Effects of Si-doping on the Microstructure of AlGaN/GaN Multiple-quantum-well
R. Liu and F. A. Ponce
Department of Physics and Astronomy, Arizona State University, Tempe, AZ 85287
S-L. Sahonta and D. Cherns
H. H. Wills Physics Laboratory, University of Bristol, Bristol BS8 1TL, UK
H. Amano and I. Akasaki
Department of Materials Science and Engineering, Meijo University, Nagoya 468, Japan
ABSTRACT
The effects of silicon-doping on the microstructure of Al
0.07
Ga
0.93
N/GaN multiple-
quantum-well (MQW) have been studied by TEM. Significant changes of surface
morphology and dislocation core structures have been observed due to Si-doping in the
Al
0.07
Ga
0.93
N barriers. Threading dislocations create surface pits in the MQW as a result
of Si doping. With an increasing doping level, the pits change the shape from small
faceted pyramid to large cone. The formation mechanism of the surface pits has been
discussed from both dynamics and kinetics points of view. We have also observed
nanopipes constrict to form closed core screw dislocations in the MQW due to Si-doping.
INTRODUCTION
Applications of UV-LEDs using GaN active regions have been obstructed by the
weak luminescence efficiency [1]. In comparison to InGaN quantum well (QW), GaN
QW appears to suffer more from the non-radiative recombination nature of threading
dislocations and the quantum-confined Stark effect caused by the internal field [1]. This
is simply because GaN QW is free of alloy fluctuation, which is believed to be essential
for a high radiative recombination rate in InGaN QW [1]. For further improvement of the
luminescence efficiency, Si doping in barrier layers is promising, because the doping
design is expected to improve radiative efficiency by screening the internal field that
separates the electron-hole wave function [2]. Si doping in InGaN barrier layers in
InGaN/InGaN MQW has been found to improve the optical gain of the laser diodes [3].
A study of the optical properties of Si-doped AlGaN/GaN MQW has been reported
earlier [2]. In the present work, the doping effect on the microstructure is studied by
TEM.
EXPERIMENTAL
The studied samples are a set of three Al
0.07
Ga
0.93
N/GaN MQW structures grown on
GaN by metal-organic vapor phase epitaxy. Detailed growth parameters were reported
earlier [4]. The MQW consists of 5 periods of AlGaN/GaN structure. The difference
between the samples is the Si doping level in the AlGaN barriers. The first sample is
nominally undoped. And the other two have very high doping levels, which are reflected
by their high carrier concentrations: about 4x10
19
cm
-3
and 9x10
19
cm
-3
respectively. In
comparison, the background carrier concentration for the nominally undoped one is about
1x10
16
cm
-3
. The microstructure of these samples has been characterized by TEM in
Y5.33.1Mat. Res. Soc. Symp. Proc. Vol. 798 © 2004 Materials Research Society
cross-section and plan-view. The microscopes used in this study are JEOL 4000FX
operated at 400kV and Philip CM200 FEG operated at 200kV. The foil specimens were
prepared by mechanical polishing followed by Ar
+
ion milling.
RESULTS AND DISCUSSION
Surface pits at the ends of threading dislocations
One significant effect of Si-doping on the microstructure is inducing surface openings
at the ends of threading dislocations. Figure 1 shows TEM images of the undoped
sample taken in cross-section and plan-view. In Fig. 1(a), a mixed type threading
dislocation propagates from the underlying GaN layer into the MQW structure and does
not create a visible opening on the surface. Similarly, Fig. 1(b) shows a nanopipe
propagates through the MQW structure without changing its shape. Figure 1(c) shows a
plan-view image of this sample. There is no evidence of surface pits at the ends of the
threading dislocations. However, the microstructure is significantly changed when the
AlGaN barriers are doped with Si. Figure 2 shows a set of TEM images of the second
sample with a Si-doping level about 4x10
19
cm
-3
. In Fig. 2(a), a cross-section image
shows many surface pits. Close examination indicates that they in fact originate from
threading dislocations. Since the image was taken under g=0002 diffraction condition
g=0002
a
g=0002
nanopipe
b
g=11-20
nanopipe
30nm
c
Figure 1. Bright field TEM images of the
Al
0.07
Ga
0.93
N/GaN MQW on GaN without Si-
doping in the barriers: (a) and (b) are taken in
cross-section, and (c) in plan-view. Threading
dislocations form no pits on the film surface.
