Photoemission study of cobalt interaction with the oxidized Si(1 0 0)2 × 1 surface

Article abstract of DOI:10.1016/j.susc.2006.03.041

The interaction of cobalt atoms with an oxidized Si(1 0 0)2 × 1 surface was studied by photoelectron spectroscopy with synchrotron radiation at room and elevated temperatures. The SiO_x layer grown in situ on the crystal surface was ～0.3 nm thick, and the amount of deposited cobalt was varied within eight atomic layers. It was found that Co atoms could penetrate under the SiO_x layer even at room temperature in the initial growth. As the Co amount increased, a ternary Co-O-Si phase was formed at the interface, followed by a Co-Si solid solution. Silicide synthesis associated with the decomposition of these phases started under the SiO_x layer at ～250 °C, producing cobalt disilicide with a stable CaF₂-type of structure.

Full text of DOI:10.1016/j.susc.2006.03.041

Surface Science 600 (2006) 2449–2456
www.elsevier.com/locate/susc
Photoemission study of cobalt interaction with
the oxidized Si(100)2 · 1 surface
a
a,
a
a
M.V. Gomoyunova , I.I. Pronin *, D.E. Malygin , N.R. Gall ,
b
b
D.V. Vyalikh , S.L. Molodtsov
a
Ioffe Physico-Technical Institute of the Russian Academy of Sciences, Polytekhnicheskaya str. 26, 194021 St. Petersburg, Russia
b
Institut fuer Festkoerperphysik, 01062 Dresden, Germany
Received 10 October 2005; accepted for publication 31 March 2006
Available online 27 April 2006
Abstract
The interaction of cobalt atoms with an oxidized Si(100)2 · 1 surface was studied by photoelectron spectroscopy with synchrotron
radiation at room and elevated temperatures. The SiOx layer grown in situ on the crystal surface was ꢀ0.3 nm thick, and the amount of
deposited cobalt was varied within eight atomic layers. It was found that Co atoms could penetrate under the SiOx layer even at room
temperature in the initial growth. As the Co amount increased, a ternary Co–O–Si phase was formed at the interface, followed by a Co–
Si solid solution. Silicide synthesis associated with the decomposition of these phases started under the SiOx layer at ꢀ250 ꢀC, producing
cobalt disilicide with a stable CaF2-type of structure.
ꢁ 2006 Elsevier B.V. All rights reserved.
Keywords: Silicon; Cobalt; Oxides; Silicides; Solid-phase epitaxy; Synchrotron radiation photoelectron spectroscopy
1. Introduction
by the surface pre-oxidation [7–9]. However, the results
of these studies do not fit each other well enough, probably,
because of the application of different oxidation techniques
and different conditions for solid CoSi2 synthesis. In partic-
ular, the author of [7,8] oxidized the Si(100) surface ex situ,
while in [9] oxygen was adsorbed on an atomically pure Si
crystal in vacuum. The Co deposition and annealing were
carried out at different temperatures. Moreover, the work-
ers arrived at different conclusions concerning the mecha-
nism responsible for the improvement of properties of
CoSi2 films synthesized with oxidation. For example, this
improvement is interpreted in [7] as being due to fact that
the oxide layer acts as a diffusion barrier in the solid-phase
reaction, which limits the Co flux into the substrate. In the
later publication [8], the emphasis is placed on the layer
capping preventing the atomic re-arrangement in the sub-
strate during this reaction. A positive effect of the oxide
layer as a diffusion barrier is also mentioned in [9]. On
the other hand, the authors of this work state that the
deposited Co layer has a single crystal face-centered
The study of the Co/O/Si(100) system is of basic and
technological interest, especially, for the production of
low dimensional CoSi2/Si(100) structures widely used in
solid-state microelectronics and in the rapidly developing
area of nanoelectronics. The available information on the
initial growth of such structures on a Si(100)2 · 1 surface
by solid-phase epitaxy still remains ambiguous [1–4]. This
is because the interaction between Co and Si atoms is deter-
mined by poorly controllable factors, which makes it diffi-
cult to reproduce optimal conditions for the growth of
perfect CoSi2 films. The template technique suggested in
[5,6] is capable to overcome some of the difficulties but it
does not resolve the problem as a whole.
It has recently been shown that the solid-phase epitaxy
of CoSi2(100) on a Si(100)2 · 1 surface can be improved
*
Corresponding author. Tel.: +7 812 247 9948.
