Quantum confinement effect in self-assembled, nanometer silicon dots

Article abstract of DOI:10.1063/1.122923

The first subband energy at the valence band of self-assembled silicon quantum dots grown by low-pressure chemical vapor deposition on ultrathin SiO₂/Si substrates has been measured as an energy shift at the top of the valence band density of s

Full text of DOI:10.1063/1.122923

Quantum confinement effect in self-assembled, nanometer silicon dots
S. A. Ding, M. Ikeda, M. Fukuda, S. Miyazaki, and M. Hirose
Citation: Applied Physics Letters 73, 3881 (1998); doi: 10.1063/1.122923
View online: http://dx.doi.org/10.1063/1.122923
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/73/26?ver=pdfcov
Published by the AIP Publishing
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
93.180.53.211 On: Wed, 19 Feb 2014 07:11:51

APPLIED PHYSICS LETTERS
VOLUME 73, NUMBER 26
28 DECEMBER 1998
Quantum confinement effect in self-assembled, nanometer silicon dots
S. A. Ding,a) M. Ikeda, M. Fukuda, S. Miyazaki,b) and M. Hirose
Department of Electrical Engineering, Hiroshima University, Kagamiyama 1-4-1,
Higashi-Hiroshima 739-8527, Japan
͑
Received 6 July 1998; accepted for publication 22 October 1998͒
The first subband energy at the valence band of self-assembled silicon quantum dots grown by
low-pressure chemical vapor deposition on ultrathin SiO /Si substrates has been measured as an
2
energy shift at the top of the valence band density of states by using high-resolution x-ray
photoelectron spectroscopy. The systematic shift of the valence band maximum towards higher
binding energy with decreasing the dot size is shown to be consistent with theoretical prediction.
The charging effects of the silicon dots and the SiO layer by photoelectron emission during the
2
measurements have been taken into account in determining the valence-band-edge energy. © 1998
American Institute of Physics. ͓S0003-6951͑98͒02052-X͔
The observations of negative conductance associated
dot height was accurately measured by atomic force micros-
with the resonant tunneling through SiO /Si quantum dot/
copy ͑AFM͒ images, while the diameter determined from
AFM was a little larger than the real size because of the
diameter of AFM tip apex. Therefore, high-resolution scan-
ning electron microscopy was used to calibrate the average
dot diameter which was by about 30% overestimated in the
case of AFM. High-resolution transmission electron micros-
copy was also used to confirm the formation of single crys-
talline Si dot with a hemispherical shape.13 The average dot
height or diameter was determined by fitting a log normal
function to the measured size distribution. For the dots pre-
pared in this study, the standard deviations for the dot height
and diameter were about 40% of the average height and
about 20% of the average diameter, respectively, indepen-
dent of the deposition temperature.
2
1
SiO double barrier structures and the visible photolumines-
2
cence from nanometer Si quantum dots at room
2
–4
temperature
have been recently reported. The subband
formation in nanometer Si dots has been suggested by a
blueshift of the optical absorption edge with decreasing the
4
–6
average dot size. However, the dot size dependence of the
energy band gap as determined by the onset of absorption
spectrum has not been well explained by theories.7
–12
For
better understanding of the quantum size effect in Si dots,
direct measurement of the subband energy is needed. Silicon
quantum dots have been produced by laser breakdown of
2
5
silane, laser ablation of silicon in inert gas, or low-pressure
6
chemical vapor deposition ͑LPCVD͒. Recently, self assem-
bling of silicon quantum dots with a fairly uniform size dis-
tribution and a high areal density has been achieved by con-
The high-resolution XPS measurements were carried out
by ESCA-300 ͑Scienta Instruments͒, using monochroma-
tized Al K␣ radiation ͑1486.6 eV͒ with an acceptance angle
of 3.3°. The instrument energy resolution defined by the full
width at half maximum ͑FWHM͒ of Ag Fermi edge is 0.3 eV
trolling the early stages of LPCVD on SiO /c-Si
2
1
3
substrates. It is also found that the OH termination of the
SiO surface by dilute HF treatment enhances silicon nucle-
2
ation sites, resulting in a significant decrease of the dot size
and an improvement of the size distribution. Thus, precise
for 75 eV pass energy. The base pressure during measure-
Ϫ10
ments was maintained in the 10
Torr range. We kept the
control of chemical states of the SiO surface enables us to
same measurement conditions through all the experiments as
4 kW x-ray power, 75 eV pass energy, and 0.5 mm entry slit
width. The Si 2p core-level and the valence band spectra for
2
fabricate nanometer scale silicon dots with an areal density
1
1
Ϫ2
of more than 10 cm . In this letter, we report on the size
dependence of the valence band maximum for self-
Si dots formed on SiO surfaces at different temperatures
2
assembled Si quantum dots on SiO by using high-resolution
x-ray photoelectron spectroscopy ͑XPS͒.
