Formation of interfacial phases in the epitaxial growth of …

Loading...

Taking too long?

Reload document

| Open in new tab

Pure Appl. Chem., Vol. 74, No. 9, pp. 1651–1661, 2002.
© 2002 IUPAC

Formation of interfacial phases in the epitaxial
growth of Sb on Si(111)-7 × 7 reconstructed
surface*

Vinod Kumar Paliwal1,2, A. G. Vedeshwar2, and S. M. Shivaprasad1,‡

1Surface Physics Group, National Physical Laboratory, New Delhi 110 012, India;
2Department of Physics and Astrophysics, University of Delhi, Delhi 110 007, India

Abstract: Understanding the evolution of the Sb/Si(111) interface is of great interest in the
formation of devices of nanodimensions. We have undertaken in situ (~10–11 torr) studies of
Sb adsorption (at room temperature) and its desorption on the 7 × 7 reconstructed Si(111)
surface, by complementary techniques such as X-ray photoelectron spectroscopy (XPS),
Auger electron spectroscopy (AES), low-energy electron diffraction (LEED), and electron
energy loss spectroscopy (EELS). For room-temperature (RT) Sb adsorption, the overlayer
grows in the Frank van der Merwe mode, forming an interface state of δ(7 × 7) in the sub-
monolayer Sb coverage regime. Adsorption of 1.0 monolayer (ML) Sb at RT shows an
abrupt shift of 0.8 eV in the peak position of the Sb 3d5/2 transition owing to band-bending
caused by a metallic (7 × 7) to a semiconducting (1 × 1) surface phase transformation.
Changes observed in full width at half-maximum (fwhm) and Sb 3d3/2 and 3d5/2 branching
ratio are discussed. Thermal annealing experiments provide evidence for agglomeration of
Sb islands, before the multilayer and monolayer desorption. During this desorption process,
we have observed two novel surface phases of (5 × 5) at 0.4 ML and (5√3 × 5√3 – R30°)
at 0.2 ML, stable at higher temperatures.

INTRODUCTION

Formation of compositional and doping superlattices of nanodimensions enable the tailoring of
advanced materials with novel properties, owing to the domination of quantum effects over the free-car-
rier distribution in this size regime [1,2]. Modern growth techniques such as molecular beam epitaxy
(MBE) have enabled the formation of superlattices of monolayer dimensions and, thus, the practical
realization of introducing artificial periodicity and, consequently, the band structure in a desired way,
an example of which is the formation of δ-doped silicon structures. Since this process involves the for-
mation of a dopant layer sandwiched between the silicon layers, the study of metal layer growth on sin-
gle-crystal silicon surfaces has assumed importance. Such studies also address the issue of band-bend-
ing and Schottky barrier formation that are dependent on surface and interface states. The properties of
metal/semiconductor interfaces have been a topic of great technological interest and scientific challenge
over several decades. Multidirectional approaches to understand the deviations from the Schottky–Mott
rule have yielded novel results, aiding the understanding at the atomistic level. It is now clear that apart
from just metal-induced gap states (MIGS), there are other factors involved at the real metal/semicon-
ductor contact, such as defect densities, growth kinetics, and interfacial strain. Since the kinetics of for-
mation plays a dominant role in determining the interface and overlayer characteristics, a surface sci-

*Pure Appl. Chem. 74, 1489–1783 (2002). An issue of reviews and research papers based on lectures presented at the 2nd IUPAC
Workshop on Advanced Materials (WAM II), Bangalore, India, 13–16 February 2002, on the theme of nanostructured advanced
materials.
‡Corresponding author: Fax: +91-11-5852678; E-mail: prasad@csnpl.ren.nic.in

1651

1652

V. K. PALIWAL et al.

ence approach to such studies is imperative. Thus, there are several studies in the literature that are
directed toward understanding the initial stages of formation of surface phases in a metal/semiconduc-
tor epitaxy, probed in situ during growth by surface-sensitive techniques such as AES, LEED, EELS,
XPS, ultraviolet photoelectron spectroscopy (UPS), scanning tunneling microscopy (STM), etc.

