Effect of indium-doped interlayer on the strain relief in GaN films grown on Si(111)

Article abstract of DOI:10.1002/pssa.200723162

Crack-free GaN films have been achieved by inserting an In-doped low-temperature (LT) AlGaN interlayer grown on silicon by metalorganic chemical vapor deposition. The relationship between lattice constants c and a obtained by X-ray diffraction analysis shows that indium doping interlayer can reduce the stress in GaN layers. The stress in GaN decreases with increasing trimethylindium (TMIn) during interlayer growth. Moreover, for a smaller TMIn flow, the stress in GaN decreases dramatically when In acts as a surfactant to improve the crystallinity of the AlGaN interlayer, and for a larger TMIn flow, the stress will increase again. The decreased stress leads to smoother surfaces and fewer cracks for GaN layers by using an In-doped interlayer than by using an un-doped interlayer. In doping has been found to enhance the lateral growth and reduce the growth rate of the c face. It can explain the strain relief and cracks reduction in GaN films.

Full text of DOI:10.1002/pssa.200723162

phys. stat. sol. (a) 205, No. 2, 294–299 (2008) / DOI 10.1002/pssa.200723162
a
p s s
www.pss-a.com
applications and materials science
Effect of indium-doped interlayer
on the strain relief in GaN films
grown on Si(111)
Jiejun Wu*, 1, Lubing Zhao1, Guoyi Zhang1, Xianglin Liu2, Qinsheng Zhu2, Zhanguo Wang2, Quanjie Jia3,
Liping Guo3, and Tiandou Hu3
1
Research Center for Wide-gap Semiconductors, State Key Laboratory for Artificial Microstructures and Mesoscopic Physics,
School of Physics, Peking University, Beijing 100871, P.R. China
2
Key Laboratory of Semiconductor Materials Science, Institute of Semiconductors, Chinese Academy of Sciences, P.O. Box 912,
Beijing 100083, P.R. China
3
Institute of High Energy Physics, Chinese Academy of Science, Beijing 100039, P.R. China
Received 23 April 2007, revised 27 September 2007, accepted 15 November 2007
Published online 6 February 2008
PACS 68.37.Hk, 68.37.Ps, 68.55.–a, 78.55.Cr, 81.05.Ea, 81.15.Gh
*
Corrsponding author: e-mail jiejunw@red.semi.ac.cn or jiejunwu@pku.edu.cn
Crack-free GaN films have been achieved by inserting an In- decreases dramatically when In acts as a surfactant to im-
doped low-temperature (LT) AlGaN interlayer grown on sili- prove the crystallinity of the AlGaN interlayer, and for a lar-
con by metalorganic chemical vapor deposition. The relation- ger TMIn flow, the stress will increase again. The decreased
ship between lattice constants c and a obtained by X-ray dif- stress leads to smoother surfaces and fewer cracks for GaN
fraction analysis shows that indium doping interlayer can re- layers by using an In-doped interlayer than by using an un-
duce the stress in GaN layers. The stress in GaN decreases doped interlayer. In doping has been found to enhance the
with increasing trimethylindium (TMIn) during interlayer lateral growth and reduce the growth rate of the c face. It can
growth. Moreover, for a smaller TMIn flow, the stress in GaN explain the strain relief and cracks reduction in GaN films.
© 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
1 Introduction Epitaxy of GaN on silicon offers great firstly by Amano and coworkers [10] for the growth of
advantages: low cost, good thermal and electrical conduc- thick crack-free AlGaN on GaN, was employed by Dadgar
tivities, and the potential for monolithic integration of et al. [11] to reduce the crack density in the GaN layer
GaN-based devices with conventional microelectronics [1]. grown on Si substrate. It is found that the inplane lattice
However, early attempts to grow GaN on Si suffered from parameter of the LT-Al(Ga)N interlayer can be tuned to
two main problems: thermal and lattice mismatch causing control and compensate growth stress during subsequent
cracks even below device-relevant layer thickness, and the growth of the high-temperature (HT) GaN layers on
strong meltback etching leading to a deterioration of the Si(111) substrates [12]. Furthermore, several kinds of
GaN layer and the Si substrate. Therefore, low-temperature interlayers, including graded AlGaN interlayers [13],
(LT) GaN (as typically employed on sapphire) is not an ef- Al(Ga)N/GaN superlattices [14] and multi-interlayers [15],
fective buffer layer for Si substrate [2] and other buffer have been investigated to increase the critical thickness for
materials including AlN [3], SiC [4], AlAs [5], BP [6], and the appearance of cracks in GaN layers on Si substrates.