were precisely measured at photoelectron take-off angles of
5° and 10°, respectively. In addition the Si 2p and valence
band density of states for a chemically cleaned, hydrogen-
terminated p-Si͑100͒ surface were measured for comparison.
Measured valence band and Si 2p core-level spectra for a
2
After chemical cleaning of p-type Si͑100͒ wafers ͑10
⍀
cm͒, they were oxidized at 1000 °C in a 2% dry oxygen
diluted with nitrogen to grow about 3.2–5.5 nm thick SiO2
layers on the Si͑100͒ surfaces. The SiO surface was subse-
Si dots/5.5 nm SiO /Si(100) structure and a hydrogen-
2
2
quently treated by a 0.1% HF solution for 1 min to form
surface Si–OH bonds which enhance the dot density and
improve the uniformity of dot size.13 Si quantum dots were
terminated bulk Si͑100͒ surface are compared in Fig. 1. The
valence band spectrum for a 5.5 nm SiO /Si structure ͑de-
2
noted by VBS-O͒ is also shown in Fig. 1͑a͒. By comparing
self assembled on the SiO surfaces at 565–580 °C using
the valence band spectra for the Si dots/SiO /Si structure
2
2
LPCVD, in which a pressure of pure silane was maintained
at 0.1 Torr. The dot size was controlled by the deposition
temperature and time as already reported elsewhere.13 The
͑denoted by VBS-D͒ with the VBS-O͒, the feature of the
VBS-D below 5 eV is referred to the energy band arising
from Si 3p states of Si dots. The Si substrate signal through
5
.5 nm thick SiO is negligible at a photoelectron take-off
2
a͒
angle of 10°. The valence band edge of the Si dots exhibits a
shift of about 0.6 eV to the higher binding energy from that
CREST, Japan Science and Technology Corporation ͑JST͒.
Electronic mail: miyazaki@sxsys.hiroshima-u.ac.jp
b͒
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
003-6951/98/73(26)/3881/3/$15.00 3881 © 1998 American Institute of Physics
3.180.53.211 On: Wed, 19 Feb 2014 07:11:51
0
9

3882
Appl. Phys. Lett., Vol. 73, No. 26, 28 December 1998
Ding et al.
