surface science

ELSEVIER Surface Science 325 (1995) 33-44

Surface electronic structure of a single-domain Si(111) 4 X 1-In

surface: a synchrotron radiation photoemission study
T. A b u k a w a a,* M . S a s a k i u F. H i s a m a t s u a, T. G o t o a T. K i n o s h i t a c A . K a k i z a k i c
S. K o n o ,
a Research Institute for Scientific Measurements, Tohotm University, Sendai 980-77, Japan
b Department of Physics, Faculty of Science, Tohoku University, Sendai 980-77, Japan
c Synchrotron Radiation Laboratory, Institute for Solid State Physics, The University of Tokyo, Roppongi, Tokyo 106, Japan
Received 22 August 1994; acceptedfor publication 6 October 1994

Abstract

The electronic structure of a Si(111)4 X 1-In surface has been studied by angle-resolved photoelectron spectroscopy
(ARPES). Using a 1.1° off-axis Si(lll) wafer as substrate, a single-domain Si(111)4 x 1-In surface has been prepared in
order to determine the dispersion of surface state (SS) without the obscurity arising from multi-oriented 4 x 1 domains.
Three SSs that cross the Fermi level have been found. Thus, the Si(lll)4 X 1-In surface is concluded to be metallic. The
dispersions of the metallic SS appeared to be almost one-dimensional, suggesting one-dimensional metallic bonds among In
atoms. Completely occupied SSs have been also found. The characteristics of SSs are discussed in relation to the existing
structural models for the Si(111)4 × 1-In surface.

Keywords: Angle-resolvedphotoemission; Indium; Metallic films; Metal-semiconductor non-magneticthin film structures; Silicon; Surface
electronic phenomena; Vicinal single-crystal surfaces

1. Introduction pear in recent STM studies [5,6]. The ordered phase

of ~ X v~-appears at a nominal coverage of 1 / 3
Indium adsorption on a S i ( l l l ) substrate induces ML and is c o m m o n among metal adsorbates of
several ordered phases up to In coverages of 1 - 2 Group IlL As the S i ( l l l ) v ~ - X V~--In surface has
ML (1 M L = 7.8 X 1014 a t o m s / c m 2, being the sur- been studied extensively by UPS [7-9], KRIPES
face atomic density of the trtmcated Si(111)1 X 1) (k-resolved inverse photoelectron spectroscopy) [ 9 -
[1-4]. The phase diagrams based on RHEED obser- 11], STM [5,6], EELS [12,13] and X-ray diffraction
vation [1-4] suggest the following sequence of or- [14], the atomic and electronic structures of the
dered phases with increase in In-coverage: 7 X 7 - - X v~--In surface are now well understood.
V'3 × V ~ - - - ~ × 3V~1---4 X 1 - - ( 1 X 1)R30 °. A The ordered phase of 4 X 1 is a unique phase of
new ordered phase, v~-X V~-, was reported to ap- In/Si(111) in its lower symmetry over the threefold
symmetry of the substrate. Therefore, there are at
least three symmetrically equivalent domains of the
* Corresponding author. Fax: +81 22 211 9119. 4 X 1 phase. Fine STM images obtained for several