E-mail address: Igor.Pronin@mail.ioffe.ru (I.I. Pronin).
0039-6028/$ - see front matter ꢁ 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.susc.2006.03.041

2450
M.V. Gomoyunova et al. / Surface Science 600 (2006) 2449–2456
structure and that it is this factor, which facilitates the epi-
taxial growth of CoSi2(100).
3. Results and discussion
To try to resolve this controversy, we used photo-
electron spectroscopy of the core levels [10], which was
successfully applied to the Co/Si(100) system in earlier
experiments [1,11–17]. Even the preliminary results of this
study [18] led us to the conclusion that cobalt atoms depos-
ited onto in situ oxidized Si at room temperature were
capable to penetrate under the oxide layer. This conclu-
sion, however, was based only on a qualitative analysis
of the Si 2p spectra taken during the deposition. The pres-
ent study is an extension of our effort to examine the Co/O/
Si(100) system and includes a fitting of the Si 2p spectra
and a spectral analysis of valence and core level electrons
of cobalt. In addition to the interaction of Co atoms with
an oxidized Si surface at room temperature, we examined
the solid-phase reaction of CoSi2 formation during the
annealing of samples with deposited Co films.
3.1. Oxidation of the Si(100)2 · 1 surface
The initial stage of silicon oxidation has attracted the
attention of many researchers because of a wide applica-
tion of SiO2/Si structures in solid-state electronics
[20–25]. The typical Si 2p spectra for pure and oxidized
Si surfaces are shown in Fig. 1a and b. The oxidation
was performed at T = 500 ꢀC and exposure 10 L. These re-
sults were obtained from the raw data by subtraction of the
secondary-electron background fitted firstly by a polyno-
mial and secondly by the Shirley method. One can also
see the results of the standard least-squares fitting proce-
dure to the spectra, which was performed as described in
our earlier work [17]. The spectra were decomposed using
the Voigt functions which represent the convolution of
Lorentzian and Gaussian functions. The first one accounts
for the lifetime of a core-level hole, while the second one
2. Experimental
The experiments were carried out with the Russian–
German beamline at the Berliner Elektronenspeicherring
fur Synchrotronstrahlung (BESSY) at a photon energy
¨
(hm = 130 eV) providing a maximum surface sensitivity of
the Si 2p core levels. The photoelectron spectra were re-
corded in a wide solid angle oriented around the normal
to the surface of the sample at room temperature. The
overall energy resolution of the analyzer and the mono-
chromator was set to 130 meV FWHM. The substrates
were Si(100) P-doped wafers with the P concentration of
1017 at/cm3, which were well oriented with missorientation
less than 0.1ꢀ. The samples were ex situ prepared by a wet
chemical treatment using the method of Shiraki [19]. To
clean the surfaces in vacuum the samples were flashed at
1200 ꢀC and cooled slowly with the rate of less than
50 ꢀC/min to provide a well-defined (2 · 1) LEED pattern
and the surface uncontaminated with carbon and oxygen.
The oxidation of the sample surface was carried out
both at room temperature and at 500 ꢀC by exposing it
to an oxygen atmosphere at ꢀ10ꢁ7 mbar. The exposure was
varied from a few units to 104 L (1 L = 1 · 10ꢁ6 Torr s)
corresponding to a saturated surface coverage by oxygen.
The chemical composition of the surfaces was monitored
using the photoemission spectra. Cobalt was evaporated
from a thoroughly degassed source, in which a wire of
highly purified deposited material (99.99%) was heated by
electron bombardment. The Co flux was 1.8 ML/min in
the standard operation mode. The Co source was cali-
brated in two ways: by using a standard quartz microbal-
ance and by attenuating the W photoelectron signal of a
special tungsten sample pre-exposed to a Co flux. We esti-
mated the error of the Co coverage to be less than 20%.
The coverage of 6.8 · 1014 at/cm2 equal to the Si concen-
tration on the substrate was taken as a Co monolayer.
All the photoelectron spectra were measured at room tem-
perature in vacuum of 1.5 · 10ꢁ10 mbar.
Fig. 1. Si 2p core-level spectra with subtracted background taken at
hm = 130 eV from the sample with the clean (a) and oxidized (b)
Si(100)2 · 1 surface. The dots represent the experimental data, and solid
lines are the fitting results and fitted components.