The study of the adsorption of Group V metal antimony on single-crystal silicon surfaces is being
intensely pursued to understand the mechanisms underlying several technologically important issues,
such as surfactant-mediated Ge/Si heteroepitaxy [3,4] and δ-doped structures [1,2]. The extra valence
electron of Sb adatoms on Si allows various ways to minimize the number of surface dangling bonds,
which makes it an interesting candidate to study the effect of the growth kinetics. Most of such studies
have concentrated on the adsorption of Sb on Si substrates held at high temperatures [5]. Among the Si
surfaces, the (111) surface has attracted enormous scientific interest owing to its complicated recon-
struction. The 7 × 7 reconstruction of the Si(111) substrate has been explained by the dimer-adatom-
stacking fault (DAS) model [6], which shows a quasi-continuous distribution of states within the bulk
band-gap of Si. The (7 × 7) unit cell consists of two triangular (1 × 1) subunits with faulted and
unfaulted stacking, respectively. These subunits are bound by dimer-row domain walls, which intersect
to produce the corner holes. This rest-adatom double layer is capped by Si adatoms. High-resolution
electron energy loss spectroscopy (HREELS) and angle-resolved ultraviolet photoelectron spectroscopy
(ARUPS) studies [7,8] have shown that some surface states lying close to the Fermi level disperse near
the zone boundary and cross the Fermi level, resulting in a metallic character of the Si(111)-7 × 7 recon-
struction. A large amount of work has already been reported in the literature especially for Sb adsorp-
tion on Si(111)-7 × 7 surface. The (√3 × √3 – R30°) structure obtained at a substrate temperature of
~650 °C, where only a monolayer that sticks on the surface [9–11] has been widely studied. The
√3 × √3 – R30° [written as (√3 × √3) below] at 1.0 monolayer (ML), d(2 × 1) at 0.8 ML, and
5√3 × 5√3–R30° [as (5√3 × 5√3) below] at 0.6 ML phases obtained at high temperatures have been
reported earlier by LEED studies [12]. However, with the advent of the high-resolution STM, the
(√3 × √3) at 0.33 ML was observed [13], and a detailed atomistic view of the structures, especially of
the (5√3 × 5√3) phase at 0.6 ML, emerged [14,15].

The growth kinetics of a metal layer on a reconstructed semiconductor surface has displayed the
potential to enable the tuning of interfacial electronic properties. Our previous studies on Ag/Si(111),
Ag/Si(001), and Mn/Si(111) [16–18] have prompted us to revisit the Sb/Si(111) interface formation by
parametric control. With this goal in view and to understand the role of kinetics and the substrate recon-
struction on growth modes and superstructural phases, we have undertaken here the studies of the
Sb/Si(111) interface formation at RT at low deposition rates and its thermal desorption in extremely
good ultrahigh vacuum (UHV) conditions. The RT adsorption studies result in a layer-by-layer growth
of Sb without disturbing the substrate reconstruction up to a critical coverage where a phase transition
that influences the electronic properties is observed. Thermal annealing studies provide evidence for
several epitaxial rearrangements as the Sb coverage is reduced.

EXPERIMENTAL
The experiments are performed in situ in two UHV chambers at a base pressure of 3 × 10–11 torr. In one
chamber with a hemi-spherical sector analyzer (HSA) Mg Kα (1253.6 eV) source and 0–5 kV electron
gun for XPS and AES, respectively, are housed, while the other chamber has a cylindrical mirror ana-
lyzer (CMA) enabling AES, EELS, and LEED (using 4-grid electron optics) studies. The analyzers are
a 0.18 % resolution CMA for AES and EELS, and a 25 meV resolution HSA for XPS, respectively.
AES is used as a common technique to ensure compatibility of measurements. The sample is a rectan-
gular piece (12 × 10 mm) cut from a Si(111) wafer (p-type of resistivity of 100 ohm-cm) which is
cleaned by a modified Shiraki process [19] to remove the hydrocarbons and form a thin SiO2 epilayer
before being introduced into the UHV chambers. The sample heating is accomplished by a combina-
tion of radiative, resistive, and electron-bombardment heating. The sample is flashed in vacuum at

© 2002 IUPAC, Pure and Applied Chemistry 74, 1651–1661

Formation of interfacial phases in the epitaxial growth of Sb

1653

1200 °C for several seconds and cooled gradually to RT to obtain a very clean (7 × 7) surface, as ascer-
tained by AES and LEED. The temperature is measured with an error of ±20 °C by a calibrated W-Re
thermocouple and an optical pyrometer. We report here studies using an Sb flux rate of 0.03 ML/min
from a home-made Ta-Knudsen cell, at different substrate temperatures (1.0 ML is defined as the den-
sity of a bulk truncated Si(111) surface, which is 7.85 × 1014 atoms/cm2). Thorough degassing and low
sublimation temperature of Sb ensures that the base pressure remains below 8 × 10–10 torr even during
Sb adsorption. Thermal stability studies are done by annealing the sample to the desired temperature
for 2 min and cooling it to RT before making AES, LEED, and EELS measurements.