HfN [7] have been investigated. Meanwhile, several tech- For the metalorganic chemical vapor deposition (MOCVD)
niques, such as selective epitaxial growth (patterning) [8, method, one of the best results has been achieved by insert-
9] and inserting strain-relief interlayers [10–15], have been ing several layers of AlN/GaN superlattices. Its total criti-
employed to increase the critical thickness for the appear- cal thickness is increased to 3.8 µm while maintaining per-
ance of cracks in GaN layers grown on Si(111) substrates. fect quality [14]. However, multi-interlayers have to be
The low-temperature (LT) Al(Ga)N interlayer, introduced used and a superlattice interlayer increases the technique
© 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Original
Paper
phys. stat. sol. (a) 205, No. 2 (2008)
295
complexity. In our prior studies, a simple method has been
All samples are investigated by photoluminescence
applied for the growth of crack-free GaN layers by insert- (PL) measurements at 10 K using the 325 nm line of a
ing an isoelectronic In-doped LT-AlGaN interlayer using He–Cd laser, high-resolution X-ray diffraction measure-
MOCVD [16]. In doping of the LT-AlGaN interlayer can ments (HR-XRD), atomic force microscopy (AFM), and
lead to a reduced crack density as well as improved struc- scanning electron microscopy (SEM). The Al fraction of
tural and optical properties of GaN layers. In this paper, the LT-AlGaN interlayer is determined by energy disper-
indium is found to relieve the strains in the AlGaN inter- sive spectroscopy (EDS) and Rutherford backscattering
+
layer and the GaN layer. This benefits the growth of GaN spectrometry (RBS) using a 2.022 MeV He beam with a
on Si substrate.
Recently, isoelectronic In doping has been demon-
strated to serve as a useful surfactant for the growth of
backscattering angle of 165°.
3 Results and discussion Figure 1 shows the rela-
HT-GaN [17–19]. Experiments have shown that the pres- tionship between the stress in the GaN layers and the TMIn
ence of In leads to a smooth morphology [17] with high flow rate during the growth of LT-AlGaN interlayers. The
crystal quality [18] and good optical properties [20, 21]. In residual stress σ in GaN epilayers is calculated according
xx
addition, calculations have shown that In termination is the following formula:
favorable [22] and the rate of adatom diffusion is increased
2S
13
by the presence of this adlayer [23], thus giving rise to
ε
=
ε ,
xx
(1)
(2)
zz
S + S
12
enhanced step-flow growth for In-doped GaN layers [19].
In this study, indium is observed to enhance the lateral
growth and reduce the growth rate of the c face of an LT-
AlGaN interlayer.
11
2
(C + C ) C - 2C
11
12
33
13
σ
=
ε ,
xx
xx
C
33
(0)
(0)
(0)
where strain ε = (c - c )/c , ε = (a - a )/
zz GaN GaN GaN xx GaN GaN
a
, C and S are elastic constants and elastic compli-
2 Experiments GaN epilayers were grown in a hori-
zontal-type low-pressure (~76 Torr) metalorganic chemical
vapor deposition (MOCVD) system. Trimethylgallium
(TMGa), trimethylaluminium (TMAl), trimethylindium
(0)
GaN
ij ij
ances for GaN, respectively. The lattice constant c of GaN
layers is measured by XRD in a 2θ–ω scan mode. The
following parameters are adopted for these calcula-
(TMIn) and ammonia (NH ) are used as precursors for
3
–3
–1
–4
tions: S = 3.086 × 10 GPa , S = –9.96 × 10 GPa ,
–1
11
12
Ga, Al, In and N, respectively. The Si(111) substrate was
chemically cleaned (a 3% HF solution for 30 s) before
being loaded into the reactor.