0039-6028/95/$09.50 © 1995 Elsevier Science B.V. All rights reserved

SSDI 0039-6028(94)00693-8
34 T. Abukawa et aL / Surface Science 325 (1995) 33-4¥

phases of I n / S i ( l l l ) by Nogami et al. [5,6] revealed (ARPES) using synchrotron radiation (SR) has been
that the 4 X 1 surface consists of ridges running applied. Dispersions of surface states have been de-
along the (110) directions and thus visualized the termined by measuring ARPES spectra along many
three-domained nature of the surface. The 4 × 1 symmetric axes of the surface Brillouin zone (SBZ)
surface with an equal three-domain probability com- of single-domain 4 × 1.
plicates the study of this surface. Recently, the atomic
arrangement of the 4 X 1 surface has been investi-
gated by /z-probe Auger electron diffraction (AED) 2. Experimental
[15] and impact collision ion-scattering spectroscopy
(ICISS) [16,17]. In the /z-probe AED study [15], a Photoemission experiments were performed on
single oriented region of the 4 X 1 surface was suc- the ISSP beam line BL-18A at the Photon Factory,
cessfully probed by a /z electron beam and a struc- National Laboratory for High-Energy Physics, KEK.
tural model with 1 ML In-coverage has been pro- The beam line is equipped with a constant-
posed. Stevens et al. [16] performed their ICISS deviation-angle grazing-incidence monochromator
experiment on a single-domain 4 x 1 surface made [19] and an angle-resolved photoelectron spectrome-
on a vicinal S i ( l l l ) substrate (2.6 ° off from (111)). ter (VG ADES500). The photoelectron spectrometer
They, however, deduced another structural model was equipped with a twin-anode (A1, Mg) X-ray
with 1 / 2 ML of In. Therefore, a general consensus source, a He resonance lamp, an angle-integrated
has not been achieved about the atomic structure of spherical electron analyzer, low-energy electron
the 4 X 1 surface yet. There is another controversy diffraction (LEED) optics and a sample manipulator.
about the electronic structure of the 4 X 1 surface. An extra water-cooled graphite Knudsen cell with a
Kelly et al. [13] reported that their HREELS spec- quartz thickness monitor was installed for In deposi-
trum exhibited an elastic peak accompanied by a tion. The overall energy and angular resolutions at
long tail, indicative of a metallic character of the ARPES measurement were 140 meV and 1° at h u =
4 X 1 surface. However, Stevens et al. [16] reported 21.2 eV, respectively. The energy resolution was
a scanning tunneling spectrum averaged over a wide estimated from an ARPES spectrum for the E F
range of 4 X 1 surface that showed a forbidden gap, cut-off of a Ta sample holder. For the ARPES
indicating that the 4 X 1 surface is semiconducting. measurements, we mainly used an SR at h u = 21.2
Very recently, t)fner et al. [18] have reported a eV to investigate the dispersion of SSs, and an SR at
k-resolved direct- and inverse-photoemission study h u = 25 eV to confirm the k-perpendicular depen-
on a three-domain 4 X I surface. They found three dence of the dispersions. We have also performed a
occupied SS bands and one empty SS band with no core-level photoemission measurement of the 4 X 1
states across the Fermi level (EF), again indicating a surface. The result of the core-level photoemission
semiconducting nature of the 4 X 1 surface. How- will be presented elsewhere [20].
ever, as they used the three-domain 4 x 1 surface, To make a single-domain 4 X 1 surface, we used
dispersions of SSs were not determined, thus there a mirror-polished S i ( l l l ) wafer (22 X 3.5 X 0.4
remains an obscurity. The STS spectrum by Stevens mm3), whose surface normal was 1.1 ° off toward the
et al. [16] may not necessarily be a representative of [011] direction. Therefore, the mean terrace width is
density-of-states of the 4 X 1 surface. In order to estimated to be 160 ~,, which is wider than that of
understand the electronic structures of the 4 × 1 the vicinal surface used previously [16]. Cleaning of
surface with confidence, one should be able to depict the sample in situ was performed by direct-current
the dispersion of surface states without the obscurity heating at 1250°C for five seconds, followed by
of the three-domain surface. Thus, an angle-resolved 1000°C-annealing for a few minutes. The tempera-
photoemission study of a single-domain Si(111)4 x ture of the substrate, 300-1250°C, was monitored by
1-In surface is highly desired. an infrared a n d / o r an optical pyrometer. After the
In the present study, a single-domain Si(111)4 X in-situ cleaning, a sharp 7 X 7 LEED pattern was
1-In surface with wide terrace widths was prepared observed and no impurities were detected by XPS.
and angle-resolved photoelectron spectroscopy As the misoriented angle of the substrate was not
 T. Abukawa et al. / Surface Science 325 (1995) 33-44 35

large, spot-splitting due to a regular array of steps

was not observed. Indium was deposited at room
temperature (RT) onto the 7 X 7 substrate from the
Knudsen cell at a deposition rate of 0.2 M L / m i n .
The absolute amount of In deposition was deter-
mined by a quartz thickness monitor. The In cover-
age after annealing was monitored with XPS In 3d
intensity, calibrated with the thickness monitor. The
uncertainty of the In coverage determination w a s
estimated to be ~ 15%.

3. Preparation of singleldomain 4 × 1-In surfaces

To optimize the condition of making a single-do- Fig. 1. LEED pattern of a single-domainSi(lll)4X 1-In surface.
main 4 X 1 surface, In coverages between 0.25 and The "4X" direction in the pattern corresponds to the [112]
2.0 ML and conditions of annealing (temperature, direction of the substrate(cf. Fig. 2). The electron-beamenergy is
period and the DC current direction) were varied. 89 eV.
Under optimized conditions, we observed f 3 X vr3-,
4 X 3vF3-, 3 ~ X 3g~-, three-domains 4 X 1 (ab- ature of post-annealing. We found that a nearly
breviated as T4 X 1, hereafter), single-domain 4 X 1 perfect $4 X 1 phase could be prepared when the
($4 X 1, hereafter), ~ × v~- and (1 × 1)R30 ° peri- sample was annealed at 500°C for 5 min after the
odicities in LEED. The appearance of these periodic- deposition of 1.4 ML In onto the RT substrate. When
ities, except for 4 X 3v~-, is consistent with former the deposited amount of In was larger than 1.4 ML,
reports [1,5,6]. Newly observed 4 × 3v~- was a rect- the high-coverage phase of (1 × 1)R30Q could not be
angular lattice and was observed only in a narrow removed from: the $4 X 1 phase by annealing before
range of the coverages for vr3 X v~- and ~ × 3~, the low,coverage phase T4 X 1 appeared. When the
i.e. ,~ 0.4 ML. LEED patterns of the 4 × 3v~- and deposited amount o f In was less than 1.4 M L , the
(1 X 1)R30 ° surfaces always showed one-fold sym- v/ff × V~- phase was observed with the $4 X 1 phase
metry, which exhibits the single-domained nature of after annealing and could not be removed by anneal-
the surfaces on the present vicinal surface. On the ing before the T4 X 1 phase appeared. Even ff the
other hand, ~ - X v~- and 3vr~- × ~ surfaces al- deposited amount was 1.4 ML, annealing at a lower
ways showed three-fold symmetric LEED patterns, temperature brought the f7- X ~ phase and a longer
showing the existence of symmetry-equivalent (shorter) annealing period at 500°C brought the T4
multi-domains on the present vicinal substrate. X 1 ((1 X 1)R30 °) phase. We could, however, re-
Since the $4 × 1 surface tended to appear to- peatedly produce a near-perfect $4 X 1 surface with
gether with the (1 × 1)R30 °, v/ff × v ~ - a n d T4 × 1 the optimized procedure. A LEED pattern of the
surfaces, a strict condition was to be searched under single-domain 4 x 1 surface is shown in Fig. 1. The
which a pure $4 X 1 phase extends on a wide area S 4 X 1 spots are very sharp and spots due to the
on the sample surface; The relationship among these other 4 X I domains and other periodicities are hardly
phases in terms of coverage was concluded from the seen.
present study as follows: T4 X 1 _< $4 × 1 = v ~
× v~- < (1 × 1)R30 °. The nominal coverage for the
$4 X 1 surface was estimated to be 1 ML from XPS 4. Results of ARPES study
In 3d intensity. The intermixture of the other phases
with $4 × 1 depended on the amount of In deposited In Fig. 2, surface Brillouin zones (SBZs) of the
onto the RT substrate, and on the period and temper- $4 x 1 surface are shown with solid lines in the
36 T. Abukawa et al. / Surface Science 325 (1995) 33-44