M.V. Gomoyunova et al. / Surface Science 600 (2006) 2449–2456
2451
simulates the phonon broadening and the instrumental
resolution. The FWHM of the Lorentzian (0.07 eV), the
branching ratio of the 2p3/2 and 2p1/2 components (0.5)
and their splitting (0.61 eV) were taken to be constant.
The FWHM of the Gaussian, the relative binding energies
of the individual spectra components as well as their inten-
sities were varied during the fitting.
Si4+ mode has the lowest intensity. Generally, the intensity
ratio of the oxide modes depends on the oxidation condi-
tions, primarily on the crystal temperature and exposure
to oxygen. When the oxygen exposure and the oxidation
temperature are varied, the results do not change qualita-
tively, but there is a tendency for increasing amount of
atoms in Si3+ and Si4+ oxidation states as the exposure
and temperature increased. Of course, there is a certain ef-
fect of the initial state of the crystal surface, in particular,
the presence of various defects.
It is important to note the presence of two additional
modes (a and b) in the spectrum with negative and positive
shifts. No doubt, it would be impossible to get a good fit
between the experimental data and the simulation without
these modes. Similar modes were also found in some recent
studies [22,24,28–30]. The workers attribute them to Si
atoms contacting the oxide layer rather than the oxygen
atoms. Our data obtained from different oxidized Si sam-
ples show that the intensity of these modes is lowest, as
compared with the other modes, when the oxidation degree
is minimal. This observation is also an evidence that these
features are associated with the SiOx/Si interface.
The spectrum for the initial Si surface shown in Fig. 1a
fits the available data [26,27] and includes a bulk mode B
and five surface modes. The first two modes (Su and Sd)
are due to the upper and lower atoms in the dimers of
the reconstructed Si(100)2 · 1 surface. The energy shifts
of these lines relative to the bulk mode are ꢁ0.51 and
0.09 eV, respectively. The third mode S1 is due to the atoms
of the second layer with an energy shift of 0.18 eV. Finally,
the modes S2 and S02 for the respective types of atoms in the
third crystal layer have the ꢁ0.24 and 0.33 eV shifts.
The spectrum of oxidized Si differs essentially from that
of a pure Si substrate. The oxygen treatment of the sample
decreases the intensity of the B mode and leads to a nearly
complete disappearance of the surface modes: the step on
the right slope of the 2p3/2 line due to the Su mode disap-
pears, and the dip between the 2p1/2 and 2p3/2 sublevels
of the Si doublet becomes greater, coinciding with the posi-
tion of the S1 and S02 modes of pure Si. One can conclude
from these changes that the initial reconstruction of the
surface vanishes during its oxidation.
The major spectral changes are the presence of new fea-
tures with the positive energy shifts up to ꢀ4 eV. Some of
the features are seen as a fine structure. According to the
data of [21–25] Si oxidation gives rise to new modes due
to four oxidation states of Si atoms: Si1+ (Si2O), Si2+
(SiO), Si3+ (Si2O3), and Si4+ (SiO2). These states differ in
the number of oxygen atoms surrounding the Si atom, with
which they interact. The analysis of the oxidized Si spectra
registered in our experiments directly demonstrates all of
the four oxidized Si phases, indicating the growth of a SiOx
layer with a complex composition. The Si2+ mode is found
to have the highest intensity, and it can be seen directly in
the experimental spectrum; its energy shift is 1.78 eV. The
Table 1 presents the values we have found for the energy
shifts, half-widths and intensities of the maxima for all of
the modes. They show a good agreement with the respec-
tive values available in publications [21–25].
Note that the Su mode of the upper dimer atoms in the
Si 2p spectra of the Si(100)2 · 1 surface disappears even at
small O2 exposure (5 L at room temperature). This indi-
cates (within the accuracy of the method) that the oxide
layer covers the whole surface and is continuous. The layer
thickness can be estimated from the intensity attenuation
of the pure Si lines during the oxidation, assuming the
mean free path of Si 2p electrons to be 0.71 nm [21]. The
thickness value obtained in this way for the sample expo-
sure of 10 L is 0.25 nm; this estimation was made on the
assumption that the oxide thickness was uniform. The sili-
con oxide phases Si1+ and Si2+ seem to be localized directly
at the SiOx/Si interface, while the Si3+ and Si4+ phases are
formed on the sample surface, as observed in [28].