–4 –1
S = –5.566 × 10 GPa , C = 390 GPa, C = 145 GPa,
13
11
12
C
= 106 GPa, C = 398 GPa [25, 26]. The unstrained
33
13
(0) (0)
bulk GaN constants of c and a are 0.5186 nm and
GaN
GaN
Our process starts with a deoxidation of the substrate
0.3189 nm [1], respectively. The results shown in Fig. 1
indicate that, with the increase of TMIn flow during the
growth of LT interlayer, the tensile stress in the subsequent
HT-GaN interlayers is decreased, accompanied by the nar-
rowing of the full width at half-maximum (FWHM) of the
surface under a H flow at high temperature (1080 °C).
2
Then, a few monolayers of Al are deposited on the silicon
prior to an AlN (~30 nm) buffer layer growth at the same
temperature (1080 °C). Subsequently, a deposition of about
200 nm of In doping or undoped Al Ga N interlayer
0.25
0.75
follows the AlN buffer, before the growth of a 1.5–2.0 µm
thick GaN layer at 1050 °C. The typical temperature used
for the growth of the Al Ga N interlayer is 1000 °C. H
0.25
0.75
2
is used as the carrier gas for all the processes. The rela-
tively low temperature for the interlayer is used to create
additional compressive strain to counterbalance the tensile
strain in the subsequent GaN film deposited on silicon [12],
while maintain a proper crystalline quality and interface
smoothness. Meanwhile, the LT-Al Ga N interlayer for
x
1–x
samples without a HT-GaN layer is deposited under the
same conditions in order to obtain the same Al fraction in
the LT-AlGaN interlayer. In addition, the other two LT-
AlGaN samples have been grown on Si(111) in which the
TMAl flow rate is reduced intentionally (about 10 at% Al)
in order to satisfy the demands of PL measurement by us-
ing the 325 nm line of a He–Cd laser. These two low-Al
content of AlGaN samples have only been used in PL
measurements. For the rest of the GaN samples grown on a
LT-AlGaN interlayer, the Al content of the LT-AlGaN in-
terlayer has been kept constant at 25 vol%. The detailed
growth procedures have been described elsewhere [24].
Figure 1 Dependence of the residual stress in a HT-GaN layer
(᭿) determined from HRXRD 2θ–ω scan curves, and the FWHM
of the HRXRD rocking curves for a (0002) HT-GaN layer (᭹) on
the TMIn flow rate during the growth of a LT-AlGaN interlayer.
www.pss-a.com
© 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

a
p s s
296
Jiejun Wu et al.: Effect of indium-doped interlayer on strain relief in GaN films
ω-scan XRD. The narrower XRD linewidth reflects the
However, when the TMIn flow is increased further
improved crystal quality of the epilayer, partly due to the (>16 µmol/mol), the stress in the GaN film begins to in-
reduction of the tensile stress in the epilayer. Experimental crease again and the FWHM of GaN is also increased. The
data further indicate that, by introducing even a small In crystalline quality of GaN is reduced for a larger TMIn
flow rate in the growth of the LT interlayer, a dramatically flow. The reasons are not clear and need to be researched
reduced tensile stress is observed in the subsequent GaN further. Neugebauer [27] studied the effects of In on the
layer. The GaN epilayer grown on the In-doped LT inter- energetics and atomic structure of the GaN surface. His
layer exhibits less tensile stress and higher crystal quality. calculation results showed that, in the case of GaN films
When the TMIn flow is up to 16 sccm, the stress is reduced grown in the presence of high In flow, indium can substi-
by up to 0.61 GPa. The FWHM for the 1.5 µm HT-GaN tute for Ga and the formation of In–N bonds might have
sample is reduced to 8.2 arcmin, while it is smaller than the occurred. (Ga,In)N alloys might actually form in the bulk
13 arcmin for an undoped sample. Typical (0002) rocking film. The formation of (Ga,In)N alloys will lead to the de-
curve FWHM for optimized GaN on AlN/Si(111) range crease of compensative stress provided by the AlGaN in-
from 8–16 arcmin [14, 15]. Furthermore, the optical prop- terlayer and the increase of tensile stress in the GaN upper
erties of samples are determined by LT-PL spectra at 10 K. layers. More tensile stress will induce more cracks to form
The FWHM of the neutral-donor band (D°X) luminescence on the surface of the GaN. So when the TMIn flow is up to
peak of the GaN film grown on a In-doped interlayer is 24 µmol/mol, the tensile stress is increased in the GaN
9 meV and is smaller than that of a GaN layer on an un- films and cracks appear on the surface again by inserting
doped interlayer (~14 meV). The best results reported for the In-doped AlGaN interlayer on Si(111) substrates.
