APPLIED PHYSICS LETTERS 92, 171904 共2008兲
Structural and optical properties of nonpolar GaN thin films Z. H. Wu,1 A. M. Fischer,1 F. A. Ponce,1,a兲 B. Bastek,2 J. Christen,2 T. Wernicke,3 M. Weyers,3 and M. Kneissl3,b兲 1
Department of Physics, Arizona State University, Tempe, Arizona 85287-1504, USA Otto-von-Guericke Universitat, Institut für Experimentelle Physik, PO Box 4120, 39016 Magdeburg, Germany 3 Ferdinand-Braun-Institute, Gustav-Kirchhoff-Str. 4, 12489 Berlin, Germany 2
共Received 29 February 2008; accepted 10 April 2008; published online 29 April 2008兲 ¯ 0兴 A correlation between the structural and optical properties of GaN thin films grown in the 关112 direction has been established using transmission electron microscopy and cathodoluminescence spectroscopy. The GaN films were grown on an r-plane sapphire substrate, and epitaxial lateral overgrowth was achieved using SiO2 masks. A comparison between the properties of GaN directly grown on sapphire and GaN laterally grown over the SiO2 mask is presented. The densities and dimensions of the stacking faults vary significantly with a high density of short faults in the window region and a much lower density of longer faults in the wing region. The low-temperature luminescence spectra consist of peaks at 3.465 and 3.41 eV, corresponding to emission from donor-bound excitons and basal-plane stacking faults, respectively. A correlation between the structural defects and the light emission characteristics is presented. © 2008 American Institute of Physics. 关DOI: 10.1063/1.2918834兴 Growth of GaN along nonpolar directions is expected to improve the efficiency of light emitting diodes 共LEDs兲. This is accomplished by removing the internal fields normal to the InGaN quantum wells, which are present in the standard 关0001兴 growth direction.1–3 The most common nonpolar ¯ 0兲 a-plane and 共11 ¯ 00兲 crystal orientations are the 共112 m-plane of GaN. Since nonpolar GaN is not readily available in the bulk form, a-plane GaN epilayers can be grown on the ¯ 02兲 共r-plane兲 sapphire substrates, where lattice and 共11 thermal-expansion mismatch lead to the presence of a high density of extended defects.4,5 The technique of epitaxial lateral overgrowth 共ELO兲 can be used to reduce the defect density in the wing region,6–8 but it significantly increases the manufacturing complexity due to the additional mask patterning process. Therefore, in order to commercialize nonpolar GaN based LEDs, it remains critical to understand the formation mechanisms of extended defects and to find ways to minimize them. The extended defects in nonpolar GaN consist mainly of threading dislocations 共TDs兲 and of basalplane stacking fault 共SF兲 segments. While TDs are not active radiative centers, the SFs have characteristic luminescence centered at ⬃3.4 eV.9 SFs terminate in partial dislocations, thus the density of partial dislocations in the film will depend on the density of SF segments, and it is expected to severely affect the optical properties of the material. In this study, we used transmission electron microscopy 共TEM兲 and cathodoluminescence 共CL兲 microscopy to establish the correlation between the defect distribution and optical properties for the a-plane GaN epitaxy. For ELO films, the window region has a weak optical emission and a high density of SFs that are short in dimension. The wing region has opposite characteristics, i.e., a strong optical emission and a low density of longer SFs. a兲
Electronic mail:
[email protected]. Also at Technical University of Berlin, Institute of Solid State Physics, Hardenbergstr. 36, 10623 Berlin, Germany.
b兲
The GaN films were epitaxially grown by metal-organic vapor phase epitaxy on the r-plane sapphire substrates. ELO was achieved using a SiO2 mask. The films were grown to a thickness of about 30 m. Plan-view samples of the top film surface region were prepared for TEM using standard mechanical polishing and argon-ion milling techniques. We used a field-emission TEM operated at 200 keV. The TEM specimen was characterized by CL and then by TEM. The TEM images were taken from the same region that was used for CL imaging and a direct correlation between microstructure and luminescence was thus obtained. TEM images were also taken from other regions which were more suitable for a detailed TEM analysis. In order to establish a direct correlation between the CL and TEM, it is critical to locate an area suitable for both measurements. CL measurements require a region in the specimen with sufficient thickness for adequate light emission intensity. In contrast, TEM analysis usually requires the opposite, i.e., a thin sample for optimum electron transparency. Therefore, a region in the TEM sample has to be carefully chosen so that it is sufficiently thick to provide good signal for CL and thin enough for TEM analysis. The CL spectrum from one such an area is shown in Fig. 1. It exhibits two emission peaks associated with a donor-bound exciton 共DBE兲 at 3.465 eV and a SF emission at 3.41 eV.9 Corresponding monochromatic CL images taken at the DBE and SF emission peaks are shown in Figs. 2共a兲 and 2共b兲, respectively. Figure 2共c兲 shows a bright-field TEM image for the ¯ 00兴, same area taken under diffraction conditions of g = 关11 which is known to give a strong contrast for the SFs. The shape of the chosen area is somewhat uneven as a result of the compromise between CL and TEM requirements. As shown in Figs. 2共a兲 and 2共b兲, the DBE and SF intensities are weak in the window region but they appear strong and complementary to each other in the wing region. An anticorrelation between emission of DBEs and SFs has been reported.9,10 The observed distribution for the DBE emission
0003-6951/2008/92共17兲/171904/3/$23.00 92, 171904-1 © 2008 American Institute of Physics Downloaded 02 Jun 2008 to 62.141.180.1. Redistribution subject to AIP license or copyright; see http://apl.aip.org/apl/copyright.jsp
171904-2
Wu et al.
