BE) scales are referenced
to the Fermi level of a sample holder clip. The intensity of
each PES spectrum has been normalized to the current in
the synchrotron storage ring. The intensity of a spectral
component is defined as the integrated line profile. The rel-
ative magnification of each spectrum is given in the figures.
0
500
1000
1500
ZTB exposure time [s]
Fig. 1. Estimated film thickness as a function of deposition time. Growth
conditions: Sample temperature 400 ꢀC and ZTB background pressure
2 · 10À8 mbar. The film thickness is estimated from the attenuation of the
component in the Si 2p PES signal attributed to SiC bulk.

2392
P.G. Karlsson et al. / Surface Science 601 (2007) 2390–2400
to SiC bulk. The Si 2p intensities were obtained from spec-
tra measured using the photon energy 758 eV. The mean
free path (k) is set to 20 A. The choice of k is supported
by a comparison made for ZrO2 on Si(100), where the film
thickness estimated from an independent X-ray reflectivity
measurement is in good agreement with that obtained by
Si(100)-(2 · 1) where the fingerprint of crystalline ZrO2 is
observed at an earlier stage [23]. The similarity to
Si(111)-(7 · 7) is not unexpected given that the Si overlayer
on SiC(0001) and the Si(111) surface both have hexagonal
symmetries. Another property that can be important for
the geometric properties of the ZrO2 film is the number
of dangling bonds per surface atom [24]. The Si overlayer
on SiC(0001) and the Si(111) surface both have one dan-
gling bond per surface atom while the Si(100) surface has
two dangling bonds per surface atom. It has been argued
that two dangling bonds are more favorable for epitaxial
ZrO2 growth [24]. This is consistent with the observation
of a delayed formation of crystalline ZrO2 on Si-rich
SiC(0001)-(3 · 3) and Si(111)-(7 · 7) as compared to
growth on Si(100)-(2 · 1). However, it is most likely that
also the presence of hydrocarbon fragments from the
ZTB molecules strongly influence the geometric properties
of the film at the initial stages of growth.
˚
˚
PES using k = 20 A at 758 eV photon energy. Note that
zero film thickness is defined for the Si-rich SiC(0001) sur-
face, which actually has silicon layers on top of the SiC
substrate. Fig. 1 shows that the growth series encompasses
˚
˚
thicknesses ranging from 4 A to 51 A. Given that the lattice
˚
˚
parameters for t-ZrO2 are a = b = 3.6 A and c = 5.2 A [22],
the films range from approximately monolayer thickness
up to an equivalent of about 10 t-ZrO2 layers. However,
this estimate is rather crude since the formation of an ex-
tended interface must be taken into account, cf. below.
Selected O 1s XAS spectra are presented in Fig. 2. The
spectral structures become sharper for each deposition
step, indicating that the ZrO2 overlayer becomes increas-
ingly ordered, i.e. crystalline. The spectrum measured for
3.1.2. Photoemission of core levels
˚
the 51 A thick film is similar to that of a thick film on
˚
˚
The Zr 3d PES spectra measured for 4 A and 51 A thick
films are presented in Fig. 3. It is found that each of all Zr
3d spectra in the growth series can be described by one
Si(100). Since the spectrum for the ZrO2/Si(100) system
is typical for tetragonal ZrO2 it can be concluded that the
tetragonal phase also forms at sufficient film thickness on
SiC(0001)-(3 · 3) at 400 ꢀC [23]. In contrast, the signature
of t-ZrO2 is not as pronounced in the O 1s XAS spectrum
˚
at a film thickness of 28 A. This is most clearly seen by
Zr 3d
hν = 290 eV
observing the differences in the photon energy region
542–545 eV. The film thickness dependence of the O 1s
XAS spectrum is similar to that reported for film growth
on Si(111)-(7 · 7) but different from film growth on
×
3.6
1.0
(hν = 758 eV)
1000 °C
×
O 1s XAS
ZrO2/Si(100)
× 1.0
1700 s
51 Å
×
6.5
× 1.3
× 12
1700 s
51 Å
40 s
4 Å
1280 s
43 Å
40 s
4 Å
188
186
184
182
180
178
Binding energy (eV)
530
535
540
545
550
555
560
Fig. 3. Zr 3d PES spectra for selected situations. The total energy
resolutions are 150 meV (290 eV photon energy) and 650 meV (758 eV
photon energy). Each of all Zr 3d spectra in the growth series can be
described by one spin–orbit pair attributed to Zr4+ (bottom and middle
Photon energy (eV)
Fig. 2. O 1s XAS spectra for selected deposition steps. The photon energy
resolution is 150 meV. The structures become sharper for each deposition
step, indicating that the ZrO2 overlayer becomes increasingly crystalline.
The spectrum measured for the 51 A thick film is similar to that of
tetragonal ZrO2 [23].
spectra). Upon subsequent annealing to 1000 ꢀC
a new additional
component having an asymmetric line shape has appeared (top spectrum,
large dots). The spectrum does not change dramatically when changing the
surface sensitivity of the measurement (top spectrum, small dots).
˚

P.G. Karlsson et al. / Surface Science 601 (2007) 2390–2400
2393
˚
spin–orbit pair. The Zr 3d5/2 BE is 183.6–183.9 eV, with the
lower values found for thinner films. This BE range moti-
vates an assignment to Zr4+ species [23]. Introduction of re-
duced species, which should be shifted at least À1 eV from
the Zr4+ component [25], cannot be motivated. A broaden-
ing of the peaks for thinner films is furthermore observed.