repeated zone scheme and part of the SBZ of the

S i ( l l l ) l × 1 is also shown with broken lines. Lateral
orientations of the S i ( l l l ) substrate and symbols for
symmetry points used in this study are also shown in
Fig. 2. In the ARPES experiment, we scanned the
photoelectron-detection angle of both the polar (0)
and azimuthal (th) angles in order to probe SS
dispersions along several symmetrical axes of the
4 × 1 SBZs in Fig. 2. Unless otherwise stated, p-
polarized SR of hu = 21.2 eV was used at an inci-
dent angle of 45 °. Each spectrum was measured in m2
the range of binding energy ( E B) of E B < 6 eV at a
50 meV step, with a maximum count of ~ 20 000 to
obtain a good signal-to-noise ratio. IP
Typical spectra along the F1-X 1 direction, i.e. the
[110] direction (cf. Fig. 2), are shown with dots in • dl
Fig. 3 with relevant spectral features labeled. To
compare the spectra between the [110] and [101]
directions (cf. Fig. 2), spectra along the [101] direc- 6 X ) ' 6 •
tion are also shown with solid lines for 0 = 16° and Binding Energy (eV)
20°: the [101] direction is equivalent to [110] on a
Fig. 3. Typical ARPES spectra (dots) for the single-domain
T4 x 1 surface, but inequivalent on the $4 × 1 sur- S i ( l l l ) 4 X l - [ n surface measured along the symmetry axis of
face. The spectral features are different between the F 1 - X 1 ([110]) of Fig. 2. The p-polarized SR of h v = 21.2 eV is
[110] and [101] directions, thus demonstrating the incident at a polar angle of 45 °. The spectra with solid lines at
single-domained nature of the present sample. 0 = 16° and 20 ° are those along the [101] direction in Fig. 2: this
is to compare ARPES spectra between the [110] and [101] direc-
The experimental dispersion is usually presented
tions.
as an EB-klt diagram, where kll is the parallel com-
ponent of the photoelectron wave-vector at a binding
energy of E B. In the present study, we have used a direct-image method, in which each spectrum is
numerically processed into a gray-scale bar and then
directly laid according to its EB-kll relationship. The
kyj procedure is explained as follows, with the aid of
~[112] /1 or] Fig, 4. A raw spectrum, shown as dots in Fig. 4a, is
converted into the gray-scale bar just below, where
(M--) (© x% the brightness of the scale bar is directly proportional
A .................. ...... A.
to the photoelectron intensity. However, a lot of
_ _

X'2 \ M-'2
/ information is smeared out due to the poor contrast
k of the gray scale. To enhance the contrast, we have
taken the second derivative of the spectra with re-
spect to E B (before the derivation, ll-point smooth-
M~\ kx ing was applied to the raw spectrum). The sign-in-
verted second derivative of the spectrum is shown
/
/ both with a solid line and with a gray-scale bar in
Fig. 4b. The spectral features are now enhanced in
Fig. 2. Surface BriUouin zone (SBZ) of the single-domain
both the solid line and the gray-scale bar. However,
S i ( l l l ) 4 × 1-In surface in the repeated zone scheme (solid lines).
The broken lines indicate the SBZ of the S i ( l l l ) l X l surface.
the contrast in the gray scale is less at larger E B.
Symbols for the symmetry points used in the present study are This is mainly due to the presence of broader fea-
shown. The symbols in parentheses are those for the 1 × 1 SBZ. tures at larger E B because the second derivative
 T. Abukawa et aL/ Surface Science 325 (1995) 33-44 37

