Nanometer-scale Si selective growth on Ga-adsorbed voids in ultrathin SiO2 films

Article abstract of DOI:10.1016/S0039-6028(99)00511-7

We examined nanometer-scale Ga selective doping by Si growth on Ga-adsorbed voids in ultrathin silicon-dioxide on Si(111) surfaces. The doping processes were observed by scanning tunneling microscopy (STM). Voids in ultrathin oxide films were plugged with a (√3 × √3)-Ga structure, and the selective growth was performed by introducing disilane gas (Si₂H₆). Si crystals were selectively grown only in the voids at 460-550°C. Two-dimensional nucleation was found to start from the edge of the voids. Incorporated Ga atoms mostly segregated during the selective growth and were reconstructed to the (√3 × √3) structure by annealing at 600°C. These results show that Ga doped dots of nanometer-scale can be formed by selective epitaxial growth using an ultrathin silicon-dioxide mask.

Full text of DOI:10.1016/S0039-6028(99)00511-7

Surface Science 431 (1999) L565–L569
www.elsevier.nl/locate/susc
Surface Science Letters
Nanometer-scale Si selective growth on Ga-adsorbed voids in
ultrathin SiO films
2
*
Yoshiki Nitta , Motoshi Shibata 1, Ken Fujita, Masakazu Ichikawa
Joint Research Center for Atom Technology, Angstrom Technology Partnership (JRCAT-ATP),
c/o National Institute for Advanced Interdisciplinary Research (NAIR), 1-1-4 Higashi, Tsukuba, Ibaraki 305-0046, Japan
Received 15 February 1999; accepted for publication 13 April 1999
Abstract
We examined nanometer-scale Ga selective doping by Si growth on Ga-adsorbed voids in ultrathin silicon-dioxide
on Si(111) surfaces. The doping processes were observed by scanning tunneling microscopy (STM). Voids in ultrathin
ǰ
ǰ
oxide films were plugged with a ( 3× 3)-Ga structure, and the selective growth was performed by introducing
disilane gas (Si H ). Si crystals were selectively grown only in the voids at 460–550°C. Two-dimensional nucleation
2
6
was found to start from the edge of the voids. Incorporated Ga atoms mostly segregated during the selective growth
ǰ
ǰ
and were reconstructed to the ( 3× 3) structure by annealing at 600°C. These results show that Ga doped dots of
nanometer-scale can be formed by selective epitaxial growth using an ultrathin silicon-dioxide mask. © 1999 Elsevier
Science B.V. All rights reserved.
Keywords: Epitaxy; Gallium; Oxidation; Scanning tunneling microscopy; Semiconductor–semiconductor thin film structure; Silicon;
Silicon oxides; Surface segregation
Selective growth of nanometer-scale is an
important technique for nanofabrication. Fujita
et al. have previously reported nanometer-scale Si
selective growth using voids in an ultrathin silicon-
windows. This technique of utilizing the voids in
the ultrathin SiO film can be applied for fabricat-
2
ing nanometer-scale dot structures. If the dopant
is incorporated into the Si crystals selectively
grown in the voids, the doping dots of nanometer-
scale can be formed.
dioxide (SiO ) film [1]. They formed the ultrathin
2
SiO film on the Si(001) surface by thermal oxida-
2
tion and the windows of nanometer-scale were
In this letter, we describe nanometer-scale Si
selective growth in Ga-adsorbed voids on the
Si(111) surface. The Si(111) surface was selected
because the reconstructed structure of Ga-
adsorbed surface on Si(111) is simple compared
with that of the Si(001) surface [2,3], and is
suitable for investigating the growth mode.
Ga-adsorbed Si(111) surfaces have two main
opened by thermal decomposition of oxide film.
Then Si crystals were selectively grown in the
* Corresponding author. Fax: +81-298-54-2577.
E-mail address: nitta@jrcat.or.jp (Y. Nitta)
1 Present address: Opto-Electronics Laboratory, Oki Electric
Industry Co. Ltd, 550-5 Higashi-asakawa, Hachioji, Tokyo
193-8550, Japan.
0039-6028/99/$ – see front matter © 1999 Elsevier Science B.V. All rights reserved.
PII: S0039-6028 ( 99 ) 00511-7