This can be related to e.g. an increased inhomogeneity for
the situations where the film is likely to have a less defined
stoichiometry and geometric structure.
The corresponding O 1s PES spectra (not shown) reveal
only one component at about 531.7 eV BE throughout the
growth series. The O 1s binding energies previously re-
ported for ultrathin SiOx on Si [26,27] and ZrO2/ZrSixOy
on Si [23,28] are similar to within 0.5 eV. Since the FWHM
of the O 1s line is typically at least 1 eV, the small difference
in BE could at most give rise to a broadening and not to
two well-resolved peaks. The information gained from
the O 1s spectra is thus limited with respect to the chemis-
try involved in the growth and interface formation. How-
ever, the O 1s BE is useful for a determination of the
location of the conduction band edge for ZrO2 as discussed
below.
sition (4 A) the relative shift in BE amounts to +0.34 eV.
˚
The BE remains at a constant value up to 28 A. For thicker
films there is an additional shift of about 0.15 eV. It cannot
be excluded that sample charging contributes to this effect.
An additional C 1s state at about 285.5 eV BE appears
after deposition. This feature is denoted C1 in Fig. 4a.
Fig. 4b shows the integrated intensities of the C1 state
and the C 1s peak due to SiC. The peak related to SiC dis-
plays an exponential decay with exposure time. In contrast,
the C1 state has a rather constant intensity up to 320 s of
˚
exposure time (about 20 A film thickness) after which it de-
creases slowly.
We propose that the C1 state in Fig. 4a is associated
with methyl groups stemming from the ZTB precursor.
This assignment is supported both by the C 1s BE and
the intensity behavior for increasing film thickness. For
ZTB-mediated ZrO2 growth on Si(100) and Si(111) a C
1s state due to methyl groups was observed at 285.3 eV
[23]. The intensity behavior of the C 1s peak due to methyl
groups upon growth on Si(100) and Si(111) is found to be
very similar to that shown in Fig. 4b. We argue that the
point at which the methyl C 1s peak begins to decrease
coincides with the point of interface completion [23]. From
this follows that the methyl species is most likely correlated
to the presence of Si surface atoms. The amount of surface
Si atoms is expected to decrease once an extended ZrO2
film forms. However, the decrease observed for films
The C 1s PES spectra are presented in Fig. 4a. For the
Si-rich SiC(0001)-(3 · 3) surface there is one symmetric
peak at 283.0 eV BE attributed to SiC bulk. Upon ZTB
exposure the SiC bulk peak is shifted, most probably be-
cause of changes in the band bending. After the first depo-
b
a
C 1s
h
ν
= 350 eV
SiC
SiC*
5
4
3
2
1
0
SiC
C1
C1
×
3.2
1000 °C
800 °C
×
26
26
×
1700 s
51 Å
40 s
4 Å
×
1.5
1.0
0.34 eV
×
SiC(0001)-(3x3)
286
285
284
283
282
281
0
400
800
1200
1600
Binding energy (eV)
Exposure time [s]
Fig. 4. (a) C 1s PES spectra for selected situations. The total energy resolution is set to 250 meV. Upon ZTB exposure the SiC bulk peak is shifted, most
probably because of changes in the band bending. Small amounts of ZTB by-products are detected (C1 species). Upon annealing to 1000 ꢀC the two C 1s
components denoted SiC and SiC* can both be assigned to SiC bulk species. The C 1s high binding energy component (denoted SiC) is attributed to areas
where the ZrO2 overlayer remains. The C 1s low binding energy component (denoted SiC*) is attributed to areas where there is a metallic compound
overlayer. (b) Integrated C 1s intensities for the SiC bulk component and the precursor induced C1 component, assigned to methyl groups.

2394
P.G. Karlsson et al. / Surface Science 601 (2007) 2390–2400
˚
thicker than 20 A is slow. A feasible explanation for this is
the tendency for intermixing of Si4+ with ZrO2 as discussed
in more detail below.
Si 2p
SiO2/SiC(0001)
ZrO2/Si(111)
hν = 130 eV
Moreover, for growth on Si(111) and Si(100) the inter-
face region contains carbonaceous species which give rise
to C 1s states having about 283–284 eV BE [23]. These
are assigned to different hydrocarbon fragments with vary-
ing degree of coordination to Si [29,30]. Fig. 4a shows that
˚
the spectrum for a 51 A thick film has some intensity be-
tween the bulk and C1 features. Thus, it cannot be ex-
cluded that several different carbon species are present
also for growth on the Si-rich surface. However, the pres-
ence of the SiC bulk peak precludes a more detailed quan-
titative assessment of the ZTB related structures.
1700 s
51 Å
× 36
Selected Si 2p PES spectra measured using two different
photon energies are presented in Figs. 5a and 5b. By com-
paring the more surface sensitive measurement using
130 eV photons (Fig. 5a) with the less surface sensitive
measurement using 350 eV photons (Fig. 5b) it is possible
to distinguish between bulk and surface species.
