Adsorption of elemental S on Si(100)2×1: Surface restoration

Article abstract of DOI:10.1103/PhysRevB.55.4435

Adsorption of elemental S at room temperature causes the transition of the reconstructed Si(100)2×1 surface to its original bulk-terminated Si(IOO)1×1. The S adsorbate forms initially a (2×1) structure at 0.5 ML on the Si(100)2×1 surface, a (1×1) at 1 ML on the Si(100)1×1, and above 1 ML sulfur is imbedded into the Si substrate. The sticking coefficient of S is constant and equal to unity for the first 2 ML. Deposition of S at room temperature up to 1 ML increases the work function by 0.3±0.05 eV. The S adsorbate is strongly bound to the Si substrate in a molecular Si-S form. The Si-S bond energy is greater than that of Si-Si, which may be the driving force of the Si(100)2×1→Si(100)1×1 transition.

Full text of DOI:10.1103/PhysRevB.55.4435

PHYSICAL REVIEW B
VOLUME 55, NUMBER 7
15 FEBRUARY 1997-I
Adsorption of elemental S on Si„100…2؋1: Surface restoration
Aris Papageorgopoulos and Adero Corner
Department of Physics and Center for High Performance Polymers and Ceramics, Clark Atlanta University, Atlanta, Georgia 30314
M. Kamaratos and C. A. Papageorgopoulos
Department of Physics, University of Ioannina, P.O. Box 1186, GR-451 1 0 Ioannina, Greece
͑Received 26 January 1996͒
Adsorption of elemental S at room temperature causes the transition of the reconstructed Si͑100͒2ϫ1 surface
to its original bulk-terminated Si͑100͒1ϫ1. The S adsorbate forms initially a ͑2ϫ1͒ structure at 0.5 ML on the
Si͑100͒2ϫ1 surface, a ͑1ϫ1͒ at 1 ML on the Si͑100͒1ϫ1, and above 1 ML sulfur is imbedded into the Si
substrate. The sticking coefficient of S is constant and equal to unity for the first 2 ML. Deposition of S at room
temperature up to 1 ML increases the work function by 0.3Ϯ0.05 eV. The S adsorbate is strongly bound to the
Si substrate in a molecular Si-S form. The Si-S bond energy is greater than that of Si-Si, which may be the
driving force of the Si͑100͒2ϫ1→Si͑100͒1ϫ1 transition. ͓S0163-1829͑97͒01507-5͔
I. INTRODUCTION
structure and electronic properties of the growth of chalco-
genide protective films such as GaS and InS.
The passivation of semiconductor surfaces is an area of
intrinsic scientific interest and of great technological impor-
tance. Silicon and other semiconductors, such as GaAs and
InP, are well known for their potentially wide use in high-
speed electronics and long-wavelength optical circuits
Besides the importance in applications, there is a recent
rising scientific interest in the structural and electronic prop-
erties of chalcogen elements ͑S,Se͒ on Si͑100͒2ϫ1 surfaces.
Theoretical calculations suggest that adsorbates change the
structure of the Si͑100͒2ϫ1 surfaces.1
4,15
To our knowledge,
1
,2
͑optoelectronics͒.
They have also demonstrated great
this does not agree with the up-to-date relevant experimental
results. More specifically, the Si͑100͒ surface is easily recon-
structed with a small amount of heating. The surface has a
structure different from that of the bulk, and the reconstruc-
tion of the clean surface occurs in order to reduce the number
of broken dangling bonds.14 In other words, the reconstruc-
tion of the surface minimizes the high energy of broken co-
valent bonds, which would exist on an ideal bulk-terminated
plane.15 The clean Si͑100͒ surface shows a strong ͑2ϫ1͒
reconstruction in the low-energy electron diffraction ͑LEED͒
pattern, observed for the first time by Schlier and Farnsworth
value, mainly in space technology, as solar cells
3
͑
photovoltaics͒. Their efficiency, however, is reduced by
4
5
electron, x-ray, and ␥ ͑Ref. 6͒ radiation damage. Nonradi-
1
ative recombination of charge carriers is another such cause.
To prevent damage in the surfaces involved without reducing
their efficiency the above semiconductors are passivated.
This is done by depositing protective films ͑dielectric win-
dow layers͒, such as chalcogenides: sulfur, gallium sulfide,
and indium sulfide.7
–10
Most of the studies concerning sur-
face passivation of semiconductors were carried out with the
use of chemical vapor deposition techniques under atmo-
spheric pressure. The analysis of the deposited layers with
the former techniques has been obtained ex situ after the
completion of films a few hundred nm thick.