GaN on Si(111) using a single interlayer process are 14–
The decreased stress benefits the reduction of the crack
17 meV [11, 13]. A peak width as narrow as 5–6 meV has density in the surface of GaN layers. The SEM morpholo-
been reported for layers grown on AlGaN/GaN superlat- gies of the GaN surfaces are shown in Fig. 2. As seen, the
tices [14]. Therefore, a method of inserting an In-doped sample without any LT interlayer is full of a high-density
LT-AlGaN interlayer is proven to effectively improve the crack network (Fig. 2d). The cracking network appears
crystalline and optical properties for GaN epilayers grown when the thickness of the HT-GaN layer exceeds 1.0 µm.
on Si(111).
These cracks are caused by excessive biaxial tensile strain.
Figure 2 SEM plane-view images of (a) 1.5 µm and (b) 2.0 µm HT-GaN layer with In-doped interlayers, (c) 1.5 µm HT-GaN layer
with an undoped interlayer, and (d) 1.0 µm HT-GaN layer without an interlayer. The insets show the AFM surface images.
© 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.pss-a.com

Original
Paper
phys. stat. sol. (a) 205, No. 2 (2008)
297
The cracking is an effective way to release the tensile
stress by activating a slip system around the crack tips
when the GaN layer is over the critical thickness. In all
samples, on inserting a LT-AlGaN interlayer, whether In-
doped or not, the crack density is significantly reduced.
However, by using an In-doped LT-AlGaN interlayer,
the maximum crack-free layer thickness is up to 1.5 µm
(Fig. 2a) and the thicker (2.0 µm) film has intact regions
2
with area up to several mm with the crack density reduced
–1
to 10 cm (Fig. 2b). In the sample with an undoped inter-
layer, cracking is still observed in the surface and the crack
–1
density is up to 70 cm when the HT-GaN thickness is
only 1.5 µm (Fig. 2c).
The decreased stress also benefits the reduction of the
surface roughness of GaN layers. The AFM images of the
GaN surfaces are shown in the insets of Fig. 2. The surface
of the film grown on the In-doped AlGaN interlayer shown
in the inset of Fig. 2a, is quite smooth and is characterized
Figure 3 Low-temperature photoluminescence spectra at 10 K of
LT-AlGaN interlayers with and without In doping. The inset
shows the (0002) HRXRD 2θ–ω scan curves of LT-AlGaN
by terraces separated by monolayer-high steps. The film interlayers with and without In doping.
grown on the undoped AlGaN interlayer shown in the inset
of Fig. 2c, is rough and has heavy pits on the surface. As an In-doped AlGaN layer compared with an undoped sam-
seen, the roughness of the GaN layer is considerably im- ple with the same Al fraction. The increased qualities and
proved by the use of an In-doped interlayer. The rms reduced tensile strain can be used to explain why In doping
2
roughness measured on a 2 × 2 µm scan is only 0.6 nm, a LT-AlGaN interlayer can reduce the crack density and
and smoother than that of GaN with an undoped LT inter-
layer (rms = 2.1 nm). Therefore, an In-doped LT-AlGaN quently grown HT-GaN layer.
interlayer is shown to be an effective method not only to
In previous studies of In-doped GaN layers, Widmann
improve the surface quality, but also to avoid crack forma- et al. [28] reported the reduction of deep-level emission,
tion in an overgrown thick HT-GaN layer [16].