160 s
10 Å
× 2.9
Each Si 2p spectrum for the Si-rich SiC(0001)-(3 · 3)
surface can be described by four spin–orbit pairs [8]. These
correspond to the SiC bulk species (SiC) and different spe-
cies within the silicon overlayer (S1, S2 and S3). The fits
were performed under the constraint that the width and rel-
ative position for each component are independent of pho-
ton energy. The lower resolution attained at higher photon
energy (200 meV compared to 60 meV) is of no conse-
quence as it is considerably smaller than the intrinsic width
of the individual components (above 600 meV).
For each pair of Si 2p spectra we have employed a sim-
ilar fitting procedure. Each spectrum is fitted using six
spin–orbit pairs: One for the bulk (SiC), three originating
from the silicon layer (SL1, SL2 and SL3) and two for oxi-
dized silicon species (SiOX and Si4+). In each spectrum the
Gaussian FWHM broadening is 0.75 eV for the bulk com-
ponent, 0.60 eV for each of the three silicon layer compo-
nents and 1.10 eV for each of the two oxidized silicon
components.
40 s
4 Å
× 1.7
S3
S2
SiC(0001)-(3x3)
S1
× 1.0
106
104
102
100
98
Binding energy (eV)
Fig. 5a. Selected Si 2p PES spectra measured using 130 eV photons from
the film growth series. The total energy resolution is 60 meV. Labels are
defined and discussed in the text. At the top relative binding energy
positions from references are marked. The SiO2/SiC(0001) marker is set
relative to the SiC bulk peak position and indicates from left to right the
positions for Si4+, Si1+ and SiC bulk [8]. The ZrO2/Si(111) marker is
relative to the Si4+ position and indicates from left to right the positions
for Si4+, Si3+, Si2+, Si1+ and Si bulk [23].
The identification of the bulk component from the sur-
face sensitivity dependence is consistent with the C 1s spec-
tra with regard to the BE shifts. The Si 2p and C 1s spectra
measured using 350 eV photons are for each situation re-
corded in sequence and with the same experimental set-
tings. Therefore, the relative binding energy shifts are
very accurate. The Si 2p bulk peak is shifted +0.33 eV at
oxide film. The spectral shapes of the low binding energy
˚
˚
states are very similar from 4 A to 51 A. That is, fits can
be performed in which the SL1, SL2 and SL3 states have
approximately fixed relative intensities. The BE:s of the
three components for the silicon layer after ZTB exposure
are however not equivalent to the S1–S3 states of the clean
reconstructed surface.
The most plausible interpretation is that the SL1–SL3
states are due to Si atoms originating from the Si adlayer.
Previous results show that extensive oxidation using O2 re-
sults in complete oxidation of the Si atoms in the Si adlayer
[8]. A complete oxidation does apparently not occur upon
reaction with the ZTB precursor at 400 ꢀC. This can be due
˚
4 A film thickness, nearly identical to the +0.34 eV found
in the C 1s spectra. It has previously been reported that
the BE:s of the core levels trace the shift of the valence
band maximum for heterojunctions of semiconductors
and insulators [31–33]. That the binding energies of the
components attributed to SiC bulk species in the Si 2p
and C 1s core level spectra respond approximately equally
to changes in the band bending is thus a very reasonable
result.
Fig. 5a clearly shows that there are states at lower BE:s
than the bulk component also after the formation of an

P.G. Karlsson et al. / Surface Science 601 (2007) 2390–2400
2395
Turning next to the oxidized Si species a comparison to
previous studies is valuable for the identification. At the
top of Fig. 5a relative Si 2p3/2 BE positions from two ref-
erence systems, SiO2/SiC(0001) and ZrO2/Si(111), are
indicated. The SiO2/SiC(0001) marker is set relative to
the SiC bulk peak position and indicates from left to right
the positions for Si4+, Si1+ and SiC bulk [8]. The ZrO2/
Si(111) marker is relative to the Si4+ position and indicates
from left to right the positions for Si4+, Si3, Si2+, Si1+ and
Si bulk [23].
Si 2p
= 350 eV
hν
1700 s
51 Å
×
23
The assignment of the Si4+ state is motivated by the
binding energy [5,6]. Following the binding energy distri-
bution of the oxidized silicon species previously found for
growth on Si(111)-(7 · 7) [23] the state labeled SiOX may
contain contributions from both Si2+ and Si3+ species,
see Fig. 5a. The Si1+ component is possibly obscured by
the SiC bulk component and/or a part of the SL3 compo-
nent. Noteworthy is that the binding energy shift between
the SL2 and Si4+ components is about 3 eV, which is sim-
ilar to the binding energy shift found between the silicon
bulk and Si4+ components for growth on Si(111)-(7 · 7)
[23]. Thus, the SL2 feature may have a contribution from
Si0 species.
A summary of how the intensities of the components
change as functions of film thickness is presented in
Fig. 6a–c. The Si layer value is defined as the sum of the
components attributed to the silicon layer (Si layer =
SL1 + SL2 + SL3), and relative intensity refers to the inten-
sity relative to the intensity of the SiC bulk component for
each spectrum.
160 s
10 Å
×
×
×
1.9
40 s
4 Å
1.2
0.33 eV
SiC
SiC(0001)-(3x3)
In Fig. 6a the relative intensities of the Si layer, SiOX
and Si4+ components are plotted as functions of estimated
film thickness. The spectra measured using 130 eV photons
have been used. The relative intensity for the Si layer is
1.0
106
104
102
100
98
Binding energy (eV)
˚
˚
approximately constant from 4 A to 51 A. The relative
intensity of the component attributed to Si4+ species in-
creases monotonically with increasing film thickness. In
contrast, the relative intensity of the SiOX component in-
Fig. 5b. Comparably less surface sensitive Si 2p PES spectra for the same
selected situations as in Fig. 6a. The spectra are measured using 350 eV
photons. The total energy resolution is 200 meV.