16
in 1959. Several models for the ͑2ϫ1͒ reconstruction have
been proposed.17 It has been recently accepted by most re-
searchers that dimers are the main building blocks of the
reconstructed surface of Si͑100͒. The question, however, of
whether the dimers are symmetric or buckled remains un-
clear, as reported by Chadi.18 Today, new evidence of asym-
metric ͑buckled͒ dimers is supported by most of the
The in situ analysis of the initial stages of the interface
formation in ultrahigh vacuum ͑UHV͒ is necessary for the
understanding and subsequent improvement of the deposi-
tion process. The interface region between the adsorbate and
the substrate is that which primarily controls the growth of
the films. The understanding of the film growth, therefore,
requires deposition methods that allow control of the build-
ing process of the structure at the atomic level.11 A begin-
ning with elemental S deposition on Si͑100͒2ϫ1 in UHV
would be appropriate. The knowledge of the behavior of S
alone on Si and other semiconductors is very important be-
cause ͑a͒ of the interest that has arisen with respect to the
possibility of pretreating the surfaces of the semiconductors
with S to protect and stabilize these surfaces against degra-
experimental1
9–23
and theoretical
14,15
investigators who have
14
worked on this problem. Kruger and Pollman calculated
that buckled dimers are energetically favored over symmetric
ones by 0.14 eV per dimer.14 The restoration of recon-
structed semiconductor surfaces to their original bulk-
terminated surface has been achieved, lately, by different ad-
sorbates. Specifically, ideal ͑1ϫ1͒ terminations of ͑111͒
surfaces were reported for As on Ge͑111͒ ͑Ref. 24͒ and
Si͑111͒ ͑Refs. 25 and 26͒ and for Cl on Ge͑111͒.27 Adsorp-
tion of S on clean Ge͑100͒2ϫ1 changed the ͑2ϫ1͒ structure
to ͑1ϫ1͒. The system S/Ge͑100͒1ϫ1 was regarded as an
12,13
28
28
dation resulting in improved subsequent processing,
and
ideally terminated surface. Weser et al. have experimen-
tally investigated the behavior of S on Si͑100͒. They have
͑b͒ it will help to obtain a better understanding of the binding
0
163-1829/97/55͑7͒/4435͑7͒/$10.00
55
4435
© 1997 The American Physical Society

4436
ARIS PAPAGEORGOPOULOS et al.
55
not observed an ordered S adlayer. Moriarty, Koenders, and
3
0
Hughes reported, recently, that room-temperature adsorp-
tion of sulfur resulted in the formation of an overlayer with
the underlying Si͑100͒ retaining the ͑2ϫ1͒ reconstruction.
They also mentioned that annealing of the S/Si͑100͒2ϫ1 sur-
face to 325 °C leads to the desorption of the sulfur overlayer
and the appearance of coexisting c͑4ϫ4͒ and ͑2ϫ1͒ surface
reconstruction. In contrast to this paper, theoretical studies
performed by Kaxiras,15 and later by Kruger and Pollman,
suggested that adsorption of group VI elements ͑S or Se͒ on
Si͑100͒2ϫ1 can lead to the restoration of the ideal bulk-
terminated geometry on the semiconductor surfaces. From
the above discussion it is apparently clear that additional
effort on the study of S and Se on Si͑100͒2ϫ1 is necessary.
Most of the sulfur adsorption studies up to now have
14
taken place with the exposure of the substrates to H S gas.
2
FIG. 1. Auger peak-to-peak heights ͑AppH͒ of the S͑151 eV͒
and the work function ͑⌬⌽͒ change as a function of the number of
doses of S deposited on clean Si͑100͒2ϫ1 surfaces at room tem-
perature.
To remove the H from the surface the substrates were
2
heated to temperatures equal to or greater than 200 °C. Hy-
drogen, however, cannot be removed selectively. These
mixed systems do not show any well-defined long-range
2
8
order. For a detailed understanding of the adsorption kinet-
ics of S on Si surfaces at room and lower temperatures, it is
important to deposit elemental sulfur.
tion of the number of doses of S deposited on clean
Si͑100͒2ϫ1 surfaces at room temperature. These measure-
ments are shown in correlation with the observed LEED pat-
terns, which will be discussed later in the next section. As
seen in this figure, the Auger peak-to-peak height of S͑151
eV͒ initially increases linearly with an increasing number of
S doses. Near the ninth dose, the curve forms a break ͑slope
change of the S Auger peak height versus S dose curve͒.
Also above the 9th dose the S Auger peak height is increas-
ing linearly up to the 18th dose of our measurements where
it forms a second break. Above the 18th dose the S͑151 eV͒
peak height starts to level off. This Auger peak height versus
S doses curve is characteristic of a layer-by-layer growth.