the binding energy. In the present study, the gray-

scale bars of D(EB)o, ¢ are directly laid in an Es-kll
coordinate, in order to find more precisely the dis-
persion of SSs; the resulting diagram is called a
k_._. gray-scale EB-kll diagram, hereafter.
The gray-scale EB-kll diagram along the F 1 - X 1
direction is presented in Fig. 5a, which consists of 33
gray-scale bars converted from ARPES spectra mea-
sured for - 14 ° < 0 < 50 ° at 2 ° intervals. Similar
gray-scale E g - k H diagrams along the F~-X~ direc-
tion, which is the direction of four-fold periodicity,
and the X 1 - M I - X 2 - M 2 - X 3 direction are presented
in Figs. 5b and 5c, respectively. In Fig. 6, gray-scale
EB-kll diagrams are shown for the direction of F3 to
X 3, the midpoint (A) of Fa--X ~ to the midpo'__mt(/~)
of X a - M 2 and for the direction of X --I2 to M 2 (cf.
Fig. 2). The diagrams in Fig. 6 are only for E a < 2
, 2 , , eV since no surface-related bands could be identified
6 0 (E~) for E B > 2 eV in these directions. We can easily
Binding Energy (eV) recognize several dispersions of states and estimate
Fig. 4. Comparison of spectra represented by line profiles and by the strengths and widths of the states by the bright-
gray-scale bars. (a) A typical raw spectrum in line prone and a ness and width of the bright bands in the diagrams.
gray-scale representation, (b) same as in (a) but for the second
derivative of the spectrum, (c) same as (b) but for a linewidth-cor- Dispersions of SS bands, m l , m2, m3, dl, d 2 (and
rected second derivative, D(E B) (see text). d), as eventually deduced from Figs. 5 and 6 are
schematically (not to scale) illustrated in Fig. 7. We
describe below how these states are deduced and the
represents curvatures of features. A partial cause of general characteristics of the states.
the peak-width of a SS is the lifetime broadening, In Fig. 5a, dominant bands are labeled as m2, m3,
A Eh, of the final-hole state, which could be consid- dl, d2, b and p, and boundaries of the bulk band
ered, very simply, as proportional to E B [21]. As the projected onto the 1 × l SBZ [22] are shown with
total line-width is given by the convolution of A E h white broken lines: the position of the valence-band
and the instrumental energy resolution (AEinst) , the maximum was estimated from the core-level shift of
effect of lifetime broadening can be corrected for via Si2p between the clean 7 × 7 [23] and 4 × 1-In
a formula: surfaces [20]. In Fig. 5b, dominant bands are labeled
and the boundaries of the bulk-band projection are
~2
shown in the same manner as in Fig. 5a. First, band
D( EB)o,q~ = --ffAE 2 -}- AEi2st X -~BI( EB)o,&, b, which is also visible in the raw spectra in Fig. 3,
appears as a conspicuous structure in Figs. 5a and
(1)
5b. Although band b is totally inside the projected
where I(EB)o, ¢ is a normalized intensity of a spec- bulk band (PBB), it should be a surface resonance
trum measured at (0, ~b), AEms t = 0.14 eV and because the dispersion does not change at another
AEh~(aEB)eV with a being a constant, a is excitation energy, h u = 25 eV. However, the disper-
roughly estimated to be 0.1 for the better contrast of sion does not exhibit four-fold periodicity along
the gray-scale bar in a wide energy range, especially F 1 - X q in Fig. 5b. This can be understood as follows.
at higher binding energies. An example of D ( E B) is Namely, if band b is due to substrate-related surface
shown in Fig. 4c. One can see that the height of the resonances, such as a back-bond state, the effect of
peaks appearing in Fig. 4c is nearly proportional to the "surface n m k l a p p " process in photoemission
that of the raw spectrum for the complete range of would be weak. Thus, the dispersion tends to follow
38 T. Abukawa et aL / Surface Science 325 (1995) 33-44

a 0.0 1.0 C 0.0 1.0

kx (~,-~) ky(,~-~)
Fig. 5 (continued).

that of the 1 × 1 SBZ. Faint bands around E B = 3 - 4

eV in Fig. 5b may represent a trace of the 4 × 1
umklapp process.
The observed band dispersions for Ea < 1.5 eV in
Figs. 5a and 5b are significantly different between
the F1-X ~ and F1-X ~ directions. There are five
dispersive bands, m2, m3, dr, d 2 and p, along the
F t - X 1 direction, while a flat band d and a dispersive
band p are observed along the F1-X ] direction.
Three of the five bands, m2, m 3 and d 2, appear in a
widely opened bulk band gap around the K point of
the 1 × 1 SBZ along the F1-X ~ direction, thus they
can be considered as SS bands. As another band d 1
penetrates into the bulk band gap near the X 1 point,
it should be a mixture of surface state and resonance.
SS band m 2 disperses upward from X1 to F1 and

Fig. 5. Gray-scale E B - kll diagrams made from the linewidth-cor-

rected second derivatives of ARPES spectra: (a) along F 1-X1, (b)
along F1-X], (c) along X 1 - M I - X 2 directions. Dominant struc-
b 0.0 1.0 tures and surface states are labeled and the edges of the projected
ky (.~-1) bulk band are shown with white broken lines.
 T. Abukawu et al. / Surface Science 325 (1995) 33-44 39

o 1
crosses E v at k x = 0.5 A - . Namely, m 2 is a metal-
lic band. m 3 disperses upward from the X t point
toward the next F point and also crosses EF, thus
this is also metallic. The fiat band d is the only band
whose dispersion satisfies the 4 X periodicity in Fig.
5b. W h e n this band appears in the bulk band gap
around I ' 3 (M), the intensity is enhanced, which is a

Fig. 7. Schematic illustration (not to scale) of dispersions of

surface states on the Si(111)4×1-Iu shown in the Cartesian
coordinate. Axes kx and ky are reciprocal axes as in Fig. 2. The
vertical axis represents the binding energy. The top plane of the
coordinate shows the Fermi surfaces and the side planes show
dispersions along the X ' - F - X ' and M-X-M axes.