Y5.33.2
to show threading dislocations with a screw component in strong contrast, those of pure
edge nature only show faintly (ideally should be invisible). In Fig. 2 (b), we observe a
nanopipe propagating from the GaN template changes to a full dislocation and form
opening on the surface. In short, it appears that all the threading dislocations, regardless
of the Burgers vector, form surface pits due to Si-doping. Morphology of the surface pits
is quite complex. In cross-section, they look like tulips; while plan-view imaging shows
that they are bound with six-fold symmetry facets. The actual morphology probably
consists of two parts: one is the faceted pyramid bottom part and the other is the cone-
shape top. However, in the higher doped sample, the surface pits are completely cone-
shape. In addition, we also notice that threading dislocations start the formation of the
surface pits earlier with a higher doping level. The surface pits formed almost
immediately when the deposition of the MQW started (Fig. 3(a)). The differences
between the two samples may give some clues when we analyze the pit formation
mechanism.
Pit formation mechanism
In InGaN/GaN MQW structures, similar surface pits have been observed by several
groups [5-6]. The pits were found to be bound with six {101
1} facets. There were some
arguments about the formation mechanism of these defects [5-8]. Relaxation of lattice
g=0002
a
g=0111
c
nanopipe
g=0002
b
nanopipe
Figure 2. Bright field TEM images of the Al
0.07
Ga
0.93
N:Si/GaN MQW on GaN with a
doping level about 4x10
19
cm
-3
: (a) and (b) are cross-section images, and (c) is plan-view.
Threading dislocations form surface pits with six-fold symmetry. Nanopipes change to
full screw dislocations and form surface pits in the MQW.
Y5.33.3
mismatch strain was believed to be the driving force of the surface pit formation [5,7].
However, it was argued that a reduced incorporation of Ga atoms on the {101
1} facets
results in the surface pits [6]. First-principles calculations suggested that indium
segregation on the facets may reduce the surface free energy and thereby allow formation
of the relatively large size surface pits [8].
In the present work of Al
0.07
Ga
0.93
N/GaN MQW, we found no surface pits for the
undoped sample. When the barriers were doped with Si, threading dislocations formed
V-shape pits on the surface. In the sample with the higher doping level, the pits are much
larger and no specific facets are formed. This observation suggests that Si impurities
probably reduce the surface free energy, which is similar to indium in the study of
InGaN/GaN MQW [8]. There are two possible ways to explain the observations. One is
via analyzing the doping effect on the system energy (from a dynamics point of view),
and the other is to consider the doping effect on the growth kinetics (from a kinetics point
of view). Following Frank’s theory for determining equilibrium radii of hollow core
dislocations [9], it has been proposed that the pit formation energy consists of three terms:
(1) the change in surface energy resulting from replacing a hexagonal region of the (0001)
surface with the six {101
0} facets of the inverted pyramid pit, (2) the reduction in strain
energy (around threading dislocation cores), and (3) the elimination of the dislocation
core energy [8]. In addition, there may be another term need to be considered: (4) the
reduction in lattice mismatch strain energy, which results from the surface roughing
caused by pitting. The reason why the undoped sample do not exhibit surface pits might
be that the critical thickness for the pit formation is not exceeded, i.e. the first term is
larger than the sum of the other three. This speculation can be justified by comparing
lattice mismatches between AlN/GaN and InN/GaN. For AlN/GaN, the value is about -
2.4%, which is less than one quarter of the magnitude between InN and GaN (about 11%).
Thus, AlGaN alloys grown on GaN will have much reduced lattice mismatch strain
compared with similar composition InGaN alloys. In addition, Al content is only 7% in
this study. Therefore, the critical thickness for the surface pit formation is probably
larger in the currently studied Al
0.07
Ga
0.93
N/GaN MQW structure. When Si impurities
were introduced into the growth, the surface free energy might be reduced. This
decreased the threshold energy for the pit formation (the first term) to such a point that it
was even lower than the sum of the other three terms. In consequence, surface pits were
created at threading dislocation ends to minimize the system energy. However, the
validity of this explanation remains to be confirmed because Frank’s theory has been
found to be significant only for core radii less than 1nm in GaN [9]. Therefore, it is
sensible to consider an alternative explanation from a kinetics point of view, i.e. the pit
formation solely results from Si-doping effect on growth kinetics. On the growth surface,
threading dislocations generate localized strain field in the vicinity of their cores, which
may be attractive for impurity segregation. If Si impurities segregate at the ends of
threading dislocations, they may retard the growth of the dislocation core. As a result,
facets like {101
1} can be generated and open up in the following growth because Si
impurities may help to stabilize the facets. We can also explain the delay of the pit
formation in the sample with the lower doping level by the need to build up the surface Si
concentration to a point where stable facets can be formed. In short, further study is
needed in order to determine which factor is playing the key role in the pit formation.