During the S deposition on the clean Si͑100͒2ϫ1 surface at
room temperature, the work function initially increases lin-
early with increasing S coverage. Above the fourth S dose
the work-function curve deviates from linearity and its slope
becomes smaller. Near the ninth dose, the work function
reaches its maximum value, and subsequently starts to de-
crease as the number of sulfur doses increases. Above the
thirteenth dose the decrease of the work function is very
small, despite that the Auger peak height of S͑151 eV͒ con-
tinues to increase. The maximum observed value of the WF
increase was about 0.3 eV. The above WF measurements
were repeated three times, and the maximum WF value var-
ied by Ϯ0.05 eV. Figure 2 shows the energy shift of the
integrated Auger Si͑92 eV͒ peak during S deposition on
clean Si͑100͒2ϫ1. The inset shows the Si͑92 eV͒ energy
shift versus S doses. According to the inset, this shift in-
creases linearly with increasing S doses to its final value of
In this work we evaporate elemental sulfur on
Si͑100͒2ϫ1 surfaces. We study the sample using LEED, Au-
ger Electron Spectroscopy ͑AES͒, thermal-desorption spec-
troscopy ͑TDS͒, and work function ͑WF͒ measurements. The
data suggest that the presence of sulfur on the surface causes
a phase transition of the substrate. Preliminary results have
been reported elsewhere.
II. EXPERIMENT
The experiments were performed in an ultrahigh vacuum
Ϫ10
chamber ͑pϽ10
Torr͒, equipped with a cylindrical mirror
analyzer ͑CMA͒ for AES measurements, a quadrupole mass
spectrometer ͑QMS͒ for TDS measurements, a LEED sys-
tem, and a Kelvin probe for WF measurements.
Elemental sulfur was evaporated by thermal dissociation
of MoS single crystal flakes mounted on a tungsten fila-
2
ment. During dissociation of MoS the Mo remained on the
2
tungsten filament, while S was evaporated. The Si͑100͒ sub-
ϩ
strate was cleaned by Ar bombardment at Eϭ1 keV for 40
min with an ion current of 10 ␮A. After bombardment the
sample was heated to 1000 °C by passing current through a
0.05-mm Ta strip, uniformly pressed between the sample and
a Ta foil case. The temperature of the sample was measured
by a Cr-Al thermocouple. The Si specimen was considered
sufficiently clean when the Auger peak height ratios C͑272
eV͒/Si͑92 eV͒ and O͑512 eV͒/Si͑92 eV͒ were below 1%.
The estimation of the S coverages on Si͑100͒ surfaces were
based on a correlation of LEED, AES, and TDS measure-
ments, and the comparison with the measurements of S on
the Ni͑100͒ surface, which took place in the same system
under the same deposition conditions.31 The surface atomic
density of 1 ML of S on Si͑100͒ is considered equal to that
1
.5 eV at the completion of the second layer of S. During
heating and as the S is removed from the surface the shift
decreases and the energy of the Si peak goes back to that of
the clean Si surface. This is an Auger chemical shift, which
may be attributed to a strong S-Si interaction. Most likely,
the S adatoms form a compound with the Si substrate. The
fact that the shift increases linearly may indicate that the
nature of binding is the same up to the completion of the
second layer, in agreement with the following TDS measure-
ments.
14
Ϫ2
of the outermost layer of Si, N ϭ6.8ϫ10 atoms cm
.
Si
III. RESULTS
A. Auger and work-function measurements
Figure 1 shows the Auger peak-to-peak heights ͑AppH͒ of
the S͑151 eV͒ and the work function ͑⌬⌽͒ change as a func-

55
ADSORPTION OF ELEMENTAL S ON Si͑100͒2ϫ1: . . .
4437
FIG. 4. Areas under the thermal desorption peaks of SiS ͑Fig. 3͒
vs S doses on Si͑100͒ surfaces.
spectra. There is only a single peak with its maximum value
near 585 °C, the very small peaks of S , not shown here,
2
appear at the same temperature and probably are due to a
partial dissociation of the SiS molecule. This finding does
not agree with Moriarty, Koenders, and Hughes,30 who re-
ported that the annealing of the S-covered Si͑100͒2ϫ1 sur-
face to 325 °C leads to the desorption of the sulfur overlayer.
The latter disagreement will be also discussed in correlation
with other findings in the discussion. The fact that the TD
peaks of SiS remain relatively sharp at the same temperature
with increasing S coverage indicates that the nature of Si-S
binding remains the same up to the completion of the S
coverage corresponding to the completion of 2 ML, in agree-
ment with the chemical shift measurements.
Figure 4 shows the areas under the TD peaks of SiS ͑Fig.
3͒ versus the S doses on Si͑100͒ surfaces. Note that the S
doses in the TDS were 15% greater than those in the previ-
ous ͑AES and WF͒ measurements, and the completion of 2
ML of S in Fig. 4 occurs at the 15th dose. It is obvious, from
this figure, that the TDS areas increase linearly with increas-
ing number of S doses, with a break ͑slope change of the SiS
TDS areas versus S doses curve͒ occurring near the 15th
dose of S; this is where we believe that the completion of 2
ML of sulfur takes place. It is known that the slope of the
areas under the TDS versus doses of a deposited adsorbate
on a substrate is proportional to the sticking coefficient of the
adsorbate. This, in correlation with Fig. 4, implies that the
sticking coefficient of S on Si͑100͒ surfaces remains constant
up to 2 ML and subsequently becomes substantially smaller.