Table 1
Energy shifts, half-widths and relative intensities of different modes occurring in Si 2p core-level spectra of the oxidized silicon surface covered with ultra
thin cobalt layers at room temperature
Mode
Energy shift, eV
FWHM, eV
Intensity, arb. un.
Oxidized Si
1.5 ML Co
3 ML Co
4 ML Co
B
0
0.35
0.66
0.81
0.86
1.16
0.49
0.41
0.90
0.37
0.41
0.12
0.19
0.11
0.11
0.06
0.08
0.00
0.00
0.28
0.11
0.19
0.11
0.12
0.05
0.05
0.06
0.00
0.19
0.08
0.19
0.12
0.11
0.00
0.00
0.12
0.07
0.13
0.08
0.18
0.12
0.12
0.00
0.00
0.11
0.05
Si1+
Si2+
Si3+
Si4+
a
0.97
1.78
2.59
3.30
ꢁ0.31
0.35
0.58
ꢁ0.30
b
c
r

2452
M.V. Gomoyunova et al. / Surface Science 600 (2006) 2449–2456
3.2. Interaction of Co with the oxidized Si surface at
room temperature
An argument in favor of the involvement of oxygen atoms
in the growth of this phase is the absence of the respective
spectral mode for the Co–Si system we studied earlier
[16,17]. It is supported by the fact that the Si1+ intensity de-
creases since the Si atoms responsible for this mode are
mostly localized at the boundary with the non-oxidized
crystal region. Therefore, Co atoms deposited on oxidized
Si leave the crystal surface to penetrate under the oxide
layer. The fact that the deposition of the first 1.5 Co mon-
olayers does not totally attenuate the a and b modes shows
that the new phase formed at the interface occupies only
The spectra of the core Si 2p electrons registered during
the Co deposition on the oxidized Si crystal are shown in
Fig. 2. One can see that an increasing amount of deposited
metal induces considerable changes in the spectrum, which
are especially clear in the region of pure Si lines and can be
identified by the intensity drop. The greatest reduction in
the photoelectron emission occurs in the 2p3/2 region of
the Si peak, leading to a gradual decrease of its intensity
relative to that of the 2p1/2 peak. As for the spectral region,
containing oxide modes, the fine structure arising from oxi-
dation disappears with increasing Co coverage, and we ob-
serve only a slight reduction in the emission intensity.
We have suggested from the emission drop in the region
of bulk Si lines and from its slight change in the region of
oxide modes that most of the deposited cobalt leaves the
surface to go under the oxide layer [18]. Otherwise, all
the spectral modes would have to be attenuated in a similar
way, which contradicts the experimental observations.
To test this suggestion, we carried out the fitting of the
Si 2p spectra measured during the Co deposition. The re-
sults of the fitting are illustrated in Fig. 3. The quantitative
data obtained for the intensities of all spectral modes dur-
ing Co deposition are given in Table 1. One can see that 1.5
Co monolayer lowers the bulk mode by a factor of ꢀ1.5;
the a and b interface modes also decrease. Besides, a new
c mode appears with a positive energy shift of 0.39 eV.
The silicon oxide modes change but little. It is only the
Si1+ mode that shows some decrease. The attenuation of
the interface modes, the emergence of the new mode and
a good stability of the Si2+–Si4+ modes indicate that the
deposited Co atoms penetrate under the oxide layer into
the SiOx/Si interface, modifying it and producing a new c
phase there. The behavior of the latter is determined by
the three components of the system – Si, O and Co atoms.
Fig. 3. Decomposition of the Si 2p spectra taken from the oxidized
Si sample covered with different amounts of Co deposited at room
temperature.
Fig. 2. Si 2p core-level spectra of the oxidized Si(100)2 · 1 surface
covered with different amounts of Co deposited at room temperature.

M.V. Gomoyunova et al. / Surface Science 600 (2006) 2449–2456
2453
part of the crystal area and seems to grow in the island-type
way.
The increase in the coverage from 1.5 to 3 ML markedly
increases the c mode, while the b mode disappears. The lat-
ter fact suggests that the ternary interface is no longer dis-
crete but becomes continuous. In that case the a mode
should also disappear, but, instead, it rises nearly two
times. This apparent contradiction can be attributed to
the fact that the attenuating mode a is replaced by a new
r mode with practically the same energy shift. The shift va-
lue (ꢁ0.3 eV) is typical for Si atoms in Co–Si solid solution
produced during Co deposition on the Si(100)2 · 1 surface
at room temperature [11–13,17,31]. So we believe that a
similar solid solution was produced in our experiments.