LETTERS
SURFACE SCIENCE
L566
Y. Nitta et al. / Surface Science 431 (1999) L565–L569
ǰ
ǰ
structures: a ( 3× 3)-Ga structure for a 1/3
monolayer (ML) of Ga [4] and a (6.3×6.3)-Ga
structure for a 1 ML of Ga [5,6]. The Si(111)
surface is easily uniformly covered with one
domain of Ga induced structure if the coverage is
adequately controlled. In our case Ga atoms are
adsorbed in the voids which are formed by thermal
decomposition of the oxide. Therefore the surface
structures are more complicated due to the diffu-
sion and desorption of Ga atoms on the oxide.
The reconstructed surfaces of Ga-adsorbed voids
were investigated, and the two structures men-
tioned above were observed depending on the
annealing conditions. Disilane gas was supplied to
these Ga-adsorbed voids, and Si selective growth
2×1014 s−1cm−2. Through these processes, STM
topographic images of the surfaces were acquired
with the constant current between 0.06 and 0.1 nA.
Most images were obtained at sample biases of 4–
5 V because of the large band gap of SiO [9,10],
2
but lower biases of 2–3 V were applied to the
sample when the Si surface or Ga-adsorbed surface
structures were observed.
After adsorption of a 1/3 ML of Ga at 550°C,
a (6.3×6.3)-Ga patchwork pattern was observed
in the voids. In this stage some amount of Ga was
thought to be adsorbed on the oxide, because the
image was sometimes disturbed by protrusions
with the diameter of >10 nm which seemed to be
Ga droplets. Since Ga atoms on the oxide seemed
to flow into the voids, effective coverage in the
voids increased to >1/3 ML and the surface of
the voids showed the (6.3×6.3)-Ga structure.
ǰ
ǰ
was achieved in the voids of the ( 3× 3)-Ga
structure. In the previous work, Si epitaxial growth
on wide Si(111)-( 3× 3)-Ga terraces was
ǰ
ǰ
Disilane gas (Si H ) was introduced to these
(6.3×6.3)-Ga voids at 450°C, and the void surface
reported [7]. However, the growth on the narrow
2 6
ǰ
ǰ
areas of the ( 3× 3)-Ga structure has not yet
been clarified. Hence we discuss the Si selective
epitaxial growth in the voids of Si(111)-()-Ga
surface. The processes during the void formation,
Ga adsorption, and selective epitaxial growth were
observed by scanning tunneling microscopy
(STM).
was observed during gas supply. But the surface
structure of patchwork pattern showed no change
for 26 min. The growth rate of Si crystals in the
(6.3×6.3)-Ga was estimated to be <0.1 ML h−1
and was not practical for selective growth. This is
due to the strong passivation effect of the
(6.3×6.3)-Ga structure.
The samples were cut from n-type well-oriented
Si(111) wafer, whose misorientation was <0.5°.
After cleaning the sample, an ultrathin silicon-
dioxide film with a thickness of 0.3 nm was pro-
duced in the same way as previously reported [1].
The voids were formed during the thermal decom-
position of the oxide films at 720°C. During annea-
ling the sample, voids started to appear randomly
on the oxide surface. The density of the voids
increased and they extended isotropically with
increasing annealing time [8]. When the diameter
of voids reached 5–20 nm, the sample was
quenched to stop further extension of the voids.
The bottoms of the voids were flat and were
composed of Si(111)-(5×5) and Si(111)-(7×7)
In order to decrease the Ga coverage on the
ǰ
ǰ
oxide film and to form a ( 3× 3)-Ga structure,
the sample was annealed at >600°C. Fig. 1a shows
the images of Ga-adsorbed voids after annealing
at 650°C for 2 min. A straight line in this image
corresponds to an atomic step, and the height
difference between the oxidized Si(111) terraces is
a multiple of that of the silicon bilayers (0.31 nm).
In the terraces there are the Ga-adsorbed voids
with the diameter of 5–20 nm. The surfaces of the
voids are flat and there are no (6.3×6.3)-Ga
structures. The protrusions of the Ga droplets on
the oxide surface were considerably decreased.
These results shows that the evaporation of excess
Ga on the voids and the oxide caused the
(6.3×6.3)-Ga structure to disappear and a flat
reconstructions.
A
1/3 ML of Ga (2.6×
1014 cm−2) was deposited at 550°C and the sample
was annealed to form a stable structure of Ga-
adsorbed voids. Then Si was selectively grown by
ǰ
ǰ
( 3× 3)-Ga structure remained in the voids. In
recent work, Ga atoms on the oxide prove to be
easier to desorb than on Si surface [11]. Therefore
most Ga atoms on the oxide and some portion of
introducing disilane gas (Si H ) at 460–550°C. The
2 6
dose rate of disilane gas was fixed at