RESULTS AND DISCUSSION

Room-temperature adsorption

It is known that thicker layers of Sb on Si, which can be of great technological relevance, can be formed
at RT [20,21]. Surprisingly, studies of the initial stages of the Sb/Si interface formation at RT have been
almost negligible, owing to the perception of lack of superstructural phases. After the pioneering work
of Metzger and Allen [22], who reported epitaxial (1 × 1) growth of Sb on Si(111), the other significant
work at RT Sb adsorption has been by Cuberes et al. [23], who, however, report disruption of the (7 × 7)
substrate structure even by depositing 0.25 ML of Sb, resulting in a rough morphology that promotes 3D
island formation. As Sb of constant flux is exposed to the clean Si(111)-7 × 7 surface, the uptake is mon-
itored by a series of Sb 3d core-level spectra. The changes observed map the interfacial growth and high-
light the mechanisms of the interface formation. However, in the Si(2p) spectra (not shown), the changes
are observed only in intensity but not in the peak positions and fwhm. This can be attributed to the rela-
tively large attenuation length of the Si(2p) XPS electrons in the Sb overlayer, and, thus, the intensity
information is coming predominantly from Si atoms that do not participate in the interface formation.

Figure 1 shows the uptake curve for Sb adsorption on the clean Si substrate, which shows a sharp
(7 × 7) LEED pattern. The probability of inelastic scattering for the characteristic electron shows an
exponential behavior with the overlayer thickness and is utilized to monitor the overlayer, owing to the
attenuation of electrons originating from the lower layers. This plot, of the intensity ratio of the adsor-
bate (Sb 3d) and substrate (Si 2p) XPS core level spectra as a function of time of adsorption, allows us
to follow the adsorption kinetics [24]. The graph shows a linear increase in the Sb/Si ratio up to about
30 min and then a change in the slope. We have identified the break point by the procedure adopted by
Stampanoni et al. [25], which yields the dotted curve shown in the figure. The curve shows the sum of

Fig. 1 Uptake curve of the Sb(3d5/2)/Si(2p) intensity ratio as a function of Sb deposition time (in minutes). Also
shown by the dotted curve is the sum of squares of errors (SSQ) to assist in determining the 1.0 ML break at 30
min.

© 2002 IUPAC, Pure and Applied Chemistry 74, 1651–1661

1654

V. K. PALIWAL et al.

squares of errors (SSQ) in the least-square fits of a set of two straight lines near the break point. The
minimum of the SSQ identifies the break point. The linear increase and the sharp points of change in
slope shows that Sb adsorption follows Frank van der Merwe (FM) (layer-by-layer) growth mode. FM
growth for Sb on Si has been reported earlier [22] with the first break in slope yielding a calibration for
1.0 ML. From the break point, it is inferred that we have a deposition rate of 0.03 ML/min.
Nontransition-metal adatoms generally tend to be very mobile and so are expected to form small metal-
lic 3D islands. However, in our case, it is clear that at the deposition rate adopted (0.03 ML/min) we
can form several epitaxial layers, at RT, determined by the kinetics of growth. This coverage is also con-
firmed by finding the Sb(MNN)/Si(LVV) Auger ratio that is obtained by adsorbing Sb onto a substrate
heated to 650 °C, which results in the well-known √3 × √3 LEED pattern [5,9] with 1.0 ML Sb.

