Page 1
Three-dimensional islands of Si and Ge formed on SiO2 throughcrystallization and agglomeration from amorphous thin ®lms
Yutaka Wakayama*, Takashi Tagami1, Shun-ichiro Tanaka
Tanaka Solid Junction Project, ERATO, Japan Science and Technology Corporation, 1-1-1 Fukuura, Kanazawa-ku, Yokohama 236-0004, Japan
Received 1 March 1999; received in revised form 15 April 1999; accepted 15 April 1999
Abstract
The crystallization process was examined for amorphous thin ®lms of silicon (a-Si) and germanium (a-Ge) on quartz glass (SiO2)
substrate. Three-dimensional crystalline islands were formed through crystallization and agglomeration. These islands indicated a bimodal
size distribution. The mechanism of crystalline island (c-Si, c-Ge) formation was discussed on the basis of thermodynamics. In studying the
crystallization of the thin ®lms, the in¯uence of the ®lm-substrate interfacial energy should be taken into consideration. It was found that the
thickness of the as-deposited amorphous ®lms is an essential factor in determining the crystallization behavior and in controlling island size.
Above all, a high size uniformity of crystalline islands could be obtained under moderate thermal annealing conditions. q 1999 Elsevier
Science S.A. All rights reserved.
Keywords: Crystallization; Agglomeration; Interface energy; Silicon; Germanium
1. Introduction
In this study, we deal with amorphous ®lms ranging from
1 to 10 nm in thickness with a bare surface, deposited on
SiO2 substrates. Our main purpose is to investigate their
crystal growth process and to examine the possibility of
controlling of the size of crystalline islands on a nanometer
scale.
Amorphous semiconductor thin ®lms have been widely
used in such devices as a thin ®lm transistor and a solar cell.
Their crystallization processes have also been studied exten-
sively from a fundamental scienti®c viewpoint [1,2] On the
other hand, the discovery of visible light emission from
porous silicon [3], even though it has an indirect band
gap, has created a new aspect of semiconductor science. A
large number of techniques for Si nanoparticle formation
have been reported, e.g., chemical vapor deposition [4±7],
laser ablation [8,9] and ion implantation [10,11]. For these
cases, comprehension of crystal growth on a nanometer
scale is very important.
Regarding the fabrication of nanometer-scale materials,
recent research has focused on a self-assembly process. For
Ge deposition on the Si substrate, a high mismatch of lattice
constants induces three-dimensional island formation at the
initial stage of epitaxial growth, that is the Stranski±Kras-
tanov growth mode. Much effort has been made to under-
stand the growth kinetics of the Ge dots theoretically [12±
14] and experimentally [15±20]. Generally, molecular beam
epitaxy (MBE) or metalorganic chemical vapor deposition
(MOCVD) are employed for dot formation. One problem
which must be solved is the control of the size of the dots for
future device application. In these techniques, the dif®culty
in size control arises because gas-phase atoms are supplied
to form solid-state dots. Accordingly, some experimental
parameters, such as the substrate temperature or the deposi-
tion rate, should be varied to accurately control highly
mobile atoms.
Here, we demonstrate a solid-state process for growing
the crystalline islands of Si and Ge on SiO2 substrates.
Crystalline islands are formed through crystallization and
agglomeration from amorphous `thin' ®lms, the surface of
which is clean. The mechanism of crystal growth is quite
different from that of amorphous `thick' ®lms. Discussion
based on thermodynamics is also given to explain this
process.
Thin Solid Films 350 (1999) 300±307
0040-6090/99/$ - see front matter q 1999 Elsevier Science S.A. All rights reserved.
PII: S0040-6090(99)00294-1
* Corresponding author. Present address: Max-Planck Institute of Micro-
structure Physics, Weinberg 2, D-06120 Halle, Germany. Tel.: 149-345-
5582-50; fax: 181-345-5511-223.