while Shu et al. [20] found that the FWHM of the PL peak
In order to understand the effects of indium doping, we
improve the optical and crystalline qualities of a subse-
was reduced due to In doping. Yuan et al. [19] noticed that
investigated the crystal and optical properties of the LT- the peak intensity increased with thickness, and they be-
AlGaN interlayers with and without In doping measured lieved that In doping can suppress the nonradiative centers
by LT-PL spectra and HRXRD 2θ–ω. The Al mole frac- and lead to the improvement of the band-edge PL proper-
tion in the AlGaN samples is estimated by EDS and RBS ties. These results support the view that indium doping not
spectra. Results show that incorporated Al fraction remains
only decreases the strain in the film, but also increases the
nearly constant in an AlGaN interlayer whether there is In optical and crystalline qualities of materials. Reduced ten-
doping or not. In Fig. 3 the LT-PL spectra at 10 K are dis- sile stress and enhanced qualities in the AlGaN interlayer
played for In-doped and undoped AlGaN samples. In order will benefit the creation of an additional compressive strain
to satisfy the demands of PL measurement by using the to counterbalance the tensile strain during the growth of
325 nm line of a He–Cd laser, the other two LT-AlGaN
samples have been grown on Si(111) in which the TMAl in the HT-GaN grown on the In-doped LT-AlGaN inter-
flow rate is intentionally reduced (about 10 at% Al). The layers compared with undoped interlayers.
LT-PL spectra indicate that the FWHM of the D°X emis-
Figure 4 shows the AFM micrographs of LT-AlGaN
sion peak of In-doped AlGaN interlayer is smaller than that In-doped and undoped interlayers, respectively (2 µm
of the undoped interlayer (145 versus 173 meV), which
subsequent GaN layers. It can explain the extra stress relief
× 2 µm scan area). Typical cross-sectional view of the
is consistent with the decrease of FWHM values of XRD AFM scan is shown in the insets of Fig. 4. Significant
ω-scan spectra (31 versus 33 arcmin) shown in the inset of changes are observed in the morphology between the two
Fig. 3. Moreover, the peak intensity for In-doped samples samples. The undoped sample exhibits an almost coalesced
is about five times stronger than that for undoped samples. morphology and the shape of the mounded islands appears
The narrower linewidth and higher intensity demonstrate like some “terrace-like” structures. The top surface of is-
that the optical and crystal qualities of the LT-AlGaN in- lands is smoother. The In-doped sample shows a “tentacle-
terlayer have been improved by using the In-doping
like” morphology and the shape of small islands is like a
method. Improved LT-AlGaN interlayer can serve as a tentacle-shape stretched from the main terrace-like islands.
good seed layer to grow subsequent advanced HT-GaN From AFM micrographs, the average height of the terrace-
layer. In additional, a small peak blueshift is observed in like main islands is about 140 nm and 180 nm for the In-
In-doped LT-AlGaN compared to the undoped sample. doped sample and the undoped sample, respectively. Since
Such a blueshift indicates the relaxation of tensile strain in
adatoms incorporate into the islands rather than the whole
www.pss-a.com
© 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

a
p s s
298
Jiejun Wu et al.: Effect of indium-doped interlayer on strain relief in GaN films
Figure 4 (online colour at: www.pss-a.com) Surface morphologies of the LT-AlGaN interlayers observed by AFM (a) with In doping,
and (b) without In doping. The insets show their typical cross-sectional scans, respectively.
surface, the lower island height indicates that In doping has 10.3 nm and 2.4 nm for the In-doped sample and the un-
enhanced the lateral growth and the growth rate of the c doped sample, respectively. The tentacle-shaped small is-
face is reduced for the In-doped sample. The tilt angle of lands on the main island are only observed in the In-doped
side faces of the islands is estimated to be 48°. From the sample. Their average height and radius are 20 nm and
tilt angle, the most likely asymmetric face is found to be 25 nm, respectively. The larger rms roughness of the In-
(10 11). When the TMIn flow is below the critical value, doped sample can be attributed to the rougher top faces in-
the solubility of indium in bulk AlGaN is negligible. Thus, duced by the tentacle-like islands stretching from the main
indium will float on top of the surface and will not be in- islands. From the insets of Fig. 4 of the AFM cross-
corporated. The presence of an In adlayer [22] on the sur- sectional scans, the micrographs show that there are some
face quantitatively affects the diffusion of N adatoms: Lat- small peaks in the In-doped sample. These small peaks are
eral transport of N adatoms on this surface is not realized the cross section of the tentacle-like structures. The small
by diffusion on the top surface layer but below the In ad- peaks had lateral sizes of about 50 nm. In the undoped
layer [23]. This mechanism reduces the N diffusion barrier sample, such structures are not observed and therefore we
dramatically and opens an efficient and hitherto unex- believe that the formation of tentacles in the In-doped
pected diffusion channel for lateral surface transport. The sample is related to the indium adatoms on the surface. The
significantly enhanced adatom diffusivity on the surface al- tentacle-like structures that stretched from the terrace is-
lows a smooth surface morphology to be achieved.