Curves a and b of Fig. 2 are derived from the Sb 3d spectra, after suitable background subtraction
and deconvolution into Gaussian components. Curve 2a shows the positions of the Sb 3d5/2 peak as a
function of Sb coverage. The peak remains unchanged at 528.3 eV almost up to a coverage of 1.0 ML
and then suddenly shifts to 529.1 eV, showing a shift of 0.8 eV. After 1.0 ML, the peak position
decreases monotonically with the coverage up to 2.5 ML, and, therefore, the shift (which was 0.8 eV at
1.0 ML) also decreases by 0.4 eV to a value of 0.4 eV at 2.5 ML. Similar behavior is observed for the
Sb 3d3/2 transition. The shifts in the core-level peak positions in metal/semiconductor interfaces are
generally attributed either to formation of Schottky barrier owing to band-bending at the substrate sur-
face or to a chemical interaction between the adsorbate and the substrate. The chemical interaction is
highly unlikely at these temperatures for this system, as has been established by several earlier works
[26]. Further, the formation of a diffuse interface or a compound formation should have been reflected
in the uptake curve of Fig. 1. Usually, the Schottky barrier formation evolves continuously with metal
coverages up to a saturation value if the substrate surface is of semiconducting nature. However, it is
intriguing that in our case, the Sb 3d core level peaks do not undergo any shift in position in the sub-
monolayer regime, while at about 1.0 ML, there is an abrupt shift by about 0.8 eV. The observation of
the flat-band up to 1.0 ML can be attributed to the persistence of metallic nature of the (7 × 7) structure
of the substrate up to this coverage. It is well known that the (7 × 7) reconstruction leaves the Si(111)
surface with metallic nature [27]. Earlier studies on well-ordered (7 × 7) structures have shown a sharp
Fermi edge, and in HREELS, the broad and intense background observed is attributed to a continuum
of electronic transitions between continuously distributed surface states [7]. In ARUPS studies [8], at
least three different peaks are reported above and below the upper valence band edge, which do not
show significant dispersion in the (1 × 1) Brillouin zone. The surface state bound at about –0.25 eV
shows some dispersion near the zone boundary and crosses the Fermi level. Inverse photoemission

Fig. 2 (a) Shows changes in the Sb 3d5/2 peak position as a function of Sb coverage at RT. (b) Shows dependence
of the fwhm of Sb 3d5/2 peak on Sb coverage at RT.

© 2002 IUPAC, Pure and Applied Chemistry 74, 1651–1661

Formation of interfacial phases in the epitaxial growth of Sb

1655

experiments also indicate a continuous distribution of empty states in the upper half of the gap with two
maxima slightly above the conduction band edge [28]. The density and dispersion of the occupied and
unoccupied states may be responsible for the metallic character of the surface. Thus, it appears that at
the deposition rate chosen, Sb 2D clusters keep growing up to 1.0 ML, with the metallic nature of
(7 × 7) reconstruction of the Si substrate intact.

The sudden change in the peak position by 0.8 eV at 1.0 ML Sb coverage is interesting. Since we
have disregarded the formation of an interfacial compound, we attribute the shift to the formation of a
Schottky barrier. The Schottky–Mott relation gives the barrier height of about 0.4 eV for Sb/Si(111)-
7 × 7 interface. The presence of interface states shifts the charge neutrality level (ϕ
0) below the Fermi
energy (Ef). This means that the width of the depletion region and barrier height (ϕ
b) will be increased,
pulling ϕ
0 up toward Ef [29]. Therefore, the presence of interfacial states might have contributed to the
increased barrier height of 0.8 eV than that calculated by a simple Schottky–Mott relation. The forma-
tion of the barrier at exactly 1.0 ML suggests that the substrate surface ceases to have the metallic char-
acter at this coverage. This abruptness can be attributed to two plausible reasons, which may be mutu-
ally competing. Kahn and Stiles [30] have observed such a delayed onset of the barrier formation for
adsorption made at low temperatures. They attribute their results to the smallness of the cluster sizes at
submonolayer coverages, where the clusters do not form a metallic character. If we assume that this
mechanism plays a role in our studies because of the abrupt nature of the change, then it cannot explain
completely the present observations unless the substrate surface is semiconducting. The new surface
semiconducting phase and the corresponding interface states create a barrier of 0.8 eV. However, it is
interesting to note that for coverages greater than 1.0 ML, the shift of peak position (which was 0.8 eV
at 1.0 ML) decreases by 0.4 eV to a value of 0.4 eV at 2.5 ML. It is probable that the increased cover-
age of Sb quenches some of the surface states, thus shifting the neutrality level towards Ef so as to
approach the Schottky–Mott value of ϕ
b.