E-mail address: [email protected] (Y. Wakayama)1 Present address: Tsukuba Research Center, Technical Research
Laboratory, Nippon Sheet Glass Co. Ltd., 5-4 Tokodai, Tsukuba, Ibaraki
300-26, Japan
Page 2
2. Experimental conditions
Quartz glass (SiO2) substrates 1 cm2 in size were used for
Si and Ge deposition. A clean substrate surface is strictly
required for precise discussion of the crystallization process
of a-Si and a-Ge thin ®lms. Hence, the cleaning of the
substrate should be carried out carefully as follows. The
mirror-polished SiO2 substrates were chemically cleaned
by washing with isopropyl alcohol and acetone for 15 min
each. After dipping in diluted HF solution (2 wt.%) for 20 s,
they were rinsed in running ultrapure water for 10 min. The
substrates thus cleaned were introduced into a high-vacuum
chamber with a background pressure of 1 £ 1028 Torr.
Then, the substrates were baked at 6008C and maintained
at that temperature for 3 h to remove hydrocarbon contam-
ination and to obtain a clean surface.
Silicon and Ge were deposited on the SiO2 substrates at
room temperature by electron-beam evaporation. The typi-
cal deposition rates were 0.018 nm/s for Si and 0.023 nm/s
for Ge. The thickness of the as-deposited ®lms, denoted as
initial thickness di, was varied from 1 to 10 nm for Si and
was constant at 10 nm for Ge. The structure of the as-depos-
ited ®lms was con®rmed to be amorphous from transmission
electron microscope (TEM) images and electron diffraction
(ED) patterns.
Thermal annealing was performed immediately after the
®lm deposition in high vacuum for crystallization. The
temperature of the specimens was elevated at a rate of
158C/min to 550±6008C for Si and 275±3258C for Ge,
respectively. For both cases, the specimens were held at
each temperature for 30 min and cooled to room tempera-
ture.
The structure of specimens thus prepared was examined
by bright ®eld TEM (JEOL-2010) images those were taken
at an acceleration voltage of 200 kV. For plane-view TEM
observation, the SiO2 substrates were mechanically thinned
from the back surface to a thickness of less than 30 mm,
followed by Ar1 ion milling. A tapping-mode atomic force
microscope (AFM, Dimension 5000, Digital Instruments)
was used for morphological investigation after thermal
annealing.
3. Results and discussion
3.1. Formation of crystalline Si island with bimodal size
distribution
Fig. 1 show the plane-view TEM images of Si ®lms which
were annealed at 6008C. Their initial thicknesses, di, were
(a) 10, (b) 5 and (c) 1 nm. As can be seen here, round
crystallites were formed randomly on the substrate and
their mean size decreased with di. Meanwhile, the number
density of the Si crystallites grew with decreasing di. The
average radius, Ra, and number density of crystallites, D, are
plotted as a function of the initial thickness di in Fig. 2.
A cross-sectional TEM image of postannealed Si with di
of 10 nm is shown in Fig. 3. The image reveals that the Si
crystallites are hemispherical and disconnected from each
other. The height of the islands is about 30 nm and is greater
than di. Namely, three-dimensional (3D) crystalline Si
islands swelled up from two-dimensional (2D) amorphous
thin ®lms through crystallization by thermal annealing.
These results imply that the size of the islands is controlla-
ble by modifying the thickness of as-deposited amorphous
thin ®lms.
The histograms of Si island size are shown in Fig. 4, their
di being (a) 10, (b) 5 and (c) 1 nm. For all cases, the size of
more than 300 islands were measured to obtain these histo-
grams and to discuss the size distribution statistically. It is
clear that the size of the Si islands is widely distributed for
each case. For instance, the radius ranged from 12 nm to 36
Y. Wakayama et al. / Thin Solid Films 350 (1999) 300±307 301
Fig. 1. Plane-view TEM images of Si ®lms annealed at 6008C. The initial
thicknesses of the ®lms, di, are (a) 10, (b) 5 and (c) 1 nm.
Page 3
nm and standard deviation was determined to be 38% of the
average radius of 24 nm for di � 10 nm.
It is worth emphasizing that the histograms indicate
bimodal size distribution. Dotted lines are drawn in these
®gures to clarify the size distribution. They consist of two
islands groups: small islands ranging from 12 to 28 nm in
radius and large islands ranging from 28 to 36 nm in radius,
at a rough estimate, for di � 10 nm.