smaller than that in the same thick SiO layer of the SiO /Si
2
2
structure under equal measurement conditions. This is the
reason for the smaller shift of the valence band spectrum for
the Si dots/SiO /Si structure with respect to that for the
2
SiO /Si structure as shown in Fig. 1͑a͒. In order to reveal the
2
quantum size effect in nanometer Si dots, we focus attention
to the valence band edge and the Si 2p core-level peak for the
Si dots. Both the charging effect and the quantum confine-
ment in the Si dots result in an energy shift of the valence
band spectrum for the dots. We have obtained the same en-
ergy difference between the valence band maximum ͑VBM͒
and Si 2p peak energy for the bulk Si͑100͒ and the Si͑100͒
substrate of a 3.2 nm SiO /Si(100) structure, indicating that
2
the influence of the escape depth difference between the pho-
toelectrons from the valence band and Si 2p core level on the
energy shift is negligible. Therefore, it is reasonable to con-
clude that the charging effect causes almost the same amount
of energy shift for both the valence band top and Si 2p core
level of the dots. Consequently, the dot charging effect for
the Si dots/SiO /Si structure can be eliminated by employing
2
the Si 2p core level for the dots as a reference energy. Unlike
the valence-band-edge energy, the core-level energy is not
influenced by the quantum size effect in Si dots because the
wave functions are strongly localized. To keep the dot charg-
ing effect identical during the XPS measurements of the
Si 2p core-level and valence band spectra, each set of the two
spectra was obtained without any change in measurement
conditions. Since the Si 2p core-level photoelectron peak en-
ergy from the Si dots was used as a reference, the average
Coulombic charging effect of the dots could be canceled out
as a difference between the Si 2p reference energy and the
valence band edge. Thus the measured energy shift of VBM
with respect to that in bulk Si can be attributed to the quan-
tum size effect.
FIG. 1. Valence band spectra for bulk Si, Si dots/5.5 nm SiO /Si and 5.5
2
nm SiO /Si structures for a take-off angle of 10° ͑a͒, and the Si 2p core-
2
level spectra for bulk Si ͑solid line͒ and the Si dots/SiO /Si structure ͑open
2
circles͒ where the suboxide signals are also indicated ͑b͒. The spectra were
not corrected for charging effect. The Si dots were deposited at 580 °C.
of bulk Si. This shift arises not only from the dot charging
effect but also from the subband formation in the dot. It
should be noted that the VBS-O and VBS-D in the energy
range from 11 to 17 eV originate from the Si 3p–O 2p and
Si 3s–O 2p bonding orbitals in the oxide layers.14 A relative
shift of the valence band peak positions at about 12 eV for
the VBS-D appears to be less positively charged than the
VBS-O as indicated by the arrows near 12 eV. The reason
for this will be discussed later. The Si 2p spectrum shown in
In order to determine the energy separation between the
VBM and core level in the Si dots, we encountered the prob-
0
lem that the Si 2p spectral shape for dots is significantly
Fig. 1͑b͒ for the Si dots/SiO /Si structure is composed of the
broader than that for bulk Si. This arises from the differential
charging effect associated with the dot size distribution.
Similar to the Si 2p for bulk Si, all the Si 2p signals for Si
dots on SiO were deconvoluted into the Si 2p and Si 2p1/2
2
4ϩ
two strong components for the chemically shifted Si signal
0
from the SiO layer and for metallic Si signal from the dots.
2
0
Note that a contribution to the Si peak from the Si substrate
2
3/2
through the 5.5 nm thick SiO layer is negligible at take-off
spin-orbit doublet which are separated by 0.61 eV with the
2
0
angles of 5° and 10°. The Si 2p3 signal for the dots is
Si 2p1/2 /Si 2p3/2 intensity ratio of 0.5. The chemical shifts
/2
1
ϩ
2ϩ
3ϩ
broad because of the size distribution and shifted to higher
binding energy by about 0.5 eV in Fig. 1͑b͒ with respect to
the Si 2p3/2 bulk signal mainly due to the charging effect of
silicon dots. Because of the dielectric properties of the oxide
layer and the undoped Si dots, the emission of photoelec-
trons during XPS measurements creates positive charges in
for suboxides signals Si , Si , Si , and silicon dioxide
4ϩ
0
Si with respect to the metallic Si state Si were assumed to
be the same as a previous study. Taking the binding energy
15
of Si 2p3/2 for bulk Si as the energy reference for all the
other Si 2p signals from Si dot/SiO structures, the energy
2
scale of the valence band spectra has been corrected by shift-
the SiO layer and presumably also in the dots, which induce
ing each spectrum by a value given by the difference be-
2
0
the corresponding spectral shift to the higher binding energy
side. The induced positive charge density in the oxide sur-
face layer and the dots should be kept constant under steady-
state conditions because the stable spectra were obtained, but
it depends on the x-ray power density on the surface. As for
the Si dots/SiO /Si structure, in which the SiO surface is
tween the respective Si 2p3 peak position for the dots and
/2
that for bulk Si to eliminate the charging effects in Si dots.