SURFACE SCIENCE
LETTERS
Y. Nitta et al. / Surface Science 431 (1999) L565–L569
L567
minutes. It is known that the number of Ga atoms
on a Si(111) surface decreases exponentially with
the annealing time [12]. According to this expo-
nential law, Si(111)-(7×7) structures are expected
to appear gradually with increasing annealing time.
However, in our experiment, small amount of Ga
atoms remained on the oxide and they are supplied
ǰ
ǰ
to the voids to form the stable ( 3× 3)-Ga
structure. In fact, once the voids of Si(111)-(7×7)
structure appeared after repeated annealing, the
number of the voids was rapidly increased with
annealing time. Because there were no Ga atoms
ǰ
ǰ
on the oxide and the ( 3× 3)-Ga structure could
not be formed in the voids.
ǰ
ǰ
After forming the voids of the ( 3× 3)-Ga
structure, disilane gas was introduced to the sample
in the growth temperature range of 460–550°C.
Fig. 2 shows sequential images of Si selective epi-
ǰ
ǰ
taxial growth on the Si(111)-( 3× 3)-Ga void
surface at 490°C. The images were taken (Fig. 2a)
before, (Fig. 2b) 15 min after, and (Fig. 2c) 56 min
after the Si growth had started. In Fig. 2a, all the
voids were composed of the flat structures before
ǰ
ǰ
the growth and the ( 3× 3)-Ga structures were
observed at higher magnification. In Fig. 2b, the
growth starts from the edge of the voids and the
islands were grown in a layer-by-layer fashion.
:
Step edges of the growth are parallel to the ꢀ110ꢁ
direction. It is reported that typical two-dimen-
sional nucleation takes place only in the middle of
the terraces in the Si epitaxial growth on the
ǰ
ǰ
Si(111)-( 3× 3)-Ga surface using disilane gas
[7]. In our case, however, the area in which the Si
growth is allowed to proceed is likely to be smaller
than the diffusion length of Si hydride dissociated
Fig. 1. (a) STM image of voids, Ga-adsorbed at 550°C and
annealed at 650°C. (b) Magnified STM image of the void.
ǰ
from disilane molecules on the Si(111)- 3-Ga
ǰ
ǰ
(
3× 3)-Ga reconstruction is observed in the voids.
surface. Although there are some point defects in
the middle of the void as shown in Fig. 1b, there
may be more defective regions at the edge of the
void where Ga coverage is not perfect. These
regions are likely to be nucleation sites. Therefore
Si hydride is supposed to migrate in the void and
to be trapped at the edge. Ga-adsorbed voids are
covered with one bilayer of Si after ca 50 min, and
the second layer starts to grow at smaller voids in
Fig. 2c. It is expected that the nucleation rate is
Ga atoms in the (6.3×6.3)-Ga voids seemed to
desorb by the annealing. Fig. 1b shows the magni-
fied image of Ga-adsorbed void. The void with
the diameter of ca 10 nm is plugged with
ǰ
ǰ
a ( 3× 3)-Ga structure in whole. Most of
the edges of the Ga-adsorbed areas are seen to
: :
be parallel to the ꢀ112ꢁ direction because of
its tendency to align in this direction. This
ǰ
ǰ
ǰ
( 3× 3)-Ga structures in the voids did not
change after repeated annealing at 650°C for a few
proportional to the circumference, that is, 3 A,
where A is the area of the voids and the layer-by-

LETTERS
SURFACE SCIENCE
L568
Y. Nitta et al. / Surface Science 431 (1999) L565–L569
layer growth rate in the depth direction is propor-
tional to the reciprocal of areas (1/A). Therefore
the total growth rate is roughly proportional to
ǰ
1/ A and the second layer is thought to grow
faster at the smaller voids.
To investigate the effect of Ga in the voids, the
selective growth in the voids with and without Ga
were examined at 550°C. In the case without Ga,
the growth proceeded in a layer-by-layer fashion
like the case with Ga. But the nucleation occurred
both at the edges and in the middle of the voids
and the growth rate of Si was ca 2.5 times higher
than that in Ga-adsorbed voids because Si dangling
bonds were not terminated.
After the growth in the voids of the (Si(111)-
3)-Ga structure, the surface was found to be flat
ǰ
but no reconstruction was observed. Most of the
Ga atoms were expected to segregate on the sur-
face, but the growth temperature was not high
enough for the reconstruction [13]. Then the
sample was annealed at 600°C in order to confirm
the existence of Ga atoms. Fig. 3 shows the void
surface after selective growth at 550°C and subse-
quent annealing at 600°C. About five bilayers of
Si were grown only in the voids after 183 min. In
Fig. 3a, some islands are seen to be much higher
than those in other voids. This may be due to Ga
atom diffusion in these voids to the outside of the
voids during the growth. The growth rate was
accelerated, since the Ga atoms on the Si(111)
surface are known to reduce the dissociation of
disilane. In Fig. 3b, the grown island is almost
isotropic though there are a few bunched steps at
the sidewall of the island. In this image the top
ǰ
ǰ
surface of the island shows a ( 3× 3) structure.
It is likely that some portion of Ga atoms were
incorporated in the Si layer but most of them
segregated during the growth and were recon-
structed on the top surface by annealing. The
temperature dependence of the Ga-incorporation
coefficient has been previously reported [14]. By
extrapolating the relationship, the Ga concen-
tration incorporated in the grown layer at 460–
550°C is estimated to be 2–5×1018 cm−3. This
concentration means that a few Ga atoms are
incorporated in the central Si island shown in
Fig. 3b. If the higher concentration is required, the
growth temperature must be lowered.
Fig. 2. Sequential STM images of selective epitaxial growth on
Ga-adsorbed voids on Si(111) at 490°C. The dose rate of disi-
lane is 2×1014 s−1cm−2. Images were acquired: (a) before; (b)
16 min after; and (c) 56 min after the start of growth.