typical characteristic of a surface state. A s this band

is connected to both d 1 and d 2 bands, as can be seen
in Fig. 6, band d is degenerated from d 1 and d 2.
It is worth noting at this stage that the single-do-
main S i ( l l l ) 4 × 1-In surface has one mirror plane,
which is parallel to the F 1 - X ~ direction, unless In
adsorbates form a low symmetric structure. There-
fore, e v e n / o d d symmetry o f SSs can be determined
from the dependence o f A R P E S spectra measured
along the F 1 - X ] direction on the polarization of SR.
Namely, as the polarization vector of incident SR
was parallel to the mirror plane for Fig. 5b, states
with even symmetry can appear in Fig. 5b. Thus, the
band d was concluded to have even symmetry about
the mirror reflection. W e have also measured A R P E S
spectra along the 1`1-X] direction with the SR nearly
perpendicular to the mirror plane, for which states
with odd symmetry are expected to appear. The
measured A R P E S spectra showed no trace of SSs
that have odd symmetry along the F 1 - X ' ~ direction.
There is a faint band p which follows the periph-
ery of the PBB around the F1 point in Figs. 5a and
0.0 10 5b. The feature p is also visible as a w e a k shoulder
kx in the raw spectra o f Fig. 3. Since the dispersion of p
did not change with the excitation energy, one might
Fig. 6. Gray-scale E B-k/I diagrams in bulk band gap regions for
E B _<2 eV. (a) Along F3-X3, (b) from a midpoint (A) of F3-X'2 ascribe p to a SS. However, it seems more reason-
to a midpoint (~) of X3-M2, (c) along X~ -M e. able to ascribe p to a band o f high density-of-states
40 T. Abukawa et aL / Surface Science 325 (1995) 33-44

t i i
at the periphery of the PBB. Since the density-of- :d
states is generally high at the periphery of the PBB, a a) .-
faint band may appear in ARPES spectra for which
k-perpendicular is not strictly conserved in the pho- x. •
toemission process. ~ kJ " , m3
Fig. 5c is the gray-scale EB-kll diagram along the 7 %,C
direction for the 4 >< periodicity but away from the ..Q
g b) m3
center of the SBZ, so that the PBB would not .*.
overlap with surface states. The PBBs are interpo- ~ . "~ - . : , ..
lated from theory [22] although the details may not
be correct. In the upper gap, there are two SS bands
that distinctly show 4 >< periodicity. The upper and c) \-
the lower states of the two are ascribed to SS bands \ : " __ d, ~3
m 2 and dl, respectively, from the energy positions
of the states at Xa in Figs. 5a and 5c. Band m3,
%
found in Fig. 5a, can be seen as a very faint band r i i
6 4 2 0
between m 2 and d 1 around the X 2 point__of Fig. 5___c. Binding energy (eV) (EF)
This faint band m 3 overlaps with d~ at M~ a__ndM__2
Fig. 8. Examples of raw-spectra of ARPES along the F3-X 3
and disperses by ~ 0.2 eV downward from M to X.
direction: the spectra (a), (b) and (c) roughly correspond to
A flat band is visible at E B = 1.4 eV between X2 k:¢ = 0.4, 0.5 and 0.65 ~-1 of Fig. 5a, respectively. The dominant
a n d M 2 , which is ascribed to d 2 from the connection surface states are labeled.
at X and M in other EB-kll diagrams. Bands m 3 and
d 2 are expected to disperse in accordance with 4 ><
periodicity, although they are not visible in the large E B = 1 eV in Fig. 6. This band is almost flat for
part of the X 1 - M 1 - X 2 - M 2 - X 3 axis. In addition, k x _ ~ 0.3 ~ - 1 , but goes downward at k x = 0.3
two bands, s I and Sz, are visible around the lower ~ - 1 . For k x > ~ 0.5 A - 1 , band d splits into the two
gap at E B > 4 eV, although their dispersions are not bands, d I (upper) and d 2 (lower).
very clear. The two bands seem to be separated by It is worth comparing at this stage a gray-scale
~ 0.3 eV from each other and seem to show a EB-kll diagram in Fig. 6 with a set of raw spectra to
similar dispersion as that of m 2 or dp Thus, we confirm the gray-scale diagram. In Fig. 8, we show
found six SS bands, m2, m3, dÂ, d2, S1 and $2, in typical examples of raw spectra used for making Fig.
Fig. 5, that are intrinsic to the S i ( l l l ) 4 >< 1-In sur- 6a. The spectra in Figs. 8 a - 8 c roug ohly correspond to
face. those at kx=0.4, 0.5 and 0.65 A -1 in Fig. 6a,
Besides the aforementioned bands, other struc- respectively. We can confirm that m 2 (m 3) crosses
tures can be seen in a high E B region of Fig. 5. We E F in the raw spectrum of Fig. 8a (8b), and m 1, m 2,
do not treat these structures, since they are mainly m3, d 1 and d 2 can be identified all together in Fig.
inside the PBB and do not show a 4 >< 1 periodicity. 8c.
The axes of the diagrams in Fig. 6 are all along As we now know from Figs. 5 - 7 that the SS
1 >< directions but away from the center of the SBZ. bands m~, m 2 and m 3 are all metallic bands, two-di-
It is also possible to figure out dispersions along the mensional Fermi surfaces for the three SSs can be
4 >< direction by comparing the three diagrams. Sur- drawn precisely by depicting the positions in the
face states appear more clearly along the F 3 - X 3 4 >< 1 S B Z at which the three bands cross EF, as is
direction (Fig. 6a) than along the equivalent direction shown in Fig. 9. Solid dots represent measured points
of F I - X 1 (Fig. 5a), because the PBB gap opens and are tied with the solid line for each state. All the
widely around F 3 - X 3. In addition to m 2 and m3, states are unoccupied on the left sides of the Fermi
another metallic state m~ appears above m e around surfaces in Fig. 9 and the state m 1 (m 2) is just above
_X3 and crosses E F at places ~ 0.1 ~ - 1 away from (below) E F at the M point. The Fermi surfaces of m 2
X 3. Besides m1_3, an intense band d is observed at and m 3 run approximately parallel to the 4 >< direc-
 T. Abukawa et al./ Surface Science 325 (1995) 33-44 41