This conclusion has become possible only from the results
of a spectral fitting, which reveals the growth of this Si
phase during Co deposition on an oxidized Si surface.
Further, Co deposition (up to four monolayers) contin-
ues to decrease the B mode of bulk Si and slightly attenu-
ates the c and r modes (Table 1). We attribute this to a
slower penetration of Co atoms under the oxide layer be-
cause of its increasing thickness; as a result, some of the
deposited cobalt remains on the sample surface. This sug-
gestion is supported by the results of an additional experi-
ment with 8 ML Co. Following 5 ML and more, the
intensity of the oxide phases also began to decrease but
not as much for the Si bulk mode. Therefore, only some
of the adatoms can penetrate into the sample at higher
metal coverages; the other adatoms remain on the surface,
thereby decreasing the oxidized and unoxidized mode
intensities.
Fig. 5. A schematic sequence of interface growth stages with increasing
Co-coverage.
ing through the oxide layer, the excited 3d metal electrons
make the major contribution to the emission of valence
electrons to produce the principal spectral maximum. As
the Co coverage becomes larger, the peak becomes nar-
rower and shifts slightly towards the Fermi level, where
the peak characteristic of pure Co is located.
Thus, the experimental data analysis provides the fol-
lowing physical picture of Co interaction with an oxidized
Si surface at room temperature (Fig. 5). The oxide layer,
which is formed on the sample and seems to be uniform,
has a thickness of about 0.25 nm. During the initial growth
(1.5–3 ML), the deposited Co atoms leave the surface to
penetrate into the oxide-silicon interface. This gives rise
to the ternary Co–Si–O phase responsible for the c mode
in the spectrum. At larger adsorbate amounts, Co atoms
accumulate at the interface, producing a solid solution
Co–Si. It is this phase that is responsible for the r mode.
When the coverage reaches 4 ML, not all Co atoms pene-
trate through the oxide layer, but some remain on the
surface.
Consider now the valent electron spectra presented in
Fig. 4. With increasing Co coverage, the photoelectron
emission rises abruptly in the region adjacent to the Fermi
level. It is well known that the photoionization cross-sec-
tions of the Co 3d levels at the photon energy of 130 eV
are larger than those of the Si 3p and 3s states. When pass-
3.3. Solid-phase reaction between Co and Si
The process of silicide growth in the Co/O/Si system at
elevated temperatures will be discussed with reference to
the data on the sample oxidized at room temperature.
The exposure to oxygen was 104 L and the SiOx layer thick-
ness evaluated from the attenuation of pure Si modes was
0.35 nm. We deposited 8 ML Co on the sample and an-
nealed it in a stepwise manner: the sample temperature
was raised by 100–150 ꢀC steps. The annealing duration
was 5 min. Then the sample was cooled to room tempera-
ture, and the spectra of the 2p Si and 3p Co levels, as well
as those of the valence band, were registered. The respec-
tive data are presented in Figs. 6–8. The results on the sam-
ples oxidized in the other regimes coincide qualitatively
with those to be discussed below, though there are some
differences in the details.
The changes indicating the onset of solid-phase reaction
between Co and Si started at ꢀ250 ꢀC for all of the spectra.
The major effect in the Si 2p spectrum (Fig. 6) is a higher
intensity of all the lines for silicon. The greatest increase
in the emission lies in the region with the Si 2p3/2 peak,
Fig. 4. Valence-band spectra of the oxidized Si(100)2 · 1 surface covered
with different amounts of Co deposited at room temperature.

2454
M.V. Gomoyunova et al. / Surface Science 600 (2006) 2449–2456
Fig. 8. Valence-band spectra taken after the annealing of the oxidized Si
sample covered with 8 ML of Co deposited at room temperature.
Fig. 6. Si 2p core-level spectra taken after the annealing of the oxidized Si
sample covered with 8 ML of Co at room temperature.