Figure 2b shows the change in fwhm of the Sb 3d5/2 as a function of Sb coverage. The fwhm value
also remains unchanged almost up to 1.0 ML. The fwhm measured to be ~1.6 eV suddenly assumes a
value of ~2.6 eV at the critical coverage of 1.0 ML, suggesting an increased width of the transition level
depending on the nature of the interface. Such a coverage dependence of the peak width is reported ear-
lier also in the literature [30,31]. This has been attributed to the formation of isolated metallic clusters
at lower coverages, which after a critical thickness form the features typical of metallic Sb owing to
s-d hybridization, which can also be applied to view our results. As discussed earlier, in the submono-
layer regime, we can understand the flat-band and the relatively narrow peak widths owing to the
premetallic dimensions of the Sb clusters. The sudden change in the Sb 3d peak width at 1.0 ML can
also suggest an abrupt coalescence of Sb 2D clusters to form the metallic overlayer. The similarity of
our results with the low-temperature studies reported above [30] can be due to the adsorption kinetics
that can result in rearrangements of atoms in the superstructure with different energetics.

Figure 3 shows an interesting coverage dependence of the anisotropy of the intensities of the Sb
core-level spin-orbit split peaks of 3d3/2 and 3d5/2 transitions. The ratio of the intensities, which is about
0.82 at a coverage of 0.1 ML, monotonically decreases with increasing Sb coverage and attains a satu-
ration value of 0.67 at the critical coverage of 1.0 ML and remains unchanged for higher coverages.
This again highlights the criticality of 1.0 ML and also the absence of any chemical interaction for com-
pound formation. It has been argued in the past, that such anisotropic changes observed in synchrotron
X-ray photoelectron spectroscopy (SXPS) are due to the rapid variations in the cross-sections around
Cooper minima, which change the radial matrix elements and phase-shift substantially, even for a small
difference in the final state energies of 3d5/2 and 3d3/2 [32]. In our experiments, we have used conven-
tional Mg Kα radiation with hν = 1253.6 eV, which suggests that in this energy regime, such small
kinetic energy changes have minimal cross-section differences so that such a process can be discounted.
X-ray photoelectron diffraction (XPD) studies have demonstrated this anisotropy as a manifestation of
the inherent difference in the photoelectron wave functions of the two spin-orbit components, ruling out
significant interference effects [33]. As the Sb adsorption proceeds in the submonolayer regime where

© 2002 IUPAC, Pure and Applied Chemistry 74, 1651–1661

1656

V. K. PALIWAL et al.

Fig. 3 Branching ratio of the intensities of the Sb 3d3/2 and 3d5/2 core-levels as a function of Sb coverage at RT.

Fig. 4 LEED patterns (at 53 eV) obtained for several Sb coverages adsorbed at RT onto Si(111)-7 × 7 surface.
(a) Clean Si(111)-7 × 7, (b) δ(7 × 7) at 0.2 ML, (c) δ(7 × 7) at 0.6 ML, and (d) 1 × 1 at 1.1 ML.

the underlying (7 × 7) structure of the substrate remains intact up to 1.0 ML, the monotonic decrease in
the anisotropy of the spin-orbit split peaks may be related to the increasing size of the 2D clusters.

To distinguish the effect of metal-to-semiconductor phase transition, and the lack of metallicity
in small 2D clusters, we have performed LEED studies. Figure 4 shows some representative LEED pat-
terns obtained at 53 eV, at different coverages of Sb adsorption onto (7 × 7) reconstructed Si(111) sur-
face held at RT. Figure 4a shows the typical LEED pattern for the clean Si(111) surface, which mani-
fests as the (7 × 7) surface reconstruction with six sharp 1/7th fractional order spots between the intense
hexagonal integral order spots. With the adsorption of submonolayer quantities of Sb, the changes that
occur in the LEED pattern are monitored. Figure 4b is the LEED pattern at 0.2 ML, which shows that

© 2002 IUPAC, Pure and Applied Chemistry 74, 1651–1661

Formation of interfacial phases in the epitaxial growth of Sb

1657

the fractional spots inside the (1 × 1) hexagon have vanished, while those along the adjacent integral
spots prevail. Similarly modified (7 × 7) LEED has also been reported earlier by Metzger and Allen [22].
With further increase in coverage, only the fractional order spots around the integral order spots persist
as in Fig. 4c for a coverage of 0.6 ML. Increasing further the coverage to values above 1.1 ML, the LEED
shows a sharp (1 × 1) structure, shown in Fig. 4d. Such modified (7 × 7) structures for H [34], Li [35],
and Ag adsorption [16] on Si(111) have been identified as δ(7 × 7) [5,16], and, thus, we attribute our
LEED also to this structure in the submonolayer coverage regime. To understand the evolution of this
phase, we have plotted dependence of the intensity ratio of the integral and fractional order LEED spots
in Fig. 5 on the Sb coverage. It is clear from the graph that the ratio increases up to a coverage of about
0.4 ML and decreases for coverages above 0.5 ML. At a coverage of about 1.0 ML, the intensity ratio
decreases to a very low value (1/10th the maximum value), where only a (1 × 1) LEED is observed. This
suggests that the adsorption of Sb does not disturb the underlying (7 × 7) substrate for submonolayer Sb
coverages at this low deposition rate. However, at 1.0 ML, the (7 × 7) changes to the (1 × 1) phase.