Thus the valence band spectra for Si dots/SiO structures
2
with different dot sizes can be directly compared with the
bulk Si signal by eliminating the charging effect. As shown
in Fig. 2, the valence band spectrum shifts towards the
higher binding energy with decreasing average dot height
from 4.3 to 1.2 nm. By linearly extrapolating the leading
edge of the valence band to zero intensity, the VBM posi-
2
2
only partly covered by the dots with a coverage of about
0%, the Si dots and SiO layer are differentially charged
4
2
during the measurements. Therefore, the average charge den-
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
sity in the SiO layer of the Si dots/SiO /Si structure is
tions were determined for bulk Si and Si dots with different
2
2
9
3.180.53.211 On: Wed, 19 Feb 2014 07:11:51

Appl. Phys. Lett., Vol. 73, No. 26, 28 December 1998
Ding et al.
3883
Theoretical studies on the electronic structures and opti-
cal properties of Si quantum dots are not very consistent with
each other.7 Based on these calculations, the enlargement
of the band gap (⌬E) in a Si quantum dot with respect to the
energy band gap of bulk Si could be expressed as: ⌬E
–12
Ϫ␣
ϰd , where d is the diameter of a spherical dot and the
exponent ␣ varies from 1 to 2, depending on calculation
methods.7
,10,11
Wang and Zunger have compared their cal-
culated result as obtained by empirical pseudopotential plane
12
8
–11
wave theory with other calculations,
tion of ⌬E ͑eV͒ϭ88.34d
giving a fitted rela-
Ϫ1.37
(d is in Å͒ for Si quantum
dots. Changes in the VBM energy for spherical dots with
different sizes have also been calculated in Ref. 12 as indi-
cated in Fig. 3. There exists significant difference between
the dependence of calculated VBM energy shift on spherical
dot diameter and that of experimental VBM shift on hemi-
spherical dot height, while the difference becomes small if
replacing the hemispherical dot height by the diameter of
equivalent sphere as shown by the dashed curve in Fig. 3.
Therefore, the experimental VBM shifts are consistent with
the theoretical values if considering the difference between
real dot shape and theoretical model. To obtain the energy
shift for the conduction band edge, other methods, such as
x-ray absorption spectroscopy16 could be employed. It
should be noted that almost same energy shift of the valence
band edge and the conduction band edge in silicon quantum
dot has been predicted by the calculation in Ref. 12. In sum-
mary, the quantum confinement effect in nanometer Si dots
is confirmed by using high-resolution photoelectron spec-
troscopy. The measured VBM energy shift is considered to
be consistent with the theoretical calculation.
FIG. 2. Valence-band-edge spectra for bulk Si and Si quantum dots mea-
sured at a take-off angle of 5°. The valence band spectra are aligned by
taking Si 2p3/2 peaks for bulk Si and Si dots as energy references. The
deposition temperature ͑T͒ and the corresponding average dot height ͑H͒ are
tabulated in the inset.
average heights. The extrapolated VBM energy for Si quan-
tum dots is plotted as a function of the dot height as shown in
Fig. 3. The height of hemispherical dot is an important di-
mension to assess the quantum size effect, while theoretical
calculations were made mostly for spherical dot. We there-
fore introduce an equivalent sphere diameter for the hemi-
spherical dot, which is defined as the diameter of a sphere
with the same volume as the dot to preserve the total number
of electronic states. Thus, the VBM energy for the corre-
sponding equivalent dot diameter is also represented in Fig.
3. The nonuniform charging effect will occur in the Si dots
because of the size distribution. This results in broadening of
the Si 2p spectral shape and the tailing of the valence band
spectrum. Therefore, only the average values as measured at
the core-level peak positions can be used as an energy refer-
ence for the data processing. As a matter of fact, the uncer-
tainty of peak position ͑Ϯ20 meV͒ due to the nonuniform
charging effect is rather small compared to the magnitude of
VBM shifts ͑у100 meV͒ that could be attributed to the
quantum confinement. Almost the same results as shown in
Figs. 2 and 3 are obtained also at the take-off angle of 10°,
The authors are grateful to T. Tamura and Y. Hamamoto
for their help with the XPS measurements and the sample
preparation, and also thank Dr. A. Kohno and Dr. W. X. Gao
for their useful discussion. This work has been supported by
the Core Research for Evolutional Science and Technology
͑CREST͒ of the Japan Science and Technology Corporation
͑JST͒.
indicating that the charging effect in dot and SiO has been
2
properly eliminated.
1
M. Fukuda, K. Nakagawa, S. Miyazaki, and M. Hirose, Appl. Phys. Lett.
7
0, 2291 ͑1997͒.
2
3
4
5
6
Y. Kanemitsu, T. Ogawa, K. Shiraishi, and K. Takeda, Phys. Rev. B 48,
883 ͑1993͒.
T. Makimura, Y. Kunii, and K. Murakami, Jpn. J. Appl. Phys., Part 1 35,
780 ͑1996͒.
4
4
K. Shiba, K. Nakagawa, M. Ikeda, A. Kohno, S. Miyazaki, and M. Hirose,
Jpn. J. Appl. Phys., Part 2 36, L1279 ͑1997͒.
Y. Yamada, T. Orii, I. Umezu, S. Takeyama, and T. Yoshida, Jpn. J. Appl.
Phys., Part 2 35, 1361 ͑1996͒.
A. Nakajima, Y. Sugita, K. Kawamura, H. Tomita, and N. Yokoyama,
Jpn. J. Appl. Phys., Part 2 35, L189 ͑1996͒.
7
8
9
0
1
2
T. Takagahara and K. Takeda, Phys. Rev. B 46, 15578 ͑1992͒.
S. Y. Ren and J. D. Dow, Phys. Rev. B 45, 6492 ͑1992͒.
J. P. Proot, C. Delerue, and G. Allan, Appl. Phys. Lett. 61, 1948 ͑1992͒.
B. Delley and E. F. Steigmeier, Phys. Rev. B 47, 1397 ͑1993͒.
M. Hirao, Mater. Res. Soc. Symp. Proc. 358, 3 ͑1994͒.
L. W. Wang and A. Zunger, in Semiconductor Nanoclusters, edited by P.
V. Kamat and D. Meisel ͑Elsevier, New York, 1996͒, p. 161.
K. Nakagawa, M. Fukuda, S. Miyazaki, and M. Hirose, Mater. Res. Soc.
Symp. Proc. 452, 234 ͑1997͒.
J. L. Alay and M. Hirose, J. Appl. Phys. 81, 1606 ͑1997͒.
F. J. Himpsel, F. R. McFeely, A. Taleb-Ibrahimi, and J. A. Yarmoff, Phys.
Rev. B 38, 6084 ͑1988͒.
J. He, D. D. Klug, J. S. Tse, C. I. Ratcliffe, and K. F. Preston, Appl. Phys.
Lett. 71, 3194 ͑1997͒.
1
1
1
13
1
1
4
5
FIG. 3. Comparison between measured energy shift of the valence band
maximum in Si quantum dots and theoretical result in Ref. 12. The equiva-
lent diameter of the hemispherical dot is defined as the diameter of sphere
with the same volume as the dot. The dashed line should be compared with
16
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
the theory ͑open circles͒.
9
3.180.53.211 On: Wed, 19 Feb 2014 07:11:51