Appl. Phys. Lett. 92, 171904 共2008兲
FIG. 1. CL spectrum of a carefully selected area in a plan-view specimen, which was prepared for a suitable correlation of optical and structural characteristics. Two strong peaks are observed at 3.408 and 3.465 eV, corresponding to emission from basal-plane stacking faults and donor bound excitons, respectively.
can be straightforwardly correlated to the SF distribution in Fig. 2共c兲, where the window region exhibits a much higher density of SFs than the wing region. The DBE emission is strongest in the regions free from extended defects, and its intensity decreases with increasing density of extended defects. In spite of its high SF density, the window region shows a weaker SF emission intensity when compared to the wing region where the SF density is lower. In addition, the monochromatic image of SFs in the window region is spotty and discontinuous in contrast to the bright broad stripes ob-
FIG. 3. TEM images from a mask/window region from a plan-view sample ¯ 0兴. The mask/window boundaries are denoted by left oriented along 关112 arrows and the coalescence boundary is denoted by the right arrows. The images are taken under various diffraction conditions in order to assess the ¯ 00兴, was obtained by slightly rotatnature of the defect structure. 共a兲 g = 关11 ¯ ing the TEM specimen around the 关1100兴 axis giving a clear contrast for the stacking faults. 共b兲 g = 关0002兴, which was obtained by slightly rotating the TEM specimen around the 关0001兴 axis, highlights the presence of partial ¯ 110兴, which dislocations that laterally terminate the stacking faults. 共c兲 g = 关2 was obtained by rotating the TEM specimen around the 关0001兴 axis by 30° ¯ 110兴 axis, shows the presence of full screw and then slightly around the 关2 dislocations 共with Burgers vector equal to a basal-plane unit vector兲.
served in the wing region. This unexpected feature is due to the different nature of SFs in the window and mask regions. A detailed TEM analysis on a thinner sample area is shown in Fig. 3, where bright-field TEM images taken under three diffraction conditions show the spatial distribution of SFs, Frank-type partial dislocations 共involving an extra basal plane兲, and full basal-plane dislocations. The distribution of SFs in Fig. 3共a兲 is similar to that in Fig. 2共c兲, with a high density in the window region and a significantly lower density in the wing region. Figures 3共b兲 and 3共c兲 show that the density of dislocations is high in the window region and is very low in the wing region. The dislocations in the window region are observed with higher definition in Fig. 4. They are identified as either partial dislocations that terminate or bound the SFs, or as full TDs with Burgers vector equal to a basal-plane unit vector. The distribution of extended defects in Fig. 3 can explain the observed monochromatic CL images in Fig. 2. The window region contains a high density of full basal-plane dislocations and of SFs that are short and each bounded by two partial dislocations. This combination of full and partial dislocations is related to the low luminescence intensity of the region. The wing region 共grown over the mask兲 in Fig. 3 exhibits long SFs with a much lower density of full and partial dislocations, and it is associated with a strong DBE
FIG. 2. Correlation between optical and structural properties of the same region as in Fig. 1. 共a兲 Monochromatic CL images corresponding to emission peaks of 共a兲 the 共D0 , X兲 共3.465 eV兲, and 共b兲 BSFs 共3.41 eV兲. 共c兲 Two¯ 00兴. The width of the TEM beam bright-field TEM image taken under g = 关11 image corresponds to the width of the CL images. Downloaded 02 Jun 2008 to 62.141.180.1. Redistribution subject to AIP license or copyright; see http://apl.aip.org/apl/copyright.jsp
171904-3
Appl. Phys. Lett. 92, 171904 共2008兲
Wu et al.