To understand the origin of the bimodal size distribution,
the crystallization process of the a-Si ®lms should be exam-
ined in greater detail. For this purpose, a-Si ®lms were
thermally annealed at 550±6008C. Fig. 5 shows plane-
view TEM images. The histograms of Si islands annealed
at 575 and 6008C are shown in Fig. 6. The initial thickness
of these specimens was 10 nm. For comparison, the results
of the specimen annealed at 6008C are also shown in Figs.
5c and 6b, duplicated from Figs. 1a and 4a, respectively.
From the TEM images, it was found that round crystalline
islands, indicated by arrows in Fig. 5a, had started to form
on the substrate at an annealing temperature of 5508C. The
radius of the islands is about 30 nm, which coincides with
that of large islands mentioned above. The number of such
large islands increases on raising the annealing temperature
to 5758C. Thus, the crystallization of a-Si thin ®lms takes
place through the formation of disconnected crystalline Si
islands. Up to this stage of crystallization, the size of the
islands is relatively uniform, as shown in Fig. 6a. The aver-
age radius was determined to be 31 nm and the standard
deviation was found to be reduced to 13% of the average
radius.
Compared to this size uniformity of the islands, the speci-
men annealed at 6008C, which was completely crystallized,
indicates a wide size distribution of the islands. Here, it
should be noted that high annealing temperatures led to
the formation of relatively small islands but not large
ones. Some typical islands are indicated by arrows in Fig.
Y. Wakayama et al. / Thin Solid Films 350 (1999) 300±307302
Fig. 3. Cross-sectional TEM image of postannealed Si with di � 10 nm.
Hemispheric islands are observed. Their height of about 30 nm is greater
than di.
Fig. 2. Average radius, Ra, and number density of crystalline Si islands, D,
plotted as a function of initial thickness di.
Fig. 4. Size distribution of Si islands shown in Fig. 1, their di being (a) 10,
(b) 5 and (c) 1 nm. All histograms show bimodal size distribution: large
islands with narrow distribution and small islands with broad distribution
are represented by dotted lines.
Page 4
5c. The upper limit of the island size, 36 nm, remained
constant for every thermal annealing condition for the speci-
men of di � 10 nm. No increase in the size of island was
observed, even if the annealing temperature was increased
to 7008C.
Here, we classify the Si islands into two groups: islands-
A whose size is large and has a narrow distribution and
islands-B whose size is small and has a broad distribution.
All of the histograms in Fig. 4 can be interpreted as an
overlap of the distribution of both islands-A and B.
Careful observation of Fig. 5a,b revealed that the Si
islands were surrounded by bright contrast, though such a
contrast was not observed in Fig. 5c. Fig. 7 exhibits a 3D
AFM image of a c-Si island. This image clearly demon-
strated that a ditch encircled the Si islands, which resulted
in the bright contrast in the TEM images. The depth of the
ditch roughly agreed with the initial thickness of the amor-
phous Si ®lm.
3.2. Thermodynamic considerations
On the basis of these experimental results, we consider
the crystallization process to be as follows. Change in the
free energy of the Si thin ®lm, DG1, can be described by
Eq. (1). To simplify the discussion, the shape of the crys-
tallites is assumed to be cylindrical with radius r and thick-
Y. Wakayama et al. / Thin Solid Films 350 (1999) 300±307 303
Fig. 5. Plane-view TEM images of postannealed Si ®lms of 10 nm initial
thickness: (a) 5508C, crystallization begins forming round islands as indi-
cated by arrows; (b) 5758C, the crystal growth proceeds, increasing the
number of islands and keeping their radius constant, relatively small islands
can be observed only near the large islands as indicated by the arrow, a
bright contrast is observed around the Si islands as shown in square; (c)
6008C, Crystallization is completed through the formation of relatively
small islands, as indicated by arrows.
Fig. 6. Size distribution of Si islands shown in Fig. 5b,c, annealed at (a) 575
and (b) 6008C. The average radius, Ra, is shown in the graphs. Lower
temperature produced larger islands with narrower size distribution. Higher
temperature gave rise to the formation of smaller islands in addition to
larger ones resulting in a wide size distribution. The two island groups
are denoted as islands-A and B, respectively.
Page 5
ness d.
DG1 � pr2dDHac 1 2prdsac 1 pr2 sco 2 sao
ÿ �1 pr2 sa 2 sc
ÿ � �1�Here, DHac is the free energy of the amorphous-crystal
phase transition per unit volume, s ac, s co and s ao are the
interface energies of a-Si/c-Si, c-Si/SiO2 and a-Si/SiO2 per
unit area, and s a and s c represent the surface free energies
of a-Si and c-Si per unit area, respectively.
DG1 varies markedly depending on whether the value
of {DHac 1 �sco 2 sao�1 �da 2 sc�} is positive or nega-
tive. In other words, the crystallization behavior de-
pends strongly on ®lm thickness d. If
d . 2{�sco 2 sao�1 �sa 2 sc�}=DHac, DG1 increases
once at the beginning of the crystallization until a certain
radius is reached, which is expressed in Eq. (2) where
�DG=�r � 0. Continuous crystal growth occurs afterward,
decreasing the value of DG1.
r � 2dsac
dDHac 1 sco 2 sao
ÿ �1 sa 2 sc
ÿ � �2�
On the other hand, DG1 increases consistently with an
increase in the crystallite radius if
d , 2{�sco 2 sao�1 �sa 2 sc�}=DHac. Then, the Si/SiO2
system becomes unstable as crystal growth progresses.
For the Si/SiO2 system, a negative value of DHac (, 2103 J/cm3) [21] ®rst brings about an amorphous-crystal
phase transition. On the other hand, the interface energy
increases as the crystallization proceeds because both s ac
(,3 £ 1025 J/cm2) [22,23] and sco 2 sao (,1023 J/cm2)
[24] have positive values. Then, d ù 10 nm is the critical
thickness to determine the variation in DG1. We assumed
here that the change in the surface free energy through
crystallization, that is sa 2 sc, also has a positive value
and is of the same order of magnitude as s ac (,1025 J/
cm2). The initial thickness of a-Si, di, ranges from 1 to 10
nm in our experiments. Therefore, DG1 increases as the
crystallization proceeds, making the Si/SiO2 system
unstable.
Our experimental results can be interpreted as follows:
crystal growth halted to prevent a further increase in DG1,
and the cylindrical crystallites agglomerate to form hemi-
spheric islands, those are islands-A, in order to minimize the
total free energy of the Si/SiO2 system. The ditch observed
in the AFM image indicates clearly a trace of agglomeration
from the cylindrical crystallites to the hemispheric island.
This is the reason for the presence of the upper limit of
island size. The size uniformity of the islands at an early
stage of crystal growth can also be explained by the same
reasoning. Namely, the halt in the crystal growth occurred at
a certain radius rc. The critical radius rc can be observed as a
bright contrast around Si islands in the TEM images. The
value of rc was determined to be about 43 nm on average
from Fig. 5b, which is 35% larger than the radius of the Si
islands.
Crystallization was completed at a high annealing
temperature of 6008C. Then, the islands formed subse-
quently, those are islands-B, from residual a-Si among
islands-A. The growth of islands-B is not limited principally
within the area of the residual a-Si ®lm. As a result, their
size is smaller and their size distribution is broader than
those of islands-A. The average size of the whole island,
including island-A and B, was actually reduced from 31 to
24 nm. The islands-B, of course, can be formed even at
5758C if crystal nucleation starts near islands-A. One of
the typical example was indicated by arrow in Fig. 5b. Rela-
tively small island, that is islands-B, was formed just near
islands-A.
This also will be shown for the case of the Ge/SiO2
system later. Nevertheless, the size uniformity of the islands
can be maintained as long as the number density of the
islands is suf®ciently low. The crystallization process
described here is illustrated in Fig. 8a±c.
The free energy change DG1 shows a parabolic curve with
radius r. The value of DG1 tends to increase sharply with
decreasing di, as drawn in Fig. 9a. Therefore, the thinner di
resulted in the smaller rc. The critical radius rc can be
expressed as follows
rc ����������������������������������
DGc
p di´DHac 1 sco 2 sac
ÿ �� s�3�
Here, DGc is the critical free energy, at which the agglom-
eration of Si islands occurred. For instance, DGc was esti-
mated to be of the order of 10215 J for the case of di � 10
nm, and then rc � 43 nm. For Eq. (3), we neglected the
terms s ac and sa 2 sc because these values are two orders
of magnitude less than those of DHac and so 2 sao.
Fig. 9b shows the relation between di and the average
radius of islands-A, ra. Assuming that the value of DGc is
constant regardless of di, ra as well as rc will be proportional
top
(1/di), according to Eq. (3). Our results almost support
Y. Wakayama et al. / Thin Solid Films 350 (1999) 300±307304
Fig. 7. AFM image of c-Si islands formed by thermal annealing at 5758C. A
ditch is observed around the c-Si as a trace of the agglomeration.
Page 6
this discussion, although the ®gure shows a slight curve as
indicated by the dotted line. The reason for the curvature
above di of 5 nm (belowp
(1/di) of 0.5 nm21/2) can be
explained as follows. It has been reported that a critical
size exist to induce the amorphous-crystal phase transition
for a-Si [2,25]. The critical diameter was estimated to be 2±
3 nm. Therefore, it is reasonable, for di below 3 nm, to
regard the crystallites as cylindrical from the beginning of
the crystallization. Then, change in free energy can be
described as Eq. (1), as mentioned above. For thick di
however, crystallites are spherical, to be exact, at the begin-
ning of the crystallization [24]. Then, the free energy change
accompanying the crystallization is described as
DG0 � �4=3�pr3DHac 1 4pr2sac �4�
The crystallite can not be regarded as cylindrical until the
crystallite contacts the substrate. As a result, the critical
radius, rc, becomes large for thick di. This two-step crystal-
lization process is the reason that ra tends to have large value
for thick di (smallp
(1/di)), as can be seen in Fig. 9b. This
crystallization process and the change in the free energies
are schematically illustrated in Fig. 10.
The critical radius is dependent only on the thickness of
the a-Si ®lms, di, as discussed here. It is independent of
other experimental parameters because the values of the
phase transition, surface and interface energies are peculiar
to a material system. These ®ndings suggest that the island
size can be controlled by modifying only the thickness of
the as-deposited ®lm.
For the case of the sandwiched structure of SiO2/a-Si/
SiO2 [22,24,26], crystallite growth has been reported to
halt after reaching a certain size mainly because the increase
in the Si/SiO2 interface energy, sco 2 sao, exceeds the
decrease in the a-Si/c-Si phase transition energy. Similarly,
a strong in¯uence of the interface energy on the crystalliza-
tion process has been reported for multilayered structures,
such as a-Ge/a-GeN [27] and a-Si/a-SiNx [28]. For these
multilayered structures, the ®lms remain 2D and continuous
even after crystallization. In contrast, the surface of a-Si thin
®lm is bare in the present experiments and there are no
capped layers on top of the ®lm. Accordingly, the free
surfaces permits the Si crystallites, which reach the critical
radius rc, to agglomerate by way of the reduction of the total
free energy, forming hemispheric islands.
Sakai et al. also reported the crystallization process of a-
Si ®lms with a clean surface [29]. In their study, they also
observed the ditch around Si crystal grains. They attributed
such a ditch formation to the surface diffusion of Si atoms
on the bare surface. However, the mechanism of the Si
island formation described here should be interpreted differ-
Y. Wakayama et al. / Thin Solid Films 350 (1999) 300±307 305
Fig. 8. Crystallization process of the a-Si thin ®lm on SiO2. (a) In the ®rst
stage, the cylindrical crystallites grow until they reach a critical radius rc.
(b) The crystallites with rc agglomerate, forming unconnected c-Si islands,
islands-A, in the second stage. (c) Finally, annealing at a higher tempera-
ture completes the entire crystallization process, forming small islands,
islands-B, among islands-A. Fig. 9. (a) Change in DG1 with various di are drawn as a function of the
radius of the cylindrical crystalline Si, r. The crystallites which reach rc and
DGc agglomerate to prevent a further increase in free energy. (b) Average
radius of islands-A, ra, is plotted as a function ofp
(1/di)
Page 7
ently for the following reasons. First, the thickness of the a-
Si ®lms was 50 nm in their experiment, which was suf®-
ciently larger than the value of 2{�sco 2 sao�1 �sa 2sc�}=DHac ( ù 10 nm). Therefore, it was not necessary to
take the in¯uence of the Si/SiO2 interface on the crystal-
lization process into consideration. Additionally, if the
surface diffusion brings about the ditch formation, the island
size should depend on the annealing temperature. That is, a
high temperature leads to large diffusion length, which
should, in turn, lead to the formation of large islands.
However, an upper limit of the Si island size exists and is
independent of the annealing temperature in our experi-
ments. Furthermore, the surface diffusion mechanism can
not explain the size uniformity of islands-A. Consequently,
we concluded that the agglomeration mechanism holds for
the formation of Si islands.
3.3. Formation of crystalline Ge island on SiO2
Principally, for any ®lm/substrate system which has a
positive value of sco 2 sao and satis®es the relation
d , 2{�sco 2 sao�1 �sa 2 sc�}=DHac, 3D crystalline
islands can be formed from amorphous ®lms through crys-
tallization and agglomeration. We examined the crystalliza-
tion of a-Ge thin ®lms on SiO2 substrate in the same manner.
Fig. 11 shows plane-view TEM images of Ge ®lms with di
of 10 nm, which were annealed at (a) 275, (b) 300 and (c)
3258C. The crystallization process can be likened to that of
the a-Si/SiO2 system as follows. Crystalline Ge islands of
about 20 nm in radius start to form at 2758C, and these
Y. Wakayama et al. / Thin Solid Films 350 (1999) 300±307306
Fig. 11. Plane-view TEM images of postannealed Ge ®lms of 10 nm initial
thickness: (a) 2758C, traces of detached islands are observed; (b) 3008C,
island-B is formed as indicated by an arrow even in this crystallization
stage when the nucleation starts near island-A; (c) 3258C, the size distribu-
tion of (c) is also shown in (d).
Fig. 10. (a) The crystallite is spherical when radius r is less than di/2. Then,
the free energy change follows Eq. (3). The crystallite can be regarded as
cylindrical if r exceeds di/2. Then, the free energy change follows Eq. (1).
(b) Change in the free energy accompanying the crystallization process.
The thicker di is, the larger rc becomes compared with those expected in
Fig. 9a.
Page 8
correspond to islands-A for Si. Relatively small islands,
islands-B, were observed only among or near islands-A,
as indicated by arrows in Fig. 11b. Crystallization was
completed at 3258C, as shown in Fig. 11c. Then, the size
distribution was bimodal, as shown in Fig. 11d.
The average size of Ge islands was smaller than that of Si
islands. This may be because the change in the interfacial
energy through crystallization, that is sco 2 sao, of Ge/SiO2
is greater than that of the Si/SiO2 system. Moreover, traces
showing that the Ge islands detached form the surface were
often observed for the Ge/SiO2 system, as indicated by
arrows in Fig. 11a. Some of the c-Ge islands were consid-
ered to detach from the surface during specimen preparation
for TEM observation. We assume that this is likewise due to
the increase in the interfacial energy through crystallization.
Recently, self-assembly techniques using gas-phase
deposition methods have been regarded as promising candi-
dates for the fabrication of quantum-dots. However, ambig-
uous factors, such as the diffusion length of highly mobile
atoms supplied to the substrate or the nucleation ratio of
islands, make it dif®cult to design nanometer-scale materi-
als precisely. In particular, the broad size distribution of the
quantum-dots is a crucial problem. On the other hand, the
solid-state process mentioned here provides another
approach to achieve size uniformity of semiconductor dots
on a nanometer scale.
4. Summary
We investigated the crystal growth of amorphous thin
®lms with a clean surface on the SiO2 substrate. It was
emphasized that the thickness of the amorphous ®lms is
the dominant factor affecting the crystallization process.
The crystalline islands examined here indicated bimodal
size distribution. This is a result of two-step crystallization
process as follows. First, the agglomeration of the crystal-
lites was triggered after they reached the critical radius rc to
minimize the total energy. At this step, the islands have
basically a uniform radius. Next, crystallization was
completed by further thermal annealing, forming small
islands with a broad size distribution. The main point is
that a high size uniformity of the islands could be obtained
under moderate thermal annealing conditions by taking
advantage of the in¯uence of the Si/SiO2 interface.
References
[1] J.L. Batstone, Philos. Mag. A 67 (1993) 51.