--r
X M of Ref. [18], it is possible to assign their result to the
presently found SSs. They observed a flat band at
E~ = 1 eV and two states which were only observ-
able around the boundary of the SBZ at E R = 0.5
F m3 m2 m1 X and ~ 1.5 eV. The flat band can be ascribed to band
Fig. 9. Fermi surfaces of the S i ( l l l ) 4 X 1-In surface plotted in the d I of the present study, and the two states at E B = 0.5
reduced zone scheme. Dots represent the experimental points and ~ 1.5 eV are considered to be parts o f m 2 and
crossing E F. d 2, respectively.
We have observed in the present ARPES spectra
several states that are independent on the component
tion and can be called two-dimensionally opened of the photoelectron wave-vector perpendicular to
Fermi surfaces. On the other hand, the Fermi surface
i
the surface. It is possible to classify the states into
of m 1 is closed at the M point. The fraction of two groups: one that shows the 4 × 1 periodicity and
electron occupation of each band can be estimated the other that does not.
from the occupied area in the SBZ. They are ~ 10, One of the bands not showing the 4 X 1 periodic-
~ 40 and ~ 50% for ml, m 2 and m3, respectively. ity is a faint band p observed along the edge of the
Therefore, the three metallic states are occupied by PBB. As we have described already, band p may not
two electrons per 4 X 1 unit cell and it can be further be a two-dimensional band but due to the edge of the
concluded that there are an even number of valence PBB. Another band without 4 × 1 periodicity is the
electrons on the S i ( l l l ) 4 X 1-In surface as a whole, band b appearing in Figs. 5a and 5b. We have
since the other states are all occupied. already postulated that b is due to a back-bond state
Next, the dispersions of the fully occupied bands of Si substrate, appeared by a weak "surface um-
dl, d 2 and d are described further. Bands d 1 and d 2 klapp process". One reason for this is that the
are components of a degenerated band d which dispersion of b agrees with that of the transverse
appears as a flat band for k x < 0.3 A-1, as well as back-bond state predicted for the ideal S i ( l l l ) l X 1
for all ky. Beyond k x = 0.4 A -~, d splits into two surface [24,25]. The transverse back-bond state cor-
bands, d~ and d 2, as seen in Figs. 6 and 7. The upper responds to the bonding state within the top Si
branch, d~, is observed as a faint structure in the double layer. Therefore, the postulated existence of a
diagrams in Fig. 6 and seems to have a flat disper- transverse back-bond state leads to the suggestion
sion along the 1 X direction in all diagrams. How- that the 4 X 1 surface is composed of an unrecon-
ever, d~ is not flat along the 4 × direction as one can structed S i ( l l l ) l X 1 substrate.
notice from the dispersion of d 1 in Fig. 5c. The Bands w i t h 4 × 1 periodicity are considered t o b e
lower branch, d2, disperses downward beyond kx --~ intrinsic to the S i ( l l l ) 4 X 1-In surface. First, we will
o 1
0.3 A - , but the shape of the dispersion is dependent discuss what electrons might form the intrinsic SSs.
0.__11ky. Namely, d 2 has a minimumI at the boundary The configuration of the ~,alence electrons of indium
X of the SBZ, but not at/~ and M. is 5s25p 1 and indium bulk is an sp-metal. If we
assume an In-coverage of 1 ML for the 4 X 1-In
surface, i.e. four In atoms per 4 X 1 unit cell, then
5. Discussion indium atoms contribute eight 5s-electrons and four
5p-electrons per 4 X 1 unit cell to the SSs. As for the
Ofner et al. [18] found three filled SS bands and substrate, if we assume an unreconstructed S i ( l l l ) l
one empty SS band on a three-domain 4 × 1 surface X 1 substrate, it contributes four dangling-bond elec-
by direct and inverse photoemission study. However, trons to the SSs per 4 × 1 unit cell. Thus, there are
they did not find any sign of metallic SS. Thus, they sixteen electrons per 4 X 1 unit cell which can oc-
concluded that the 4 X 1 surface is semiconducting. cupy eight SS bands. We will classify the SSs into
It is probable that the three-domain 4 x 1 surface two kinds for convenience: one is an interface state
smeared out the spectral feature of metallic SSs that comprises dangling-bond electrons and In va-
found presently. As for the observed three SS bands lence electrons and another is an overlayer state
42 T. Abukawa et al. / Surface Science 325 (1995) 33-44