FIG. 2. Energy shift of the integrated Auger Si͑92 eV͒ peak
during S deposition on clean Si͑100͒2ϫ1. The inset shows the Si
chemical shift vs S doses.
B. TDS measurements
Thermal-desorption spectra from S-covered Si͑100͒ sur-
faces have shown hardly observable peaks of S ͑amu 64͒.
2
Sulfur was mainly desorbed as a SiS compound. Figure 3
shows a series of SiS ͑amu 60͒ of thermal-desorption spectra
for different amounts of S deposited on Si͑100͒ surfaces. The
heating rate of desorption was constant, ␤ϭ30 K/s, for all
C. LEED measurements
The LEED observations show that the clean Si͑100͒2ϫ1
surface gives a good intense ͑2ϫ1͒ LEED pattern. Above the
fourth dose of S deposition on this surface, the half-order
spots become diffused and the pattern changes to ͑1ϫ1͒ with
its maximum intensity near the ninth dose. Further S depo-
sition increases the background, however, the ͑1ϫ1͒ struc-
ture remains. It appears that above the 4th dose of S the
Si͑100͒2ϫ1 reconstructed surface begins to change to
Si͑100͒1ϫ1. We believe that S adsorption on Si͑100͒2ϫ1
FIG. 3. Thermal-desorption spectra of SiS from S-covered
Si͑100͒ surfaces.

4438
ARIS PAPAGEORGOPOULOS et al.
55
forms also a ͑2ϫ1͒ structure, initially. The ͑2ϫ1͒ pattern,
near the 4th dose, should correspond to 0.5 ML of S cover-
age. Above the 4th dose, both the Si substrate and the S
adsorbate change gradually to a ͑1ϫ1͒ structure. The
completion of the ͑1ϫ1͒, near the 9th dose, corresponds to
1.0 ML of S, and coincides with the first break of the Auger
curve and the maximum increase of the work-function value
͑Fig. 1͒. Above the 9th S dose, the ͑1ϫ1͒ is retained.
The same UHV system has been used previously with the
same S source, and about the same flux of S on Ni͑100͒.31
The LEED observations showed the formation of a c͑2ϫ2͒
pattern close to the 12th dose of S on Ni͑100͒. The density of
the S overlayer, which produces the c͑2ϫ2͒ on Ni͑100͒, is
1
4
Ϫ2
8
ϫ10 atoms cm at 12 doses. At 9 doses this density
14
Ϫ2
should be about 6ϫ10 atoms cm . This is very close to
.8ϫ10 atoms cm , which is the density of 1 ML of S on
1
4
Ϫ2
6
Si͑100͒. Therefore, at the completion of the ͑1ϫ1͒ structure
on Si͑100͒, the coverage of S at 9 doses is indeed about 1
ML. It has been proposed that the initial sticking coefficient
of S on clean Ni͑100͒ is close to unity.32 This implies that
the initial sticking coefficient of S on Si͑100͒2ϫ1 is also
one. The linearity of the Auger curve up to 1 ML ͑Fig. 1͒
indicates that the sticking coefficient of S on Si͑100͒ surface
remains one, at least up to 1 ML. The slope of the Auger
curve is also constant for the second monolayer, between the
first and the second break, which occurs in about the same
time of completion as that of the first monolayer. This may
further imply that the sticking coefficient of S is nearly con-
stant during the adsorption of the second S layer, between
the 9th and 18th doses. Moreover, the TDS measurements
FIG. 5. Variation of the Auger peak-to-peak height ͑AppH͒ of
S͑151 eV͒ and the work-function change ͑⌬⌽͒ during the heating of
the S-covered Si͑100͒2ϫ1 surfaces.
ML means that only part of the second layer is initially dif-
fused into the bulk of Si at room temperature.
D. Heating of S-covered Si„100…2؋1 surfaces
͑Fig. 4͒ indicate clearly that the sticking coefficient remains
Figure 5 shows the variation of the Auger peak-to-peak
height of S͑151 eV͒ and the work-function change ͑⌬⌽͒, in
correlation with the LEED patterns, during the heating of the
S-covered Si͑100͒2ϫ1 substrate. The heating was implied in
constant and, therefore, equal to unity up to the completion
of 2 ML of S coverage.
Despite that the Auger curve remains linear and the stick-
ing coefficient remains constant up to 9th S dose, the work-
function curve deviates from linearity at the 4th dose. This
may be attributed to different sites of S atoms on Si before
and after the 4th dose. Both the chemical shift ͑Fig. 2͒ and
the TDS ͑Fig. 4͒ measurements indicate that the nature of
binding of S atoms on Si remains the same up to 2 ML of S,
which may imply that the charge transfer ͑polarization͒ for
each S adatom remains also the same. Therefore, any change
to the surface dipole moment should be attributed to the
dipole length change. Consequently, the sulfur atoms resid-
ing on the dimers of the Si͑100͒2ϫ1 surface may have
greater dipole moment ͑dipole length͒ than on sites between
Si atoms of the Si͑1ϫ1͒ surface. In the latter case, the dis-
tance between the neighboring Si atoms of the top layer is
greater than that between those of the dimers of the recon-
structed surface. Therefore, the S adatoms should be deeper
in their sites between the Si atoms of the ͑1ϫ1͒ surface
structure, with a smaller dipole length, therefore, smaller di-
pole moment and consequently a smaller work function than
that for S atoms on the dimers. This will be explained in
more detail in the Discussion. The fact that, above 1.0 ML
5
0 °C increments for 2 min each. We should emphasize that
the work function decreases in the early stages of the heating
treatment and reaches its minimum value at 300 °C while the
Auger peak height remains nearly unchanged up to 400 °C,
i.e., at which the S is not yet removed from the surface. Most
likely, heating provides the activation energy for diffusion of
more S atoms into the bulk of the Si substrate, causing a
further work-function decrease. Above 400 °C, the work
function increases while the Auger peak of S decreases dras-
tically, indicating a drastic desorption of S from the surface.
At about 550 °C, the second S layer is removed completely
and the work function increases to its maximum value of 1
ML. Above 550 °C the WF decreases again and as the S
coverage approaches 0.5 ML, the surface changes back to the
reconstructed ͑2ϫ1͒. Near 650 °C the S is completely des-
orbed from the Si substrate.
IV. DISCUSSION
As we have mentioned previously, the current experimen-
tal LEED observations show very clearly that deposition of
elemental sulfur on clean Si͑100͒2ϫ1 surfaces, at room tem-
perature, changes the Si͑100͒2ϫ1 reconstructed surface to
Si͑100͒1ϫ1. We believe that the S adatoms initially reside
on the dimers forming a ͑2ϫ1͒ at 0.5 ML and a ͑1ϫ1͒ above
͑9th dose͒, the work-function value decreases, while the
sticking coefficient and the nature of binding of S on Si
remain constant, may indicate that above 1.0 ML the sulfur
is submerged into the Si bulk near the surface. Since the WF
lowering almost stops before the S coverage completion of 2

55
ADSORPTION OF ELEMENTAL S ON Si͑100͒2ϫ1: . . .
4439
FIG. 6. Side-view schematics of ͑a͒ the top three layers of a
clean reconstructed Si͑100͒2ϫ1 surface with the dangling bonds,
͑
͑
b͒ the S-͑2ϫ1͒ ͑hemisulfide͒ structure on the Si͑100͒2ϫ1 surface,
c͒ the S-͑1ϫ1͒ ͑monosulfide͒ structure on a Si͑100͒1ϫ1 surface,
and ͑d͒ the diffused second S layer into the bulk of Si͑100͒1ϫ1.
FIG. 7. Top-view schematics of ͑a͒ the top three layers of the
clean reconstructed Si͑100͒2ϫ1 surface, ͑b͒ the S-͑2ϫ1͒ structure
on the Si͑100͒2ϫ1 surface, ͑c͒ the S-͑1ϫ1͒ structure on
Si͑100͒1ϫ1 surface, and ͑d͒ the diffused second S layer into the
bulk of Si͑100͒1ϫ1.
this coverage. In Fig. 6 we propose a surface structural
model of sulfur on Si͑100͒ surfaces. Figure 6͑a͒ shows a
schematic side view of the top three layers of the clean re-
constructed Si͑100͒2ϫ1 surface with the dangling bonds.
The deposited S atoms originally reside on bridge sites. Each
S atom is specifically bound through the dangling bonds on
the two Si atoms of the dimer ͓Fig. 6͑b͔͒, which is consistent
with the fact that S is divalent. The maximum coverage of
this state is 0.5 ML at which all the dimer sites are filled, and
the S adatoms form a ͑2ϫ1͒ structure retaining the recon-
struction of the Si substrate. This ͑2ϫ1͒ structure of the S
overlayer on the Si͑100͒2ϫ1 surface has been named the
hemisulfide state. As the S deposition on the Si substrate
continues above 0.5 ML, the S atoms that reside on available
bridge sites between dimers cause the bonds between the Si
atoms of the dimers to break, thus providing bonding for the
S adatoms. As a result of this process, the Si atoms are
gradually displaced, causing the restoration of the recon-
structed Si͑100͒2ϫ1 to a Si͑100͒1ϫ1 surface. The sulfur
adatoms remain on the bridge sites, and are bound to neigh-
boring Si atoms as shown in Fig. 6͑c͒. Thus, above 0.5 ML,
S gradually forms a S͑1ϫ1͒ structure, causing the change of
the Si substrate to an ideal bulk-terminated plane. This
S͑1ϫ1͒ structure on Si͑100͒1ϫ1 has been named the mono-
sulfide state with a maximum coverage of 1 ML. Above 1
ML, S continues to be adsorbed with the same sticking co-
efficient up to the coverage of 2 ML. Most of the second
monolayer of S is initially diffused into the bulk of Si and
later the diffusion decreases and S remains on the surface.
The heating of 300 °C provides activation energy for further
diffusion with a simultaneous decrease of the WF. Figure
ments ͑Fig. 3͒. The latter measurements show that the S is
desorbed as a SiS molecule, indicating clearly that the Si-S
bond energy is greater than that of Si-Si, which may indeed
be the dominant cause of the substrate restoration to
Si͑100͒1ϫ1. The greater S-Si bond energy than that of a
Si-Si substrate has been reported some time ago.33 The heat-
ing results ͑Fig. 5͒ indicate that, as soon as part of the S is
desorbed from the surface and the coverage is р0.5 ML the
reconstructed ͑2ϫ1͒ comes back again. Our model of S ad-
15
sorption on the Si substrate is in agreement with Kaxiras,
and later with Kruger and Pollman’s16 theoretical calcula-
tions. According to their results adsorption of S or Se causes
the surface restoration of the reconstructed Si͑100͒2ϫ1 sub-
strate to its original bulk-terminated surface. Kaxiras, in his
report, considered different structures consisting of embed-
ding and mixing of the group-VI adatoms with Si substrate
atoms. He emphasizes, however, that the restored surfaces
are stable against all of the considered alternative structures.
This is in agreement with our measurements, which indicate
an imbedding of the second S layer into the Si bulk.
Next, we will try to justify the WF variations in correla-
tion with the corresponding adsorption sites of S on the Si
substrate. As is explained in Sec. III, the surface dipole mo-
ment changes are mainly due to those of the dipole length,
which are dictated by site changes of the S on the Si sub-
strate. However, the WF changes are proportional to the di-
pole moment ͓Eq. ͑1͔͒.
From the initial slope of the WF curve ͑Fig. 1͒ the initial
dipole moment of S may be calculated by the Hemholtz
equation:
6
͑d͒ shows a possible binding of the embedded S atoms,
which we call disulfide state. Figure 7 shows a top view of
a͒ a clean Si͑100͒2ϫ1 surface, ͑b͒ the hemisulfide on
͑
Si͑100͒2ϫ1, ͑c͒ the monosulfide on Si͑100͒1ϫ1, and ͑d͒ the
disulfide on Si͑100͒1ϫ1.
p0ϭ͑1/2␲͒͑⌬⌽/⌬N͒
N→0
The break of the bond between the Si atoms of the dimers
and the subsequent rearrangement to a ͑1ϫ1͒ structure is
consistent with a strong Si-S interaction. This strong interac-
tion is suggested by the chemical shift of the Auger Si͑92
eV͒ peak during S deposition ͑Fig. 2͒ and the TDS measure-
ϭ1/2␲300ϫ10Ϫ18͑⌬⌽/⌬N͒N→0 Debye,
͑1͒
where the sulfur atomic density Nϭ⌰S 6.8ϫ1014
Ϫ2
atoms cm , and ⌰ is the coverage of S in monolayers. The
S

4440
ARIS PAPAGEORGOPOULOS et al.
55
Besides the theoretical support14,15 on the restoration of
the semiconductor surfaces to their original bulk-terminated
geometry achieved by S and Se adsorbates, there are also
several experimental results mentioned in the Introduction.
These results, however, refer to As on Ge͑111͒ ͑Ref. 24͒ and
on Si͑111͒,2
5,26
and to Cl on Ge͑111͒. Weser et al. re-
ported that S on Ge͑100͒2ϫ1 changed the ͑2ϫ1͒ structure to
1ϫ1͒ and that the S/Ge͑100͒1ϫ1 system was regarded as an
ideal terminated surface. The same authors, however, have
27
28
͑
29
not observed any S overlayer on the Si͑100͒2ϫ1 surface.
Recently, Moriarty, Koenders, and Hughes30 reported that
room-temperature adsorption of S resulted in the formation
of an overlayer on Si͑100͒2ϫ1, retaining the ͑2ϫ1͒ recon-
struction. They also report that annealing of S covered
Si͑100͒2ϫ1 at 325 °C leads to the desorption of the sulfur
overlayer. As was already mentioned, the complete removal
of S takes place by heating the substrate to 650 °C. The same
authors, in continuing their investigation, discovered coexist-
ing c͑4ϫ4͒ and ͑2ϫ1͒ surface reconstructions after the de-
sorption of S at 325 °C. Our finding that S is desorbed as a
SiS molecule shows that heating causes depletion of Si from
the surface. Annealing at relatively low temperatures would
cause a partial removal of Si from the surface, which could
change the reconstruction from ͑2ϫ1͒ to a c͑4ϫ4͒. Although
our structural models are consistent with the experimental
results, we cannot rule out completely the possibility that,
from the beginning of deposition, S forms two-dimensional
islands of ͑1ϫ1͒. Above a certain coverage, the islands coa-
lesce, leading to a uniform ͑1ϫ1͒ structure at 1 ML. More
work is needed to be done.
FIG. 8. Location of the S atoms ͑a͒ on the Si͑100͒2ϫ1 surface,
b͒ on the Si͑100͒1ϫ1 surface.
͑
initial dipole moment of S was found to be p ϭ0.4 Db ͑De-
0
bye͒. If it is considered that in the hemisulfide state the S
atoms reside on the dimers as in Figs. 6͑a͒ and 8͑a͒, and that
p ϭqd or qϭp /d, where q is the charge of each S adatom
0
0
and dϭ1.87 A, it is found that qϭ0.04e ͑where e is the
charge of an electron͒. This indicates that the charge of the S
overlayer is very small to consider the bonds of S on the Si
substrate as ionic. Most likely, the S-Si bond is covalent, in
agreement with the chemical shift ͑Fig. 2͒ and the TDS mea-
surements ͑Fig. 3͒. After the restoration, the distance be-
tween the S overlayer and the topmost layer of the Si sub-
strate decreases and becomes dϭ1.09 A ͓Fig. 8͑b͔͒.
Considering this dipole length and the finding that qϭ0.04e,
it is found that p ϭqdϭ0.2 Db. Therefore, the value of p is
0
0
smaller in the monosulfide state than in the hemisulfide. This
is consistent with the decrease in slope and deviation from
linearity of the WF curve above 0.5 ML of S ͑4th dose͒ on
Si͑100͒ ͑Fig. 1͒, when the reconstructed ͑2ϫ1͒ Si surface
starts to change to its ͑1ϫ1͒.
During the binding of S to the dimers of the Si substrate,
we cannot preclude a decrease of the Si substrate WF due to
a transition of the asymmetric dimers to their symmetric ar-
rangement. It has already been mentioned in the Introduc-
tion: the existing view is that the dimers of the Si͑100͒ are
buckled. This asymmetric deformation increases the dipole
V. CONCLUSION
The adsorption of elemental S at room temperature causes
the change of the reconstructed Si͑00͒2ϫ1 substrate to its
original bulk-terminated Si͑100͒1ϫ1 surface. The S adsor-
bate forms initially a ͑2ϫ1͒ structure at 0.5 ML on the
Si͑100͒2ϫ1 substrate and subsequently
a ͑1ϫ1͒ on
Si͑100͒1ϫ1. Above 1 ML, sulfur is imbedded into the Si
bulk near the surface. The sticking coefficient of S on the
Si͑100͒2ϫ1 surface is constant and equal to unity for the
first 2 ML. Deposition of S at RT up to 1 ML increases the
work function of the surface by about 0.3Ϯ0.05 eV. Above 1
ML, as the sulfur is diffused into the Si bulk, the work func-
tion decreases. Surface dipole moment estimations based on
the work-function measurements suggest that the Si-S bond
is covalent. The deposition of S causes a chemical shift of
the Si͑92 eV͒ peak of 1.5 eV, indicating a strong S-Si inter-
action, while the TDS measurements show that S is mainly
desorbed in the form of a SiS compound. This result supports
the argument that the Si-S bond energy is greater than that of
Si-Si, which may be the driving force of the Si͑100͒2
ϫ1→Si͑100͒1ϫ1 transition.
moment of the dimers, and the WF is greater than that of the
substrate with unbuckled ͑symmetric͒ dimers.1
8,34
The exist-
ence of the asymmetric dimers is supported experimentally
1
9,20
21
by ion scattering
and LEED experiments. Recent STM
͑
Refs. 22 and 23͒ measurements and theoretical
1
4,15
calculations
make the view of asymmetric dimers even
stronger. From our measurements it is not clear that the
dimers remain asymmetric in the hemisulfide state or that
during S deposition the asymmetric dimers change to sym-
metric. In the latter case, the increase of the work function
during S deposition in the hemisulfide state would be com-
pensated to some degree by the WF lowering during the
transition of the asymmetric dimers to their symmetric state.
The increase of the WF in the hemisulfide state, however,
was very close to 0.25 eV measured ͑in the same UHV sys-
tem͒ for 0.5 ML of S on the Ni͑100͒ ͑Ref. 31͒ surface, which
may indicate that the increase in WF up to 0.5 ML was due
to the adsorption of S alone and not to any structural change
of the dimers.
ACKNOWLEDGMENTS
The authors wish to acknowledge the support of NASA
Grant No. NCC3-286 and the NASA High Performance
Polymers and Ceramics Center at Clark Atlanta University
Grant No. NAGW-2939.

55
ADSORPTION OF ELEMENTAL S ON Si͑100͒2ϫ1: . . .
4441
1
20
A. N. MacInnes, M. B. Power, and A. R. Barron Appl. Phys. Lett.
2, 711 ͑1993͒.
R. R. Chang and D. L. Life, Appl. Phys. Lett. 53, 134 ͑1988͒.
R. K. Jain and G. A. Laundis ͑unpublished͒.
M. Yamaguchi et al., Appl. Phys. Lett. 44, 432 ͑1984͒.
A. Yamamoto, M. Yamaguchi, and C. Vemura, Appl. Phys. Lett.
M. Aono, Y. Hou, C. Oshima, and Y. Ishizawa, Phys. Rev. Lett.
49, 567 ͑1982͒.
B. W. Holland, C. B. Duke, and A. Paton, Surf. Sci. 140, L 269
͑1984͒.
6
2
3
4
5
21
2
2
R. J. Hamers, R. M. Tromp, and J. E. Demuth, Phys. Rev. B 34,
5343 ͑1986͒.
44, 611 ͑1984͒.
23
24
R. A. Wolkow, Phys. Rev. Lett. 68, 2636 ͑1992͒.
R. D. Bringans, R. I. G. Uhrberg, R. Z. Bechrach, and J. E.
Northrup, Phys. Rev. Lett. 55, 533 ͑1985͒.
R. D. Bringans, R. I. G. Uhrberg, and R. Z. Bechrach, Phys. Rev.
B 34, 2373 ͑1986͒.
R. R. G. Uhrberg, R. D. Brigans, M. A. Olmstead, R. Z.
Brachrach, and J. E. Northup, Phys. Rev. B 35, 3945 ͑1987͒.
R. D. Schnell, F. J. Himpsel, A. Bogen, D. Rieger, and W. Stein-
mann, Phys. Rev. B 32, 8052 ͑1985͒.
T. Weser, A. Bogen, B. Konrad, R. D. Schnell, C. A. Schug, W.
Moritz, and W. Steinmann, Surf. Sci. 201, 245 ͑1988͒.
T. Weser, A. Bogen, B. Konrad, R. D. Schnell, C. A. Schug, and
W. Steinmann, in Proceedings of the 18th International Confer-
ence on the Physics of Semiconductors, edited by O. Engstrom
6
7
I. Weinberg, C. K. Swartz, and R. E. Harf ͑unpublished͒.
R. Leonelli, C. S. Sundaraman, and J. F. Currie, Appl. Phys. Lett.
5
7, 2678 ͑1990͒.
S. Shikoda, H. Okada, and H. Hayashi, J. Appl. Phys. 69, 2717
1991͒.
A. N. MacInnes, M. B. Power, and A. P. Barron, Chem. Mater. 4,
1 ͑1992͒.
25
8
9
0
͑
2
2
2
2
6
7
8
9
1
1
M. B. Power, A. N. MacInnes, A. F. Hepp, and A. R. Barron, in
Chemical Perspectives of Microelectronic Materials III, edited
by C. R. Abernathy et al., MRS Soc. Symp. Proc. No. 282 ͑Ma-
terials Research Society, Pittsburgh, 1993͒, p. 659.
1
1
1
1
2
3
A. Madhukar, Thin Solid Films 231, 8 ͑1993͒.
K. P. Pande and D. Gutierrez, Appl. Phys. Lett. 46, 416 ͑1985͒.
J. Chare, A. Choujaia, C. Santinelli, R. Blanchef, and P. Victoro-
vitch, J. Appl. Phys. 61, 257 ͑1987͒.
͑World Scientific, Singapore, 1987͒, p. 97.
1
4
30
31
Peter Kruger and Johannes Pollman, Phys. Rev. B 47, 1898
P. Moriarty, L. Koenders, and G. Hughes, Phys. Rev. B 47,
15 950 ͑1993͒.
C. A. Papageorgopoulos and M. Kamaratos, Surf. Sci. 338, 77
͑1995͒.
M. Blaszczyszyn, R. Blaszczyszyn, Z. Medewski, A. J. Melmed,
and J. Madey, Surf. Sci. 131, 433 ͑1983͒.
M. Copel, M. C. Reuter, E. Kaxiras, and R. M. Tromp, Phys. Rev.
Lett. 63, 632 ͑1989͒.
͑
1993͒.
1
1
5
6
Efthimios Kaxiras, Phys. Rev. B 43, 6824 ͑1991͒.
R. E. Schlier and H. E. Farnsworth, J. Chem. Phys. 30, 917
3
3
2
3
͑
1959͒.
1
7
The Chemical Physics Of Solid Surfaces and Heterogenous ca-
talysis, edited by D. A. King and D. P. Woodruff ͑Elsevier,
Amsterdam, 1988͒, Vol. 5, p. 37.
D. J. Chadi, Phys. Rev. Lett. 43, 43 ͑1979͒.
R. M. Tromp, R. G. Smeenk, and F. W. Saris, Phys. Rev. Lett. 46,
1
1
8
9
34
Woodruff and T. A. Delchar, Modern Techniques of Surface Sci-
ence ͑Cambridge University Press, Cambridge, 1986͒.
9392 ͑1981͒.