To look for changes in the electronic structure of this system by monitoring the single-electron
transitions and loss to collective electron oscillations, we have performed EELS experiments with an

Fig. 5 Intensity ratio of fractional to the integral order LEED spots, obtained for various Sb coverages, is shown
along with the unit cell symmetry observed.

Fig. 6 EELS spectra obtained with primary beam energy (Ep) of 250 eV, at various stages of Sb adsorption at RT.

© 2002 IUPAC, Pure and Applied Chemistry 74, 1651–1661

1658

V. K. PALIWAL et al.

incident primary beam energy (Ep) of 250 eV. The EELS results are shown in Fig. 6 for clean Si sur-
face and that adsorbed with various Sb coverages. The clean Si(111)-(7 × 7) surface results in the typ-
ical pattern as reported earlier [36,37]. The dominant peak at 17.5 eV is due to loss to bulk plasmon,
while the peak at 11.0 eV is attributed to the surface plasmon loss. The peaks at 14.5 eV and 8.0 eV are
attributed to surface states, and the 4.7 eV peak has been associated with interband transitions at 3.5 and
5.0 eV (Fig. 6a). Now, as the Sb coverage increases, the surface plasmon peak diminishes and only a
broad hump is observed in the 5 to 13 eV range as shown in curve Fig. 6b for a coverage of 0.4 ML.
As the coverage increases further and attains a very thick Sb overlayer, the bulk plasmon at 17.5 eV
monotonically decreases to a value of 16 eV, the Sb bulk plasmon value (Fig. 6f). However, the surface
plasmon peak begins to reform at about 0.5 ML, attains a maximum value of 11.7 eV at about 1.0 ML
(Fig. 6d) and then decreases monotonically toward the Sb surface plasmon loss value of 11.0 eV.
Comparing with the LEED observations, in the region where the δ(7 × 7) structure is formed, the inter-
face plasmon loss appears at 0.4 ML and then attains a maximum value when the (7 × 7) phase is still
persistent. After the surface phase becomes (1 × 1), then the surface plasmon loss value for Sb takes
over, attaining the bulk value of 11.0 eV for very thick Sb overlayer.

Consolidating the XPS, AES, EELS, and LEED results, it appears that the substrate surface-phase
change is the responsible factor for the changes that occur at a critical coverage of 1.0 ML. We postu-
late a plausible mechanism of the Sb-induced metal-to-semiconductor phase transition of the Si(111)
surface. As explained earlier, we recall the well-established DAS model of the (7 × 7) reconstruction.
At submonolayer coverages, Sb atoms initially occupy the (1 × 1) sites of the faulted and unfaulted tri-
angular subunits, between and above the Si atoms, keeping the (7 × 7) structure unaltered. As adsorp-
tion proceeds, close to 1.0 ML, the Sb adatoms occupy positions on the dimer rows to initiate the shift-
ing of charges from the substrate atoms to the adsorbates. This results in a (7 × 7) to (1 × 1) phase
transition due to the filling up of the corner holes with the Si adatoms, under a Sb monolayer. The
absence of Si adatoms in the (1 × 1) phase removes the –0.25 eV surface state and consequently ren-
ders the substrate semiconducting.

High-temperature annealing
As noted earlier, several surface phases such as the (√3 × √3) at 1.0 ML, d(2 × 1) at 0.8 ML,
(5√3 × 5√3) at 0.6 ML and a (√3 × √3) at 0.33 ML have been reported in earlier studies. To under-
stand the mechanism of phase changes and the thermal stability of the system, the residual composi-
tion during desorption of the RT deposited films is studied by AES as depicted in Fig. 7 for several

Fig. 7 Shows residual Sb/Si ratio after annealing the substrate for 2 min at each temperature, for various initial Sb
coverages adsorbed at RT. The temperature ranges shown (double-sided arrows) for multilayer and monolayer
desorption are taken from ref. [22].

© 2002 IUPAC, Pure and Applied Chemistry 74, 1651–1661