originated from bonding electrons within the In over- are occupied by ten electrons. Therefore, we have
layer. The interface states may have similar proper- observed at least ten electrons out of sixteen valence
ties as the dangling-bond state of S i ( l l l ) l × 1. If the electrons expected for the 4 × 1-In surface at 1 ML
dangling bonds are completely occupied with In of In coverage. The other electrons might form
valence electrons, four fully occupied interface bands surface resonances vanishing in the bulk bands.
will form. The oveflayer states, on the other hand, The present results are discussed below on the
may have similar properties as those of the metallic basis of the structural models proposed so far [15,16].
bands of In bulk. First, the 1 ML model proposed by AED [15] is
We discuss below the surface states based on the consistent with the present results in terms of the
above consideration. An important feature of the coverage, and is expected to have an even number of
metallic bands ml, m 2 and m 3 is that they com- surface valence electrons. In this model, only the
monly have a minimum near the SBZ boundary and arrangement of the In atoms was proposed and it
that they all are unoccupied around the center of the forms narrow " r i b b o n s " extending along the 1 ×
SBZ. This is a typical characteristic of a partially direction with a regular interval in between. The
filled Px, y state. Therefore, it is supposed that the narrow " r i b b o n " is composed of 4 rows of In array,
three metallic states mainly consist of In5Px, y or- within which the arrangement of In atoms is similar
bitals and that they correspond to the metallic bonds to that of the ideal (001) surface of tetragonal In
among an In overlayer. The metallic SSs found crystal. The nearest-neighbor I n - I n distance in this
presently show a characteristic one-dimensionality in model is ~ 3.25 A, which is in good agreement with
their dispersion, where noticeable dispersions are the nearest-neighbor distance of In metal, i.e. the
present only along the 1 × direction and the Fermi metallic bond length. Therefore, it is reasonable to
surfaces nm almost parallel to the 4 × direction. expect metallic bonds along the 1 X direction within
This strong one-dimensionality indicates that the each of the "ribbons" of this model. As for the
metallic bonds are formed along the 1 × direction inter-" ribbon" interaction, the nearest I n - I n dis-
o
which must be a consequence of the arrangement of tance between neighboring In ribbons is ~ 6 A,
the In atoms. which is too far to form a metallic bond. Thus, a
On the other hand, d 1 and d 2 are completely one-dimensional metallic character is expected for
occupied surface states. There are two plausible ori- this model, which is consistent with the present
gins of the completely occupied bands. First, they findings.
are due to bonds between In5s electrons; second, As for the structural model proposed by ICISS
they are due to bonds between otherwise dangling [16], the proposed In coverage of 1 / 2 ML is incon-
bonds of the substrate and In valence electrons. sistent with the present results, although the model
Inspection of the dispersions of d 1 and d2, leads to has an even nurnber of surface valence electrons
the conclusion that the latter is more plausible; the which is consistent with the present results. The
dispersions of d 1 and d z are similar to those of the nearest-neighbor I n - I n distance in this model is 3.84
dangling-bond state on the ideal Si(111)1 × 1 sur- A, which is ~ 18% larger than the I n - I n metallic
face [24-26], although the dangling-bond state is bond length, so that it may be difficult to form
half-filled at the ideal surface. The presently found metallic bonds among In atoms in this model. Thus,
symmetry property of the d state is also consistent the 1 / 2 ML model is unfavorable, based on the
with that of the dangling bond: the dangling bond, present results. One may expect another structural
which is an sp 3 orbital pointing towards the surface model with 3 / 4 or 5 / 4 ML of In coverage, but
normal, has even symmetry about the mirror plane. those models with such coverages are inconsistent
We have also observed the surface states s 1 and s 2 with the present results because they have odd num-
along X - M at high binding energy. Although we bers of surface valence electrons.
could not obtain the dispersion in detail, sl and s 2 The registration of In atoms on the substrate is
may be related to In5s electrons. another problem. As the hollow site, T4, is the
So far, we have observed seven SSs with 4 × 1 bonding site of In on the Si(lll)v~- × v~--In surface
periodicity. If there is no degeneration, the seven SSs and also as the other hollow site, H3, is the next
 T. Abukawa et al. / Surface Science 325 (1995) 33-44 43

favorable site for In adsorption [5,27], it was be- states and resonances have been found in the gray-
lieved that the hollow sites are plausible sites of In scale EB-kll diagrams and full dispersions in the
adsorption on the 4 X 1 surface [15,16]. However, 4 × 1 SBZ have been obtained for dominant surface
Finney et al. [28] claimed a vertically bonded on-top states. In particular, full dispersions of three metallic
geometry of In at two monolayers of In on a Si(111) surface states have been determined together with
surface. 0fner et al. [18] also discussed the possibil- the Fermi surfaces of the metallic states, The shapes
ity of an on-top geometry for the (1 X 1)R30 ° sur- of the dispersions and the Fermi surfaces of the three
face, although they expected that the 4 × 1 surface metallic bands have appeared to be essentially one-
takes a similar bonding geometry as the x/3 × v~-In dimensional. It is suggested that the metallic states
surface. Therefore, it is not very surprising to expect are due to one-dimensional metallic bonding among
an on-top geometry for In adsorption on the 4 × 1 In atoms on the 4 X 1-in surface. Fully occupied SS
surface. In fact, the on-top geometry seems to give a bands have also been found, which are tentatively
favorable explanation to the present findings of sur- ascribed to bonding states between the otherwise
face states as follows. If In atoms are adsorbed on dangling bonds of unreconstructed S i ( l l l ) l X 1 sub-
on-top sites, p~ a n d / o r s orbitals of In may be used strate and the valence electrons of In atoms. Compar-
in bonds to the Si dangling bonds, so that Px,y ison of the results of electronic-structure calculation
orbitals are free to form bonds among In atoms. On with the present experimental findings is highly de-
the other hand, if In atoms are adsorbed at hollow sired.
sites, Px,y orbitals may be largely involved in the
bonding to the dangling bonds. Thus, the on-top site
seems more favorable than the hollow sites for In Acknowledgements
adsorption in terms of the formation of metallic Px, y
bands. We certainly need a theoretical verification of The authors are indebted to Mr. Y. Akama of the
this argument. Laboratory for Microelectronics, Research Institute
As we have assumed an unreconstructed Si(111)1 for Electrical Communication, Tohoku University,
X 1 surface as the substrate, the present discussion is for the determination of the surface orientation of the
a limited one. If there is a large reconstruction of the substrate S i ( l l l ) by X-ray diffraction. This experi-
substrate, it is possible to change the number of Si ment has been performed under the proposal number
dangling bonds per unit cell, thus changing all argu- 92G292 of Photon Factory, National Laboratory for
ments given so far. In order to further utilize the High-Energy Physics, KEK.
present experimental findings, theoretical studies of
electronic structures are highly desired.
References

[1] H. Hirayama, S. Baba and A. Kinbara, J. Vac. Sci. Technol.