Fig. 10 that the bulk mode of pure Si decreases monoton-
ically with increasing temperature but a new mode D with a
positive energy shift of ꢀ0.2 eV emerges. Simultaneously,
we observe a reduction in the c mode. The intensity of
the D feature rises with temperature and becomes domi-
nant at 600 ꢀC in the spectral region with the pure Si
modes. Its energy shift is close to the values reported for
CoSi2 with the CaF2-type of structure [1,11,12,31]. There-
fore, the detection of this mode is an evidence for the for-
mation of cobalt disilicide with a stable structure. This
reaction seems to involve cobalt atoms in three initial
states. First, these are the Co atoms that remained on the
sample surface after the deposition and then diffused under
the oxide layer during the annealing. Second, some of the
Co atoms enter the Co–Si–O phase, which partly decom-
poses during the annealing, as indicated by the lower inten-
sity of the respective mode. A third source of Co atoms is
the Co–Si solid solution. Note that the seemingly slow
changes of the r mode can be explained by the new spectral
feature appearing at nearly the same energy. Indeed, cobalt
disilicide is known to have a surface mode with the ꢁ0.6 eV
shift relative to the bulk mode of this material [32]; this
mode is to be located just in the spectral region we are dis-
cussing. So, we believe that an additional phase similar to
the above one arises at the CoSi2 interface with an energy
shift of ꢁ0.5 eV counted from the CoSi2 bulk mode; it is
close to the shift of the CoSi2(100) surface mode reported
in the literature [32]. This fact offers the suggestion that the
solid-phase synthesis involves the formation of an ordered
CoSi2(100) layer.
Fig. 7. Co 3p spectra taken after the annealing of the oxidized Si sample
covered with 8 ML of Co deposited at room temperature.
showing that the changes occurred approximately in the re-
verse order as compared with those registered during the
deposition. One should note that the oxidation modes per-
sist and are even enhanced up to a temperature of 600 ꢀC.
Further, temperature rise up to 800 ꢀC abruptly decreases
these modes and they practically disappear at 850 ꢀC.
Simultaneously, there is a sharp intensity rise (by a factor
of ꢀ2) in the rest of the 2p Si spectrum.
The fitting of the experimental data has revealed the de-
tails of the mode behavior in the 2p Si spectrum during the
annealing, which can help us to understand the processes
occurring at high temperatures. The results of spectral fit-
ting for the initial sample and those annealed at 250 ꢀC,
400 ꢀC and 600 ꢀC are presented in Fig. 9. The intensity
dependences of the characteristic spectral modes on anneal-
ing temperature are given in Fig. 10. One can see from
One should note an essential increase of the Si3+ and
Si4+ modes during the sample annealing, which can be
partly attributed to the Co removal from the sample sur-
face, whose amount was ꢀ3 ML. When the Co atoms dif-
fuse under the oxide layer during the annealing, their
screening effect disappears, so the spectral intensity is to in-
crease for all of the oxide components. This factor cannot,

M.V. Gomoyunova et al. / Surface Science 600 (2006) 2449–2456
2455
Fig. 10. The dependences of the total intensities representative Si 2p
modes on annealing temperature.
however, account for the dominant rise of the Si3+ and Si4+
modes as compared with the Si2+ mode. This becomes clear
from a comparison of data on Si oxidation at various tem-
peratures. In particular, it was shown in [21,22] that the
intensity of oxide modes at 600 ꢀC rises with the degree
of Si oxidation. If Si is oxidized at room temperature or be-
low it, the oxide mode intensity, on the contrary, decreases
with Si oxidation [23–25]. The annealing of Si crystals, cov-
ered with SiOx layers grown at low temperature, leads to its
restructuring and as a result the intensity ratio of different
oxide modes becomes practically the same as in case of
high-temperature silicon oxidation. Such thermostimulated
restructuring seems to occur in our case.
However, the enhancement of Si3+ and Si4+ phases is
not as great as it might seem from Fig. 10. When Si atoms
are excited by 130 eV photons, the photo-ionization cross-
sections for Si4+ and Si3+ atoms are larger than those for
uni- and divalent atoms (by about a factor of 2.2 and
1.7, respectively) because of the resonance effects [21]. So,
the actual increase in the Si concentration in the Si4+ and
Si3+ oxide phases is about two times smaller than the rise
of the respective peaks in Fig. 10.
The results of the spectral analysis of the core Co 3p
electrons shown in Fig. 7 agree well with the above descrip-
tion of solid CoSi2 synthesis under the SiOx layer. As the
annealing temperature is raised, the Co mode intensity de-
creases monotonically. Moreover, there is a noticeable shift
of its 3p line maximum towards lower energies at 250 ꢀC,
indicating the onset of silicide growth under the oxide
layer. The further temperature rise to 600 ꢀC has only a
slight effect on the line shape, which becomes similar to
the typical CoSi2 line.