[2] H. Hofmeister, J. Dutta, H. Hofmann, Phys. Rev. B 54 (1996) 2856.
[3] L.T. Canham, Appl. Phys. Lett. 57 (1990) 1046.
[4] F.M. Timofeev, A. Aydinli, R. Ellialtioglu, K. Turkoglu, M. Gure,
V.N. Mikhailov, O.A. Lavrova, Solid State Commun. 95 (1995) 443.
[5] A.J. Kenyon, P.F. Trwoga, C.W. Pitt, G. Rehn, J. Appl. Phys. 79
(1996) 9291.
[6] B.J. Hinds, A. Banerjee, R.S. Johnson, G. Lucovsky, Mater. Res. Soc.
Symp. Proc. 452 (1997) 207.
[7] T. Inokuma, Y. Wakayama, T. Muramoto, R. Aoki, Y. Kurata, S.
Hasegawa, J. Appl. Phys. 83 (1998) 2228.
[8] T. Yoshida, S. Takeyama, Y. Yamada, K. Mutoh, Appl. Phys. Lett. 68
(1996) 1772.
[9] T. Makimura, Y. Kunii, K. Murakami, Jpn. J. Appl. Phys. 94 (1996)
35.
[10] T. Shimizu-Iwayama, S. Nakao, K. Saitoh, Appl. Phys. Lett. 65
(1814) (1994) 94.
[11] P. Mutti, G. Ghislotti, S. Bertoni, et al., Appl. Phys. Lett. 66 (1994)
851.
[12] Y. Chen, J. Washburn, Phys. Rev. Lett. 77 (1996) 4046.
[13] F. Liu, J. Tersoff, M.G. Lagally, Phys. Rev. Lett. 80 (1998) 1268.
[14] H.T. Dobbs, D.D. Vvedensky, A. Zagwill, Appl. Surf. Sci. 123/124
(1998) 646.
[15] F.K. LeGoues, M.C. Reuter, J. Tersoff, M. Hammar, R.M. Tromp,
Phys. Rev. Lett. 73 (1994) 300.
[16] G. Abstreiter, P. Schittenhelm, C. Engel, E. Silveira, A. Zrenner, D.
Meertens, W. JaÈger, Semicond. Sci. Technol. 11 (1996) 1521.
[17] T.I. Kamins, E.C. Carr, R.S. Williams, S.J. Rosner, J. Appl. Phys. 81
(1997) 211.
[18] M. Goryll, L. Vescan, K. Schmidt, S. Mesters, H. LuÈth, Appl. Phys.
Lett. 71 (1997) 410.
[19] G. Medeiros-Ribeiro, A.M. Bratkovski, T.I. Kamins, D.A.A. Ohlberg,
R.S. Williams, Science 279 (1998) 353.
[20] F.M. Ross, J. Tersoff, R.M. Tromp, Phys. Rev. Lett. 80 (1998) 984.
[21] E. Donovan, F. Saepan, D. Turnbull, J. Appl. Phys. 57 (1985) 1795.
[22] P.D. Persans, A. Ruppert, B. Abeles, J. Non-Cryst. Solids 102 (1988)
130.
[23] F. Spaepen, Acta Metall. 26 (1978) 1167.
[24] T. Tagami, Y. Wakayama, S.-I. Tanaka, Jpn. J. Appl. Phys. 36 (1997)
L734.
[25] S. Veprek, Z. Igbal, F.A. Sarott, Philos. Mag. B 45 (1982) 137.
[26] H. Freistedt, F. Stolze, M. Zacharias, J. Blasing, T.P. Drusedau, Phys.
Stat. Solidi B 193 (1996) 375.
[27] I. Honma, H. Komiyama, K. Tanaka, J. Appl. Phys. 66 (1989) 1170.
[28] J. Dutta, I.M. Reaney, P. Roca, i Cabarrocas, H. Hofmann, NanoS-
tructured Mater. 6 (1995) 843.
[29] A. Sakai, H. Ono, K. Ishida, T. Niino, T. Tastumi, Jpn. J. Appl. Phys.
30 (1991) L941.
Y. Wakayama et al. / Thin Solid Films 350 (1999) 300±307 307