Y5.33.4
Closure of nanopipes
In Fig. 2(b-c), we show that nanopipes constrict to form closed core screw
dislocations when they propagate into the MQW structure. Doping of Si appears to be
responsible for the transformation, since the same defects in the undoped sample
maintain the coreless feature (Fig. 1(b-c)). In the sample with the highest doping level,
nanopipes form surface pits before they have a chance to become full dislocations (Fig.
3(b)). The origin of the core structure transformation is puzzling. It cannot be accounted
for by Frank’s model [10], because reduction in surface free energy due to Si doping
should increase the core radius rather than decrease it, provided that the strain energy
around the dislocation core remains the same. However, Si doping has been reported to
increase the surface roughness by enhancing spiral growth mode [11,12]. It is probably
an accelerated spiral growth of the nanopipe that leads to the constriction of the opening
cores.
SUMMARY
The effects of Si-doping on the microstructure of the Al
0.07
Ga
0.93
N/GaN MQW
structures have been studied. According to our observations, all threading dislocations,
regardless of the Burgers vector, generate surface pits in Si-doped AlGaN layers. With
an increasing doping level, the originally pyramid-faceted pits become cone-shape. The
formation mechanism of the surface pits due to Si-doping has been discussed from both
dynamics and kinetics points of view. In addition, Si-doping also results in closure of
nanopipes in the MQW. The effect of Si-doping on growth kinetics is probably
responsible for the transformation. This observation may provide a new way to get rid of
the detrimental effects of nanopipes on device performance.
g=0002
a
g
=11-20
b
nanopipe
Figure 3. Bright field TEM images of the
Al
0.07
Ga
0.93
N:Si/GaN MQW on GaN with a
doping level about 9x10
19
cm
-3
: (a) and (b)
are cross-section and plan-view
respectively. Threading dislocations form
deep surface pits in circular funnel shape.
Y5.33.5
ACKNOWLEDGEMENTS
This work is supported by an international NEDO (New Energy and Industrial
Technology Development Organization) program (NO 01-MB10).
REFERENCES
[1] S. Chichibu, A. Shikanai, T. Deguchi, A. Setoguchi, R. Nakai, H. Nakanishi, K. Wada,
S. P. DenBaars, T. Sota and S. Nakamura, Jpn. J. Appl. Phys. 39, 2417 (2000).
[2] B. Monemar, H. Haratizadeh, P. P. Paskov, G. Pozina, P. O. Holtz, J. P. Bergman, S.
Kamiyama, M. Iwaya, H. Amano, and I. Akasaki, phys. stat. sol. (b) 237, 353 (2003).
[3] S. Chichibu, D. A. Cohen, M. P. Mack, A. C. Abare, P. Kozodoy, M. Minsky, S.
Fleischer, S. Keller, J. E. Bowers, U. K. Mishra, L. A. Coldren, D. R. Clarke, and S. P.
DenBaars, Appl. Phys. Lett. 73, 469 (1998).
[4] T. Nakamura, S. Mochizuki, S. Terao, T. Sano, M. Iwaya, S. Kamiyama, H. Amano
and I. Akasaki, J. Cryst. Growth 237/239, 1129 (2002).
[5] C. J. Sun, M. Zubair Anwar, Q. Chen, J. W. Yang, M. Asif Khan, M. S. Shur, A. D.
Bykhovski, Z. Liliental-Weber, C. Kisielowski, M. Smith, J. Y. Lin, and H. X. Xiang,
Appl. Phys. Lett. 70, 2978 (1997).
[6] X. H. Wu, C. R. Elsass, A. Abare, M. Mack, S. Keller, P. M. Petroff, S. P. DenBaars,
and J. S. Speck and S. J. Rosner, Appl. Phys. Lett. 72, 692 (1998).
[7] Y-S Lin, K-J Ma, C. Hsu, S-W Feng, Y-C Cheng, C-C Liao, C. C. Yang, C-C Chou,
C-M Lee, and J-I Chyi, Appl. Phys. Lett. 77, 2988 (2000).
[8] J. E. Northrup, L. T. Romano and J. Neugebauer, Appl. Phys. Lett. 74, 2319 (1999).
[9] J. Elsner, R. Jones, P. K. Sitch, V. D. Porezag, M. Elstner, Th. Frauenheim, M. I.
Heggie, S. Öberg, and P. R. Briddon, Phys. Rev. Lett. 79, 3672 (1997).
[10] F. C. Frank, Acta Crystallogr. 4, 497 (1951).
[11] L. T. Romano, C. G. Van de Walle, J. W. Ager III, W. Götz and R. S. Kern, J. Appl.
Phys. 87, 7745 (2000).
[12] S. Keller, U. K. Mishra, S. P. DenBaars, and W. Siefert, Jpn. J. Appl. Phys., Part 2
37, L431(1998).
Y5.33.6