SURFACE SCIENCE
LETTERS
Y. Nitta et al. / Surface Science 431 (1999) L565–L569
L569
selective growth are likely to segregate to the top
surface. These results show that Ga-doped dots
can be formed by selective epitaxial growth by
using an ultrathin silicon-dioxide mask.
Nanometer-scale windows have recently been
fabricated by a focused electron beam [15] or a
field-emission electron beam from an STM tip
[16,17]. If dopant-adsorbed voids are formed by
these techniques, nanometer-scale dots and wires
can also be formed at predetermined areas.
Acknowledgements
This work partly supported by NEDO, was
performed at JRCAT under the joint research
agreement between NAIR and ATP.
References
[1] K. Fujita, H. Watanabe, M. Ichikawa, Appl. Phys. Lett.
70 (1997) 2807.
[2] A.A. Baski, J. Nogami, C.F. Quate, J. Vac. Sci. Technol.
A 8 (1990) 245.
[3] Y. Enta, S. Suzuki, S. Kono, Surf. Sci. 242 (1991) 277.
[4] A. Kawazu, H. Sakama, Phys. Rev. B 37 (1988) 2704.
[5] J.R. Patel, J. Zegenhagen, P.E. Freeland, M.S. Hybertsen,
J.A. Golovchenko, D.M. Chen, J. Vac. Sci. Technol. B 7
(1989) 894.
[6] J. Zegenhagen, E. Artacho, P.E. Freeland, J.R. Patel, Phil-
osophical Magazine B 70 (1994) 731.
[7] K. Fujita, Y. Kusumi, M. Ichikawa, Surf. Sci. 380
(1997) 66.
[8] K. Fujita, H. Watanabe, M. Ichikawa, J. Appl. Phys. 83
(1998) 4091.
[9] J.L. Alay, M. Fukuda, C.H. Bjorkman, K. Nakagawa, S.
Yokoyama, S. Sasaki, M. Hirose, Jpn. J. Appl. Phys., Part
1 1 (34) (1995) L653.
[10] K. Fujita, H. Watanabe, M. Ichikawa, J. Appl. Phys. 83
(1998) 3638.
Fig. 3. (a) STM image after selective growth at 550°C and subse-
quent annealing at 600°C. About five bilayers of Si have selec-
tively grown in the voids. (b) The magnified image of growth
ǰ
ǰ
area identified by the arrow in (a). A ( 3× 3)-Ga reconstruc-
[11] M. Shibata, S. Stoyanov, M. Ichikawa, Phys. Rev. B 59
(1999) 10289.
tion is observed on the grown area.
[12] M. Zinke-Allmang, L.C. Feldman, Surf. Sci. 191 (1987)
L749.
[13] H. Nakahara, M. Ichikawa, Surf. Sci. 298 (1993) 440.
[14] K. Nakagawa, M. Miyao, Thin Solid Films 183 (1989) 315.
[15] S. Fujita, S. Maruno, H. Watanabe, M. Ichikawa, J. Vac.
Sci. Technol. A 15 (1997) 1493.
[16] M. Shibata, Y. Nitta, K. Fujita, M. Ichikawa, Appl. Phys.
Lett. 73 (1998) 2179.
[17] N. Li, T. Yoshinobu, H. Iwasaki, Jpn. J. Appl. Phys. 37
(1998) L995.
In summary, Si selective growth on Ga-
adsorbed voids was performed on Si(111) surfaces
ǰ
ǰ
by using disilane gas. The ( 3× 3) structure was
formed in the voids with diameters of 5–20 nm
after suitable annealing conditions. Si was selec-
tively grown only in voids in a layer-by-layer
fashion and nucleation starts from the edge of the
voids. Ga atoms partly incorporated during the