6. Conclusion A 4 (1986) 1416.
[2] T. Aiyama and S. Ino, Surf. Sci. 82 (1979) L585.
[3] M. Kawaji, S. Baba and A. Kinbara, Appl. Phys. Lett. 34
Using a 1.1 ° off-axis S i ( l l l ) substrate, we have (1979) 748.
made a single-domain Si(111)4 X 1-In surface. The [4] S. Baba, M. Kawaji and A. Kinbara, Surf, Sci. 85 (1979) 29.
[5] J. Nogami, S. Park and C.F. Quate, Phys. Rev. B 36 (1987)
single-domain Si(111)4 X 1-In surface was obtained
6221.
at a 1 ML In coverage. Synchrotron radiation [6] J. Nogami, S. Park and C.F. Quate, J. Vac. Sci. Technol. B 6
angle-resolved photoelectron spectroscopy has been (1988) 1479.
applied to the single-domain S i ( l l l ) 4 × 1-In surface [7] G.V. Hansson, J.M. Nichols, P. M~'tensson and R.I.G.
in order to investigate the surface electronic struc- Uhrberg, Surf. Sci. 168 (1986) 105.
tures. A new scheme of compilation (gray-scale E B- [8] T. Kinoshita, PhD Thesis, Tohoku University, 1987.
[9] T. Kinoshita, H. Ohta, Y. Enta, Y. Yaegashi, S. Suzuki and
kll diagram) of sets of ARPES spectra has been S. Kono, J. Phys. Soc. Jpn. 56 (1987) 4015.
adopted in order to obtain a quantitative representa- [10] H. Ohta, T. Kinoshita, Y. Yaegashi, S. Suzuki and S. Kono,
tion of surface-state dispersions. Several surface J. Phys. Soc. Jpn. 57 (1988) 4452.
44 T. Abukawa et al. / Surface Science 325 (1995) 33-44

[11] J.M. Nichols, B. Reihl and J.E. Northrup, Phys. Rev. B 35 [20] T. Abukawa, M. Sasaki, F. Hisamatsu, T. Goto, M. Naka-
(1987) 4137. mura, T. K.inoshita, A. Kakizaki and S. Kono, to be pub-
[12] H. Hirayama, S. Baba and A. Kinbara, Jpn. J. Appl. Phys. 25 lished.
(1986) IA52. [21] S.D. Kevan and D ~ . Shirley, Phys. Rev. B 22 (1980) 542.
[13] M.IC Kelly, G. Margaritondo, J. Anderson, D.J. Frankel and [22] J. Pollmann, P. Kriiger, A. Mazur and G. Wolfgarten, Surf.
G.J. Lapeyre, J. Vac. Sei. Teclmol. A 4 (1986) 1396. Sci. 152/153 (1985) 977.
[14] M.S.Finney, C. Norris, P.B. Howes, R.G. van Silfhout, G.F. [23] F.J. Himpsel, B.S. Meyerson, F.R. McFeely, J.F. Morar, A.
Clark and J.M.C. Thornton, Surf. Sci. 291 (1993) 99. Taleb-Ibrahimi and J.A. Yarmoff, in: Proc. Int. School of
[15] N. Nakamura, K. Armo and S. Kono, Surf. Sci. 256 (1991) Physics Enrico Fermi, Vol. Course CVIII, Photoemission and
129. Absorption Spectroscopy of Solids and Interfaces with Syn-
[16] J.L. Stevens, M.S. Worthington and I.S. Tsong, Phys. Rev. B chrotron Radiation, Eds. M. Campagna and R. Rosei (Else-
47 (1993) 1453. vier, Amsterdam, 1988) p. 203.
[17] D.M. Cornelison, C.S. Chang and I.S.T. Tsong, J. Vac. Sci. [24] F. Casula, S. Ossicini and A. Selloni, Solid State Commun.
Technol. A 8 (1990) 3443. 30 (1979) 309.
[18] H. Ofner, S.L. Surnev, Y. Shapira and F.P. Netzer, Phys. [25] K.C. Pandey and J.C. Phillips, Phys. Rev. Lett. 32 (1974)
Rev. B 48 (1993) 10940. 1433.
[19] S. Suzuki, T. Kashiwakura, A. Kakizaki, T. Kinoshita, A. [26] M. Schliiter, J.R. Chelikowsky, S.G. Louie and M.L. Cohen,
Harasawa, J. Fnjii, S. Suga, M. Fujisawa, H. Kato, S. Sato Phys. Rev. B 12 (1975) 4200.
and T. Ishii, Synchrotron Radiation Laboratory, Institute for [27] For example, J.E. Northrop, Phys. Rev. Lett. 53 (1984) 683.
Solid State Physics, The University of Tokyo, Activity Re- [28] M.S. Finney, C. Norris, P.B. Howes and E. Vlieg, Surf. Sci.
port 1989 (1990) p. 60. 277 (1992) 330.