Let us now turn to the spectra of valence electrons shown
in Fig. 8. These support the above conclusion concerning
CoSi2 growth. The sample heating to 250 ꢀC leads to the
disappearance of the Co 3d peak located by about 1 eV be-
low the Fermi level and having parameters close to those of
Fig. 9. Decomposition of the Si 2p spectra taken from the annealed
samples.

2456
M.V. Gomoyunova et al. / Surface Science 600 (2006) 2449–2456
the pure Co peak. Instead, we observe a narrower feature at
the binding energy of 1.7 eV. Probably, this means the ab-
sence of metallic cobalt in the sample at 250 ꢀC, because it
is bound to silicon or oxygen. The temperature rise to
400 ꢀC gives a narrower peak because of a faster reduction
in the electron emission along its right high-energy slope. At
the temperature of 600 ꢀC, the line does not change its gen-
eral shape but it acquires two small features at E ꢀ 0.5 eV
and 1 eV, which are typical of CoSi2 [33]. This structure
supports the idea of solid CoSi2 synthesis during the crystal
annealing. The other well-defined peak at the binding en-
ergy of ꢀ7 eV is due to the oxide layer.
References
[1] J.M. Gallego, R. Miranda, S. Molodtsov, C. Laubschat, G. Kaindl,
Surf. Sci. 239 (1990) 203.
[2] H.L. Meyerheim, U. Do¨bler, A. Puschmann, Phys. Rev. B 44 (1991)
5738.
[3] U. Starke, W. Weiss, K. Heinz, G. Rangelov, Th. Fauster, G.R.
Castro, Surf. Sci. 352 (1996) 89.
[4] W.S. Cho, J.Y. Kim, N.G. Park, I.W. Lyo, K. Jeong, S.S. Kim, D.S.
Choi, S.N. Whang, K.H. Chae, Surf. Sci. 453 (2000) L309.
[5] S.M. Yalisove, R.T. Tung, D. Lorentto, J. Vac. Sci. Technol. A 7
(1989) 1472.
[6] D.D. Chambliss, T.N. Rhodin, J.E. Rowe, Phys. Rev. B 45 (1992)
1193.
[7] R.T. Tung, Appl. Phys. Lett. 68 (1996) 3461.
[8] R.T. Tung, Jpn. J. Appl. Phys. 36 (1997) 1650.
[9] Y. Hayashi, M. Yoshinaga, H. Ikeda, S. Zaima, Y. Yasuda, Surf. Sci.
438 (1999) 116.
[10] W.F. Egelhoff Jr., Surf. Sci. Rep. 6 (1987) 253.
[11] G. Rangelov, P. Augustin, J. Stober, Th. Fauster, Surf. Sci. 307
(1994) 264.
[12] G. Rangelov, P. Augustin, J. Stober, Th. Fauster, Phys. Rev. B 49
(1994) 7535.
Thus, we can conclude from the results presented that
silicide synthesis in the Co/O/Si system occurs under
the SiOx layer at the interface with the Si crystal. This
conclusion is consistent with the opinion of the author
of [8]. A thin oxide capping layer on the silicon surface lim-
its the region of the solid-phase reaction producing
CoSi2, and this prevents atomic re-arrangement on the Si
crystal surface. Such re-arrangements were, for instance,
observed in [34,35] during solid CoSi2 synthesis on a pure
Si(100)2 · 1 surface. The wafer surface was found to be
faceted, leading to a poorer morphology of the silicide film.
A pre-oxidation of the single crystal Si surface can remove
this negative effect.
[13] G. Rangelov, Th. Fauster, Surf. Sci. 365 (1996) 403.
[14] M.V. Gomoyunova, I.I. Pronin, N.R. Gall, S.L. Molodsov, D.V.
Vyalikh, Tech. Phys. Lett. 29 (2003) 496.
[15] M.V. Gomoyunova, I.I. Pronin, N.R. Gall, S.L. Molodtsov, D.V.
Vyalikh, Phys. Solid State 45 (2003) 1596.
[16] M.V. Gomoyunova, I.I. Pronin, Techn. Phys. 49 (10) (2004) 1249.
[17] M.V. Gomoyunova, I.I. Pronin, N.R. Gall, D.V. Vyalikh, S.L.