FIG. 4. Enlarged images from the rectangular region in Fig. 3 taken under TEM conditions that highlight specific extended defects. 共a兲 Stacking faults. 共b兲 Threading partial dislocations that laterally terminate the stacking faults. 共c兲 Threading full dislocations. Their apparent tilt is due to the orientation of the crystal at the imaging conditions described in Fig. 3.
and a distinct SF emission observed in the monochromatic images. There is a clear anti-correlation between the presence of dislocations and the DBE and stacking luminescence. In Fig. 5, we show the segmented nature of the BSFs in the window region. The dimensions of the SFs on the basal plane range from several hundred nanometers to several microns. They are mostly of I1 type, which has a fault transla¯ 3兴 and are bounded by partial dislotional vector of 1 / 6关022 cations. The Burgers vector of those dislocations has two components: one normal 共Frank partial兲 and one parallel 共Shockley partial兲 to the basal plane. A high-resolution TEM image of a terminated SF is shown in the inset of Fig. 5. In contrast, even though the density of SFs in the wing region is large, the density of partial dislocations is very low due to the longer dimensions 共over tens of microns兲 of the SFs. The different dimensions of the SFs in the window and wing regions depend on the SF nucleation mechanism. In the window region, a high density of slightly misaligned islands
is formed in the early stages of growth of GaN on the r-sapphire substrate as a result of the high lattice mismatch. Any imperfection or impurity on the substrate surface could generate a fault in the stacking sequence, resulting in a SF within the island. The probability of such fault formation is low. A single mistake in the nucleation results in an I1 SF 共involving a single violation of the stacking sequence兲. Two sequential mistakes result in an I2 SF and three sequential mistakes in E SF 共also referred as extrinsic SF兲. These multiple faults occur with much lower probabilities.11 The higher probability of formation should result in a large density of short I1 SFs, whose extension lengths reflect the dimensions of the nucleation islands. In the wing region, growth is along the c direction and proceeds by step nucleation and fast lateral growth. With growth along the c axis, chances of a mistake in the stacking sequence are very small 共this statement is supported by the fact that SFs are seldom observed in the c plane epitaxy兲. However, when they happen, they will be single step mistakes, resulting again in SFs of the I1 type. The SF dimensions will now be large due to low basal-plane ¯ 00典 nucleation step rates and fast growth in the lateral 具11 directions. In summary, an anticorrelation between the light emission from radiative centers and the distribution of dislocations in the a-plane GaN ELO films grown on r-sapphire substrates have been established by CL and TEM. It reveals that the extension lengths of BSFs in the window region are shorter in comparison to the wing region. This is attributed to the island growth mode in the early stages of GaN growth on r-sapphire in the window region, and the lateral overgrowth in the wing region. The authors gratefully acknowledge support from a grant from Nichia Corporation for the work at Arizona State University and the fruitful cooperation with F. Bertram at University of Magdeburg. 1
FIG. 5. Plan-view TEM image of the a-plane GaN films, showing the segmented nature of stacking faults in the window region. The termination of an I1 stacking fault by a Frank–Shockley partial dislocation is shown in the high-resolution TEM image in the inset.
P. Waltereit, O. Brandt, A. Trampert, H. T. Grahn, J. Menniger, M. Ramsteiner, M. Reiche, and K. H. Ploog, Nature 共London兲 406, 865 共2000兲. 2 A. Chitnis, C. Chen, V. Adivarahan, M. Shatalov, E. Kuokstis, V. Mandavilli, J. Yang, and M. A. Khan, Appl. Phys. Lett. 84, 3663 共2004兲. 3 B. A. Haskell, S. Nakamura, S. P. DenBaars, and J. S. Speck, Phys. Status Solidi B 244, 2847 共2007兲. 4 M. D. Craven, S. H. Lim, F. Wu, J. S. Speck, and S. P. DenBaars, Appl. Phys. Lett. 81, 469 共2002兲. 5 S. M. Bishop, J.-S. Park, J. Gu, B. P. Wagner, Z. J. Reitmeier, D. A. Batchelor, D. N. Zakharov, Z. Liliental-Weber, and R. F. Davis, J. Cryst. Growth 300, 83 共2007兲. 6 M. D. Craven, S. H. Lim, F. Wu, J. S. Speck, and S. P. DenBaars, Appl. Phys. Lett. 81, 1201 共2002兲. 7 M. D. Craven, F. Wu, A. Chakraborty, B. Imer, U. K. Mishra, S. P. DenBaars, and J. S. Speck, Appl. Phys. Lett. 84, 1281 共2004兲. 8 Z. Bougrioua, M. Laügt, P. Vennéguès, I. Cestier, T. Gühne, E. Frayssinet, P. Gibart, and M. Leroux, Phys. Status Solidi A 204, 282 共2007兲. 9 R. Liu, A. Bell, F. A. Ponce, C. Q. Chen, J. W. Yang, and M. A. Khan, Appl. Phys. Lett. 86, 021908 共2005兲. 10 J. Mei, S. Srinivasan, R. Liu, F. A. Ponce, Y. Narukawa, and T. Mukai, Appl. Phys. Lett. 88, 141912 共2006兲. 11 J. P. Hirth and J. Lothe, Theory of Dislocation, 2nd ed. 共Wiley, New York, 1982兲, p. 354.
Downloaded 02 Jun 2008 to 62.141.180.1. Redistribution subject to AIP license or copyright; see http://apl.aip.org/apl/copyright.jsp