lands top surface have been little observed in In doping of
However, In doping is observed not to improve the the GaN system [19]. Compared with the conventional
smooth surface of a LT-AlGaN interlayer. The rms rough- GaN system, the AlGaN interlayers are grown at 1000 °C,
ness is 52 nm for the In-doped sample, and is slight larger and lower than the typical temperature (1050 °C) of GaN
than that of the undoped sample of 47 nm. In order to ex- layers. Temperature may be an important parameter. In or-
plain this phenomenon, AFM measurements are performed der to prove this, the surface image of an In-doped AlGaN
to observe locally the top surface of the terrace island in layer grown in 1050 °C temperature has been measured by
both samples. The typical top-face rms roughness is about AFM. However, the same surface of tentacle-like struc-
© 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.pss-a.com

Original
Paper
phys. stat. sol. (a) 205, No. 2 (2008)
299
[5] N. P. Kobayashi, J. T. Kobayashi, P. D. Dapkus, W. J. Choi,
and A. E. Bond, Appl. Phys. Lett. 71, 3568 (1997).
[6] S. Nishimura, S. Matsumoto, and K. Terashima, Opt. Mater.
19, 223 (2002).
tures was never observed. The AFM image is similar to
that of an undoped AlGaN layer (shown in Fig. 4b). The
results indicate that the different surface is dependent on
the growth temperature. At lower temperature, In adatoms
are not easy to transport on the surface and desorb from the
surface. More In adatoms benefit the formation of nuclei
on the surface of the main islands when the main islands
themselves are grown according to the traditional mode.
Tentacle-like structures on the main terrace island top
surface are developed from the growth and coalescence
of nuclei. But details of the processes need to be studied
further.
[7] R. Armitage, Q. Yang, H. Feick, J. Gebauer, and E. R. We-
ber, Appl. Phys. Lett. 81, 1450 (2002).
[8] Y. Honda, Y. Kuroiwa, M. Yamaguchi, and N. Sawaki,
Appl. Phys. Lett. 80, 222 (2002).
[9] X. L. Fang, Y. Q. Wang, H. Meidia, and S. Mahajan, Appl.
Phys. Lett. 84, 488 (2004).
[10] H. Amano, M. Iwaya, T. Kashima, M. Katsuragawa, I. Aka-
saki, J. Han, S. Hearne, J. A. Floro, E. Chason, and J. Figiel,
Jpn. J. Appl. Phys., Part 2 37, L1540 (1998).
[11] A. Dadgar, J. Bläsing, A. Diez, A. Alam, M. Heuken, and
A. Krost, Jpn. J. Appl. Phys., Part 2 39, L1183 (2000).
[12] J. Bläsing, A. Beiher, A. Dadgar, A. Diez, and A. Krost,
Appl. Phys. Lett. 81, 2722 (2002).
4 Conclusion In this paper, high-quality GaN layers
on silicon have been grown by MOCVD using the inser-
tion of the In-doped LT-AlGaN interlayer. Isoelectronic In
doping has been found to enhance the lateral growth while
the c-face growth rate is reduced. Thus, In doping is effec-
tive in enhancing the PL intensity and crystalline quality of
the LT interlayer, and this lead to the reduction of tensile
stress in a subsequent GaN epilayer with an increase in
TMIn flow during interlayer growth. The relaxation of ten-
sile stress can control and eliminate the generation of
cracks in thick GaN films. In summary, the method using
the In-doped LT-AlGaN interlayer is feasible to achieve
the nearly crack-free 2.0 µm thick GaN films grown on
silicon with high optical and crystal properties.