˚
creases up to an estimated film thickness of about 30 A,
to the fact that the oxidation with O2 was accomplished
using much larger doses (3000 L) as well as a higher tem-
perature. However, it should be noted that the alkoxide
precursors are strong oxidants when compared to O2. This
has been demonstrated for titanium tetra isopropoxide on
Si(111), where the tendency for Si oxidation far exceeds
that of O2 [26].
Another important aspect to take into account is the
possibility for reaction of the Si atoms with hydrocarbon
fragments, thereby interfering with the oxidation process.
The study of ZTB-mediated ZrO2 growth on Si(100) and
Si(111) reveals the presence of Si 2p states shifted by
À0.3 and +0.3 eV with respect to the bulk peak [23]. These
states were assigned to result from Si interacting with
hydrocarbons, carbon and hydrogen [23]. That states in
the BE regime associated with the SL1–SL3 states appear
due to reaction of Si adlayer atoms with hydrocarbon frag-
ments is thus also a plausible explanation.
and is thereafter approximately constant. The small de-
crease in relative intensity of the Si layer signal suggests
that the part of the Si adlayer atoms that are not oxidized
is substantial (at least 80%). However, it is clear that fur-
ther studies, e.g. by performing angular scans, could prove
to be most valuable in this respect.
The absolute intensities of the Si layer and Si4+ compo-
nents are plotted as functions of estimated film thickness in
Fig. 6b. Again, the spectra measured using 130 eV photons
have been used. The absolute intensity of the Si layer com-
ponent follows approximately an exponential attenuation
function. The absolute intensity of the Si4+ component in-
˚
creases up to a film thickness of about 10 A after which it is
approximately constant.
In Fig. 6c we present measures of the relative surface
sensitivity, calculated as the magnification of the relative
intensity when changing the photon energy from 350 eV
to 130 eV. The relative surface sensitivity of the bulk

2396
P.G. Karlsson et al. / Surface Science 601 (2007) 2390–2400
Table 1
Summary of the analysis of the Si 2p spectra, giving the label, binding
a
Si layer
Si3+
energy relative to the bulk SiC peak (BE, in eV), proposed assignment and
preferential location of the various Si species within the ZrO2/4H-
SiC(0001) heterojunction
2
1
0
Si4+
Label Relative Proposed assignment
BE
Location
SL1
SL2
À1.50
À1.05
Reconstructed Si adlayer (Si–H, Si)
Reconstructed Si adlayer (Si–H, Si,
Si–C)
Interface
Interface
SL3
SiC
À0.60
Reconstructed Si adlayer (Si–C, Si1+
)
Interface
Bulk
Interface
Near surface
0.00
Si2+, Si3+
0
10
20
30
40
50
SiOX +1.12
Si4+
+2.03
Thickness(Å)
b
species indicates that there are Si4+ species in the near sur-
face region, possibly intermixed with ZrO2 (forming sili-
cate). A similar behavior has previously been observed
for HfO2/SiO2/Si(001) [34] and Al2O3/Si [35]. It was sug-
gested in Ref. [34] that emission of Si species from the
SiO2/Si interface occurs in order to release the stress in-
duced by oxidation. It can not be excluded that this also
is the case for the ZrO2/SiC(0001) system.
Si layer
Si4+
In summary, the analysis of the Si 2p spectra reveals that
the reaction with the ZTB precursor is complex. This inter-
face consist of a multitude of different Si species in contrast
to that formed upon oxidation of Si(0001)-(3 · 3) by use
of O2. In Table 1, we summarize the discussion on the Si
2p data, providing proposed assignments and locations of
the various Si species.
0
10
20
30
40
50
Thickness(Å)
c
5
4
3
2
Si layer
Si3+
Si4+
3.1.3. Band alignment
Studies of oxide growth on SiC are to a large extent
motivated by the development of SiC electronic devices.
Consequently, a discussion about the alignment of the va-
lence band edges and the resulting band offsets are in place.
We have recently demonstrated that a picture of the energy
level arrangement can be obtained by combining PES and
XAS [28].
10
20
30
40
50
Thickness (Å
)
Fig. 6. Summary of results from the delineation of the Si 2p spectra
recorded for the film growth experiments. (a) Relative intensities of the Si
2p components plotted as functions of film thickness. (b) Absolute
intensities of the Si 2p components plotted as functions of film thickness.
(c) Photon energy dependent changes of the relative intensities of the Si 2p
components. These changes reflect the depth distribution of the different
species.
Selected valence band PES spectra are presented in
Fig. 7. The Si-rich SiC(0001)-(3 · 3) surface has a valence
band spectrum similar to earlier published results [7,8]. It is
observed that the feature between 1.0 and 1.5 eV, attrib-
uted to surface states for the reconstructed surface, has dis-
˚
appeared at 4 A film thickness. The feature at around
3.1 eV consists of contributions from both the silicon layer
and SiC bulk bands. It is attenuated upon film deposition,
consistent with the results found from the Si 2p spectra. A
shift towards higher BE is observed for this feature, analo-
gous to the behavior of the Si 2p and C 1s bulk peaks. In
the BE regime 5–10 eV new features appear, which grow
with increasing exposure. These features are mainly associ-
ated with the valence band of ZrO2 [28].
The film thickness dependent shifts of the VBEs can be
determined either from the shifts of the core levels or from
valence difference spectra [28,31–33]. Using the former
method, the VBE positions for SiC and ZrO2 have been
species is thus by definition 1, and higher numbers corre-
sponds to species closer to the surface. For the SiOX and
Si4+ traces the numbers corresponding to low thicknesses
are noisy due to uncertainties in determining the relatively
small areas of the corresponding components.
The results in Fig. 6 indicate that the SiOX species has a
similar depth position as the silicon layer and this position
is fixed relative to the SiC bulk. The most straightforward
interpretation is that the Si layer and SiOX species are sit-
uated at the interface between oxygen free and oxygen rich
materials. The high relative surface sensitivity of the Si4+

P.G. Karlsson et al. / Surface Science 601 (2007) 2390–2400
2397
51 Å/1700 s
Valence band
= 130 eV
h
ν
O 1s XAS
1700s / 51Å
800°C
1000°C
+1.05 eV
VBO
O 1s PES
40 s
4 Å
530
532
534
536
538
2
Energy (eV)
˚
Fig. 8. O 1s PES and XAS spectra for the 51 A thick ZrO2 film deposited
on the Si-rich SiC(0001)-(3 · 3) surface. The energy values for the PES
spectrum are relative to the Fermi level, whereas the energy values for the
XAS spectrum correspond to the absolute photon energy. The energy
needed to match the low energy side of the PES spectrum to the XAS
threshold gives an estimate of the location of the conduction band edge
relative to the Fermi level [28].
SiC(0001)-(3x3)
1
12
10
8
6
4
2
0
Binding energy (eV)
Fig. 7. Selected valence band spectra. Species 1 is attributed to surface
states origination from the (3 · 3) surface reconstruction. Species 2 is
related to the substrate and is attenuated upon deposition. The vertical
solid lines indicate the locations of the SiC and ZrO2 valence band edges
and the resulting valence band offset (VBO). After subsequent annealing
to 1000 ꢀC there are filled states all the way up to the Fermi level.
4 Å ZrO2
51 Å ZrO2
0.60 eV
4H-SiC(0001)
4H-SiC(0001)
0.40 eV
CBEs
estimated and indicated by vertical solid lines in Fig. 7. It is
observed that the valence band offset (VBO) increases with
˚
˚
˚
the film thickness, from 2.54 eV at 4 A to 2.72 A at 51 A.
The location of the conduction band edge CBE can be
estimated by adding band gap values from the literature
(3.26 eV for 4H-SiC [36] and 5.40 eV for ZrO2 [37]). Alter-
natively, the CBE position relative to the Fermi level can be
obtained by comparing the O 1s PES spectrum to the cor-
responding O 1s XAS spectrum [28]. In Fig. 8, the O 1s
2.54 eV
2.72 eV
VBEs
Fig. 9. Band alignment of the ZrO2/4H-SiC(0001) system formed by
ZrO2 growth on the Si-rich SiC(0001)-(3 · 3) system. The locations of the
valence band edges (VBEs) are determined from valence and core level
PES data. The location of the conduction band edges (CBEs) are
estimated from literature values of the band gaps, O 1s PES and O 1s XAS
spectra.
˚
PES and XAS spectra for the 51 A thick film are put on
a common energy scale. The values for the PES spectrum
represent the BE relative to the Fermi level while the values
for the XAS spectrum represent the absolute photon en-
ergy. By matching the low energy side of the PES spectrum
to the XAS threshold, an estimate of the CBE relative to
˚
the procedure adopted here, i.e. deposition on the Si-rich
the Fermi level is obtained. In the case of the 51 A thick
SiC(0001)-(3 · 3) surface, is not optimal in this respect.
ZrO2 film the CBE is located 1.05 eV above the Fermi level.
This gives a band gap of 5.40 eV when adding the VBE po-
sition found in the valence PES spectrum, in excellent
agreement with Ref. [37].
In Fig. 9 the band alignment for 4 A and 51 A thick
films are presented. With respect to the prospect of using
this system in MOSFET applications the band alignment
is too asymmetric. In order to be feasible the ZrO2 band
edges must be shifted downwards in BE by 1.5–2 eV rela-
tive to the SiC band edges [22,33]. It can be speculated that
3.2. Thermal stability
˚
˚
3.2.1. Annealing to 800 ꢀC
The estimated overlayer thickness decreases by a few
percent upon annealing the film grown at 400–800 ꢀC.
No observable changes appear in the Zr 3d and O 1s sig-
nals. However, the C1 species has disappeared in the corre-
sponding C 1s signal, as can be seen in Fig. 4a. That is, the

2398
P.G. Karlsson et al. / Surface Science 601 (2007) 2390–2400
surface methyl groups have vanished, most probably under
the formation of desorbing species.
Si 2p3/2
= 350 eV
The Si 2p3/2 lines measured using 130 eV photons and
350 eV photons are presented in Figs. 10a and 10b, respec-
tively. Only the Si 2p3/2 line is presented to better illustrate
the changes upon heating. The spectra are obtained after
subtraction of a calculated background followed by re-
moval of a calculated Si 2p1/2 line. After annealing to
800 ꢀC the sum of the relative intensities of the components
on the low binding energy side of the bulk component has
decreased. The components on the low binding energy side
of the bulk component are furthermore more separated in
BE. It is reasonable that the extension of states towards
lower BE:s signatures the formation a new type of silicon
species, possibly involving bonds to zirconium, cf. below.
It is observed that the relative intensities of the compo-
nents attributed to oxidized silicon species increase. The
component SiOX representing intermediate oxidation
states has both increased in relative intensity and shifted to-
wards the bulk peak. This is a possible indication of oxida-
tion of SiC. To clarify, formation of Si1+ is expected at a
binding energy shifted +0.6 eV from the bulk peak [8].
The position found for SiOX is +0.7 eV from the bulk
peak.
hν
1000 °C
800 °C
400 °C
106
104
102
100
98
Binding energy (eV)
Si 2p3/2
hν = 130 eV
Fig. 10b. Less surface sensitive measurements for the same selected
situations as in Fig. 9a. The presented Si 2p spectra are those obtained
after background subtraction and Si 2p1/2 stripping. The total energy
resolution is 200 meV.
Very small changes in the valence spectrum are ob-
served. The feature around 3.1 eV becomes slightly less
pronounced, consistent with the changes in the low BE Si
2p components associated with the silicon layer.
In summary, annealing to 800 ꢀC does not lead to any
appreciable changes in the ZrO2 film. However, changes
are observed that signals the beginning of interface
decomposition.
1000 °C
3.2.2. Annealing to 1000 ꢀC
800 °C
400 °C
Upon subsequent annealing to 1000 ꢀC the estimated
˚
average thickness is reduced to around 20 A, i.e. by more
than 50%. Fig. 3 shows that the Zr 3d spectrum is dramat-
ically different compared to that of the as grown film. A
new additional component at about 5 eV lower BE than
the Zr4+ component has appeared. From the BE position
and the asymmetric line shape [38] this component is attrib-
uted to zirconium atoms in a compound having metallic
properties. Noteworthy is that the relative area between
the two zirconium species does not change dramatically
when changing the surface sensitivity of the measurement,
as can be seen in Fig. 3. This indicates that the surface re-
gion is heterogeneous, i.e. the two Zr species stem from lat-
erally different parts.
106
104
102
100
98
Binding energy (eV)
Fig. 10a. Surface sensitive measurements for the thermal stability
experiments. The presented Si 2p spectra are those obtained after
background subtraction and Si 2p1/2 stripping. The total energy resolution
is 60 meV. Labels are defined and discussed in the text.

P.G. Karlsson et al. / Surface Science 601 (2007) 2390–2400
2399
The valence spectra change dramatically upon annealing
to 1000 ꢀC as seen in Fig. 8. The most apparent effect is the
decrease of the intensity of the ZrO2 valence band. How-
ever, the most important finding is that there are filled
states all the way up to the Fermi level. This supports the
notion of the formation of a metallic compound.
The absolute intensity of the O 1s signal (spectrum not
shown) decreases upon annealing to 1000 ꢀC but the
shape of the spectrum is retained. In line with the inter-
pretation of the Zr 3d spectrum the O 1s signal is mainly
attributed to remaining ZrO2. The O 1s XAS spectrum
(not shown) reveals that the tetragonal structure is
maintained.
The C 1s spectrum is shown in Fig. 4a. The signal in-
creases and a reasonable fit can be performed using two
components, each having the same width as found for
the bulk species for the Si-rich SiC(0001)-(3 · 3) surface.
This procedure results in a binding energy separation of
0.8 eV and an intensity ratio of 0.25 between the two com-
ponents. The two C 1s components denoted SiC and SiC*
in can both be assigned to SiC bulk species. This is consis-
tent with a laterally heterogeneous film as inferred from the
Zr 3d spectra. The C 1s high binding energy component
(denoted SiC) is attributed to areas where the ZrO2 over-
layer remains. This is motivated by the similarity in binding
energy position compared to that of the as grown film. The
C 1s low binding energy component (denoted SiC*) is
attributed to areas where there is a metallic compound
overlayer.
C 1s spectrum. The two components on the low binding
energy side are surface species. Based on the observation
of Zr 3d states in a metallic environment, the state of
99.1 eV BE is tentatively assigned to compounds of the
type ZrSix. The state labeled Si in Fig. 10a has a BE of
100.0 eV, which is very similar to that of the most abundant
Si species for the Si-rich SiC(0001)-(3 · 3) surface. This
suggests that it may be associated with Si aggregates on
the surface.
The results thus show that the ZrO2 film decomposes
upon annealing to 1000 ꢀC. A heterogeneous layer is
formed that can be subdivided laterally in regions with
remnants of the t-ZrO2 film, metallic Zr silicide and, possi-
bly, areas with a Si aggregates. The thermal stability under
vacuum is similar to that observed for ultrathin ZrO2 films
on Si(100) [41,42]. It has been argued that the ZrO2/
Si(100) structure is stable up to 800 ꢀC and that a reaction
leading to the formation of zirconium silicide occurs
around 900 ꢀC [41,42]. Regarding the stability at 800 ꢀC
it should be noted that the quality of the reported data
on annealed ZrO2/Si(100) makes an observation of the
subtler details found in the present study very difficult. It
was furthermore shown that formation of ZrSi2 is most
likely, as it is the most thermodynamically stable metal sil-
icide for the ZrO2/Si system [41,42]. However, since the
substrate in the present case is SiC there may be additional
possibilities. For instance, compounds containing carbon
atoms must be considered.
The BE for the C 1s peak labeled SiC* is 0.3 eV lower
than that of SiC bulk for the Si-rich SiC(0001)-(3 · 3) sur-
face. Thus, the core level shifts of the bulk component im-
posed by ZrO2 and metallic silicide are in opposite
directions. In a simple band-bending picture [39] this
behavior is possible if the two different overlaying materials
have significantly different work functions or laterally lo-
cally different interface states. Shifts in opposite directions
can also in principle be rationalized in terms of different
screening properties but the effect of screening is expected
to be more local in nature [40].
The Si 2p spectra change dramatically in shape upon
annealing to 1000 ꢀC and the absolute integrated intensity
increases. At least four components are needed to delineate
the spectra. The components presented in Figs. 10a and
10b are symmetric but most likely there are silicon species
in an environment having metallic properties. However,
introduction of reasonable asymmetries [38] in different
combinations have limited effects on the resulting relative
intensities and binding energy positions.
4. Conclusions
The growth and thermal stability of ultrathin ZrO2 films
on the Si-rich SiC(0001)-(3 · 3) surface have been ex-
plored using photoelectron spectroscopy (PES) and X-ray
absorption spectroscopy (XAS). The films were grown
in situ by chemical vapor deposition using the zirconium
tetra tert-butoxide (ZTB) precursor. During exposure to
ZTB the samples were heated to 400 ꢀC, which results in
growth of tetragonal ZrO2. An interface is formed between
the ZrO2 film and the SiC substrate. The interface contains
Si in various suboxidation states. Confined within the inter-
face are also Si species formed by reactions with hydrocar-
bon fragments of the ZTB molecule. Si in a 4+ oxidation
state is detected in the near surface region. This shows that
intermixing of SiO2 and ZrO2 occurs, possibly under the
formation of silicate. The SiC–ZrO2 band alignment was
discussed with the aid of valence PES and O 1s XAS spec-
tra. It is found that the as-prepared system does not fulfill
the band alignment required for a MOSFET device. Subse-
quent annealing of a deposited film was performed in order
to study the thermal stability of the system. Annealing to
800 ꢀC does not lead to decomposition of the tetragonal
ZrO2 (t-ZrO2) but changes are observed within the inter-
face region. After annealing to 1000 ꢀC a heterogeneous
layer has formed. The decomposition of the film leads to
regions with t-ZrO2 remnants, metallic Zr silicide and Si
aggregates.
By comparing the more surface sensitive measurements
with the less surface sensitive measurements it is possible
to distinguish between bulk and surface species. The two
components on the high binding energy side should be
attributed to SiC bulk. In the Si 2p spectrum measured at
350 eV photon energy the intensity ratio and binding
energy difference between the two components attributed
to SiC bulk are similar to those found for the corresponding

2400
P.G. Karlsson et al. / Surface Science 601 (2007) 2390–2400
[21] J.N. Andersen, O. Bjo¨rneholm, A. Sandell, R. Nyholm, J. Forsell, L.
Acknowledgements
˚
˚
Thanell, A. Nilsson, N. Martensson, Synchrotron Radiat. News 4
(1991) 15.
The authors acknowledge the assistance from the MAX-
lab staff and the financial support from the Swedish Science
Council (VR) and the Go¨ran Gustafsson Foundation.
[22] R. Putenkovikalam, E.A. Carter, J.P. Chang, Phys. Rev. B 69 (2004)
155329.
[23] P.G. Karlsson, J.H. Richter, J. Blomquist, P. Uvdal, T.M. Grehk, A.
Sandell, Surf. Sci. 601 (2007) 1008.
[24] P.W. Peacock, J. Robertson, Phys. Rev. Lett. 92 (2004) 057601.
[25] An approximately linear relationship between the core level binding
energy and oxidation state is commonly observed for semiconductors
and insulators [F.J. Himpsel, F.R. McFeely, A. Taleb-Ibrahimi, J.A.
Yarmoff, G. Holling, Phys. Rev. B 38 (1988) 6084]. The 3d core level
shift between Zr4+ and metallic Zr silicide is 4.9 eV. From this follows
that the core level shift between Zr4+ and Zr3+ should amount to
about 1.2 eV.
[26] A. Sandell, M.P. Andersson, M.K.-J. Johansson, P.G. Karlsson, Y.
Alfredsson, J. Schnadt, H. Siegbahn, P. Uvdal, Surf. Sci. 530 (2003)
63.
[27] K. Sakamoto, H.M. Zhang, R.I.G. Uhrberg, Phys. Rev. B 68 (2003)
075302.
References
[1] L.A. Lipkin, J.W. Palmour, J. Electron. Mater. 25 (1996) 909.
[2] V.R. Vathulya, D.N. Wang, M.H. White, Appl. Phys. Lett. 73 (1998)
2161;
V.V. Afanas’ev, A. Stesmans, M. Bassler, G. Pensl, M.J. Schulz, C.I.
Harris, Appl. Phys. Lett. 68 (1996) 2141.
[3] V.V. Afanas’ev, A. Stesmans, Appl. Phys. Lett. 69 (1996) 2252;
V.V. Afanas’ev, A. Stesmans, M. Bassler, G. Pensl, M.J. Schulz, C.I.
Harris, J. Appl. Phys. 85 (1999) 8292.
[4] F. Amy, H. Enriquez, P. Soukiassian, P.-F. Storino, Y.J. Chabal, A.J.
Mayne, G. Dujardin, Y.K. Hwu, C. Brylinski, Phys. Rev. Lett. 86
(2001) 4342.
[5] Y. Hoshino, T. Nishimura, T. Yonedo, K. Ogawa, H. Namba, Y.
Kido, Surf. Sci. 505 (2002) 234.
[28] A. Sandell, P.G. Karlsson, J.H. Richter, J. Blomquist, P. Uvdal, T.M.
Grehk, Appl. Phys. Lett. 88 (2006) 132905.
[29] A. Fink, W. Widdra, W. Wurth, C. Keller, M. Stichler, A. Achleitner,
G. Comelli, S. Lizzit, A. Baraldi, D. Menzel, Phys. Rev. B 64 (2001)
045308.
[6] F. Amy, P. Soukiassian, Y.K. Hwu, C. Brylinski, Phys. Rev. B 65
(2002) 165323.
[7] C. Virojanadara, L.I. Johansson, J. Phys. C 16 (2004) S1783.
[8] C. Virojanadara, L.I. Johansson, Phys. Rev. B 71 (2005) 195335.
[9] K.Y. Gao, Th. Seyller, L. Ley, F. Ciobanu, G. Pensi, A. Tadich, J.D.
Riley, R.G.C. Leckey, Appl. Phys. Lett. 83 (2003) 1830.
[10] D.R. Clarke, S.R. Phillpot, Mater. Today 8 (2005) 22.
[11] G.D. Wilk, R.M. Wallace, J.M. Anthony, J. Appl. Phys. 89 (2001)
5243.
[30] H. Liu, R. Hamers, Surf. Sci. 416 (1998) 354.
[31] E.A. Kraut, R.W. Grant, J.R. Waldrop, S.P. Kowalczyk, Phys. Rev.
Lett. 44 (1980) 1620.
[32] E.A. Kraut, R.W. Grant, J.R. Waldrop, S.P. Kowalczyk, Phys. Rev.
B 28 (1983) 1965.
[33] R. Puthenkovikalam, J.P. Chang, Appl. Phys. Lett. 84 (2004) 1353.
[34] K.Y. Gao, F. Speck, K. Emtsev, Th. Seyller, L. Ley, M. Oswald, W.
Hansch, Phys. Stat. Sol. A 203 (2006) 2194.
[35] Z. Ming, K. Nakajima, M. Suzuki, K. Kimura, M. Uematsu, K.
Torii, S. Kamiyama, Y. Nara, K. Yamada, Appl. Phys. Lett. 88
(2006) 153516.
[36] V.V. Afanas’ev, M. Bassler, G. Pensl, M.J. Schulz, E. Stein von
Kamienski, J. Appl. Phys. 101 (2007) 034108.
[37] V.V. Afanas’ev, F. Ciobanu, S. Dimitrijev, G. Pensl, A. Stesmans, J.
Phys.: Condens. Matter. 16 (2004) S1839.
[38] S. Doniach, M. Sunjic, J. Phys. C 3 (1970) 285.
[39] C. Kittel, Introduction to solid state physics, 1986. ISBN: 0-471-
87474-4;
[12] J.P. Chang, Y.-S. Lin, S. Berger, A. Kepten, R. Bloom, S. Levy, J.
Vac. Sci. Technol. B 19 (2001) 2137.
[13] B.-O. Cho, J.P. Chang, Appl. Phys. Lett. 93 (2003) 745.
[14] W.Y. Lee, H. Li, Y.-F. Su, Mater. Sci. Forum 426–432 (2003) 137.
[15] K. Shibata, T. Oi, A. Otsuka, H. Sumitomo, K. Oshihara, Y.
Teraoka, W. Ueda, J. Ceram. Soc. Jpn. 111 (2003) 852.
[16] F. Amy, P. Soukiassian, Y.K. Hwu, C. Brylinski, Appl. Phys. Lett. 75
(1999) 3360.
[17] F. Amy, P. Soukiassian, Y.K. Hwu, C. Brylinski, Surf. Sci. Lett. 464
(2000) 691.
[18] F.S. Tautz, S. Sloboshanin, U. Starke, J.A. Schaefer, Surf. Sci. 470
(2000) L25.
[19] K. Reuter, J. Bernhardt, H. Wedler, J. Schardt, U. Starke, K. Heinz,
Phys. Rev. Lett. 79 (1997) 4818.
[20] U. Starke, J. Schardt, J. Bernhardt, M. Franke, K. Reuter, H.
Wedler, K. Heinz, J. Furthmuller, P. Kackell, F. Bechstedt, Phys.
Rev. Lett. 80 (1999) 758.
S.M. Sze, Semiconductor devices, physics and technology, 2001.
ISBN: 0-471-33372-7.
[40] T.C. Chiang, G. Kaindl, T. Mandel, Phys. Rev. B 33 (1986) 695.
[41] T.S. Jeon, J.M. White, D.L. Kwong, Appl. Phys. Lett. 78 (2001) 368.
[42] J.P. Chang, Y.-S. Lin, Appl. Phys. Lett. 79 (2001) 3824.