Molodtsov, Surf. Sci. 578 (2005) 174.
4. Summary
[18] M.V. Gomoyunova, I.I. Pronin, N.R. Gall, S.L. Molodtsov, D.V.
Vyalikh, Techn. Phys. Lett. 30 (2004) 850.
[19] A. Ishizaka, Y. Shiraki, J. Electrochim. Soc. 133 (1986) 666.
[20] S. Imata, A. Ishizaka, J. Appl. Phys. 79 (1996) 6653.
[21] F.J. Himpsel, F.R. McFeely, A. Taleb-Ibrahimi, J.A. Yarmoff, G.
Hollinger, Phys. Rev. B 38 (1988) 6084.
[22] Y. Hoshino, T. Nishimura, T. Nakada, H. Namba, Y. Kido, Surf.
Sci. 488 (2001) 249.
[23] P. Morgen, T. Jensen, C. Gundlach, L.-B. Tækker, S.V. Hoffman, K.
Pedersen, Comput. Mater. Sci. 21 (2001) 481.
[24] H.W. Yeom, R. Uhrberg, Jpn. J. Appl. Phys. 39 (2000) 4460.
[25] A. Mascaraque, C. Ottaviani, M. Capozi, M. Pedio, E.G. Michel,
Surf. Sci. 377–379 (1997) 650.
[26] T.-W. Pi, C.-P. Ouyang, J.-F. Wen, L.-C. Tien, J. Hwang, C.-P.
Cheng, G.K. Wertheim, Surf. Sci. 514 (2002) 327.
[27] H. Koh, J.W. Kim, W.H. Choi, H.W. Yeom, Phys. Rev. B 67 (2003)
073306.
[28] J.H. Oh, H.W. Yeom, Y. Hagimoto, K. Ono, M. Oshima, N.
Hirashita, M. Nywa, A. Toriuma, A. Kakizaki, Phys. Rev. B 63
(2001) 205310.
[29] T.-W. Pi, J.-F. Wen, C.-P. Ouyang, R.-T. Wu, G.K. Wertheim, Surf.
Sci. 478 (2001) L333.
[30] F. Jolly, F. Rochet, G. Dufour, C. Grupp, A. Taleb-Ibrahimi, J. Non-
Cryst. Solids 280 (2001) 150.
The analysis of photoelectron spectra has shown that
the deposition of the first cobalt monolayers on a pre-oxi-
dized Si(100)2 · 1 surface makes the metal atoms leave the
surface to penetrate under the oxide layer. This leads to the
formation of a new interface phase consisting of three types
of atoms (Si, Co and O), which is responsible for a new
mode c with a positive energy shift of 0.39 eV. This process
is followed by the growth of a Co–Si solid solution. When
the cobalt coverage increases to 4–5 ML, its removal be-
comes slower, so some of the Co atoms remain on the sur-
face to form a metallic film. The silicide synthesis involving
the decomposition of these new phases begins at a temper-
ature of ꢀ 250ꢀC. It occurs under the SiOx layer at the
layer-crystal interface, producing cobalt disilicide with a
stable CaF2-type of structure. The SiOx layer persists up
to a temperature of 800 ꢀC. It prevents the atomic re-
arrangement of the substrate surface and eventually im-
proves the quality of the CoSi2 film.
[31] F. Boscherini, J.J. Joyce, M.W. Ruckman, J.H. Weaver, Phys. Rev. B
35 (1987) 4216.
[32] R. Leckey, J.D. Riley, R.L. Johnson, L. Ley, B. Ditchek, J. Vac. Sci.
Technol. A 6 (1988) 63.
Acknowledgements
The study was supported by the Russian Foundation for
Basic Research (Project No. 04-02-17651) and by the bilat-
eral Project ‘‘Russian–German Laboratory at BESSY’’.
One of us (N.R. Gall) is grateful to the Home Science Sup-
port Foundation of Russian Federation.
[33] J. Derrien, Surf. Sci. 168 (1986) 171.
[34] J.R. Jimnez, L.J. Schowalter, L.M. Hsiung, K. Rajan, S. Heshimoto,
R.D. Thompson, S.S. Iger, Vac. Sci. Technol. A 8 (1990) 3014.
[35] C.W.T. Bulle-Lienwma, A.H. van Ommen, J. Hoonstra, C.N.A.M.
Aussems, J. Appl. Phys. 71 (1992) 2211.