[13] M. H. Kim, Y. G. Do, H. C. Kang, D. Y. Noh, and S. J. Park,
Appl. Phys. Lett. 79, 2713 (2001).
[14] E. Feltin, B. Beaumont, M. Laügt, P. de Mierry, P. Ven-
néguès, H. Lahrèche, M. Leroux, and P. Gibart, Appl. Phys.
Lett. 79, 3230 (2001).
[15] A. Reiher, J. Bläsing, A. Dadgar, A. Diez, and A. Krost,
J. Cryst. Growth 248, 563 (2003).
[16] J. J. Wu, X. X. Han, J. M. Li, H. Y. Wei, G. W. Chong,
X. L. Liu, Q. S. Zhu, and Z. G. Wang, Opt. Mater. 28, 1227
(2006).
[17] S. Keller, S. Heikman, I. Ben-Yaacov, L. Shen, S. P. Den-
baars, and U. K. Mishra, App. Phys. Lett. 79, 3449 (2001).
[18] S. Yamaguchi, M. Kariya, T. Kashima, S. Nitta, M. Kosaki,
Y. Yukawa, H. Amano, and I. Akasaki, Phys. Rev. B 64,
035318 (2001).
Acknowledgments The authors are grateful to Professor
Baohe Wei of the Petroleum Exploration and Development Re-
search Institute and Professor Yongzhong Wang of Peking Uni-
versity for their cooperation in SEM and XRD analyses, respec-
tively. The authors also thank Professor Guohua Li and Kun
Ding of the Institute of Semiconductors, Chinese Academy of
Sciences, for their help in PL observation and Dr Ruoyuan Li
for AFM measurements. This work was supported by the
National Natural Science Founds of China (No. 60376013 and
No. 60136020).
[19] H. Yuan, S. J. Chua, S. Tripathy, and P. Chen, J. Vac. Sci.
Technol. A 21, 1814 (2003).
[20] C. K. Shu, J. Ou, H. C. Lin, W. K. Chen, and M. C. Lee,
Appl. Phys. Lett. 73, 641 (1998).
[21] H. Kumano, K. Hoshi, S. Tanaka, and I. Suemune, X. Shen,
P. Riblet, P. Ramvall, and Y. Aoyagi, Appl. Phys. Lett. 75,
2879 (1999).
[22] J. E. Northrup and C. G. V. Walle, Appl. Phys. Lett. 84,
4322 (2004).
[23] J. Neugebauer, T. K. Zywietz, M. Scheffler, J. E. Northrup,
H. Chen, and R. M. Feenstra, Phys. Rev. Lett. 90, 056101
(2003).
References
[1] A. Dadgar, A. Strittmatter, J. Blasing, M. Poschenrieder,
O. Contreras, and P. Veit, phys. stat. sol. (c) 0(6), 1583 [24] Y. Lu, X. Liu, D. Lu, H. Yuan, Z. Chen, T. Fan, and
(2003).
Z. Wang, J. Cryst. Growth 236, 77 (2002).
[2] H. Ishikawa, K. Yamamoto, T. Egawa, T. Soga, T. Jimbo, [25] P. Perlin, L. Mattos, N. A. Shapiro, J. Kruger, W. S. Wil-
and M. Umeno, J. Cryst. Growth 189/190, 178 (1998).
liam, T. Sands, N. W. Cheung, and E. R. Weber, Appl. Phys.
[3] P. Kung, A. Saxler, X. Zhang, D. Walker, T. C. Wang,
Lett. 85, 2385 (1999).
I. Fergusson, and M. Razeghi, Appl. Phys. Lett. 66, 2958 [26] A. Polian, M. Grimsditch, and I. Grzegory, J. Appl. Phys.
(1995).
79, 3343 (1996).
[4] R. F. Davis, T. Gehrke, K. J. Linthicum, E. Preble, P. Ra- [27] J. Neugebauer, phys. stat. sol. (c) 6, 1651 (2003).
jagopal, C. Ronning, C. Zorman, and M. Mehregany, [28] F. Widmann, B. Daudin, G. Feuillet, N. Pelekanos, and
J. Cryst. Growth 231, 335 (2001).
J. L. Rouviere, Appl. Phys. Lett. 73, 2642 (1998).
www.pss-a.com
© 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim