Fabrication and Characterization of GaN-Based LEDs Grown
on Chemical Wet-Etched Patterned Sapphire Substrates
Y. J. Lee, H. C. Kuo, T. C. Lu, B. J. Su, and S. C. Wang
Department of Photonics and Institute of Electro-Optical Engineering, National Chiao Tung University, Hsinchu 300, Taiwan
Characteristics of GaN-based light-emitting diodes 共LEDs兲 grown on the chemical wet-etched patterned sapphire substrates 共CWE-PSS兲 with different crystallography-etched facets were investigated. An improvement of 40% on the overall quantum efficiency was achieved by adopting this CWE-PSS scheme. A Monte Carlo ray-tracing method was employed to derive the optimized condition of sapphire etching time, and the calculated result demonstrated the same trend with real device measurement. Adopting the CWE-PSS in LEDs could not only improve the epitaxial quality but also increase the extraction quantum efficiency due to crystallography-etched facets efficiently scattering the guided light to enter the escape cone on the top of device surfaces. Finally, we observed better aging behaviors of CWE-PSS LEDs, which could be due to the reduction of threading dislocations of the epitaxial layers.
© 2006 The Electrochemical Society. 关DOI: 10.1149/1.2359701兴 All rights reserved.
Manuscript submitted July 4, 2006; revised manuscript received August 9, 2006. Available electronically October 23, 2006.
III-V nitride wide-bandgap materials have attracted considerable attention in recent years.1-3The bandgap energy of AlInGaN varies from 0.8 to 6.2 eV depending on its composition. Therefore, the high-brightness light emitting diodes共LEDs兲 in these wavelength regions have been widely employed on versatile applications, such as traffic signals, back side lighting in liquid crystal display共LCD兲, and illumination lighting.4-6In general, GaN-based LEDs are grown on the top of the sapphire substrate, and a high dislocation density in the order of 108–1010cm−2is induced due to the large mismatch of lattice constant and thermal expansion between the epitaxial GaN film and the underneath sapphire substrate.7The large order of mag-nitude of dislocation density suppresses the further performance of GaN-based LEDs. Additionally, due to the significant difference of the refractive index between the GaN-based material and air, the light extraction efficiency is limited by the total internal reflection. Approximately 1/共4n2兲 of light from the active region can escape
from the top and bottom of the device, where n denotes the refrac-tive index of a semiconductor material.8Even though GaN has a lower refractive index共n ⬇ 2.5兲 than that of other semiconductor materials, only about 4% of the total emitted light can be extracted from one face according to above equation. Therefore, the major effort in fabricating LEDs is then how to get the photons that had been generated inside the active region out of the semiconductor layers. Our group has proposed several methods to enhance light extraction efficiency, including surface roughness and shaping.9-15 Recently, the single-step growth on the maskless patterned sapphire substrate共PSS兲 fabricated by dry etching was proposed and a con-siderable improvement on both internal quantum efficiency and light extraction efficiency was demonstrated.16-21Although previous in-vestigations supported the contribution of the PSS fabricated by dry etching, the sapphire surfaces are unavoidably damaged during the dry etching process. Thus, the threading dislocation would easily propagate to the top epitaxial films through GaN layers deposited on the sidewalls of the sapphire patterns whose surfaces have already been damaged, limiting further improvement of epitaxial quality.22 In this report, we propose a patterned sapphire substrate fabricated by a chemical wet etching technique; hence, the surface damage on the sapphire substrate mentioned above could be eliminated. The details of fabrication and characterization of GaN-based LEDs grown on this novel chemical wet-etched patterned sapphire sub-strate共CWE-PSS兲 are discussed and the optimized dimension of the CWE-PSS is also be calculated by the Monte Carlo ray-tracing method.
Device Fabrication
The GaN-based LEDs used in this study were grown using a low-pressure metallorganic chemical vapor deposition 共Aixtron 2600G兲 system onto the C-face 共0001兲 2 in. diam CWE patterned
sapphire substrates. The LED layer structure was comprised of a 30 nm thick GaN nucleation layer, a 2m thick undoped GaN layer, a 2m thick Si-doped n-type GaN cladding layer, an unin-tentionally doped active region of 450 nm emitting wavelength with five periods of InGaN/GaN multiple quantum wells共MQWs兲, and a 0.2m thick Mg-doped p-type GaN cladding layer. The as-grown wafer was then patterned with square mesas of 350⫻ 350 m2size
by a standard photolithographic process and was partially etched until the exposure of n-GaN to define the emitting area and the n-electrode a 300 nm thick indium tin oxide共ITO兲 was deposited as the transparent conductive layer, and Cr/Au was then deposited as n and p electrodes and was alloyed at 200°C in N2atmosphere for
5 min. Figure 1 schematically depicts a cross-sectional structure of the GaN-based LED grown on the CWE-PSS. For fabricating the CWE-PSS, the SiO2film with hole patterns of 3m diam and 3 m spacing was deposited onto the sapphire substrate by plasma-enhanced chemical vapor deposition共PECVD兲 to serve as the wet etching mask. The sapphire substrate was then wet etched using an H3PO4-based solution at an etching temperature of 300°C. The
wet-etching rate for sapphire substrates was about 1m/min in this study and can be related to the H3PO4 composition and etching temperature.23,24 Figure 2a and b shows the scanning electron mi-croscopy共SEM兲 images of the patterned sapphire substrate of the etching time of 90 and 120 s, respectively. In Fig. 2a the crystallography-etched pattern of a 共0001兲-oriented sapphire sub-strate has a flat surface of兵0001其 C-plane with a triangle shape in the center. Surrounding the triangle-shaped C-plane are three facets of兵1-102其 R-plane with an angle of 57° against the 关0001兴 C-axis. However, due to the relative fast etching rate of C-plane than that of
Figure 1. 共Color online兲 The schematic drawing of epitaxial layers and device structure with chemical wet-etched patterned sapphire substrate 共CWE-PSS兲.
R-plane, the triangle-shaped flat surface of the兵0001其 C-plane in the pattern center finally vanishes with the increase of etching time. As shown in Fig. 2b, the 兵0001其 C-plane disappeared and only the 兵1-102其 R-plane is observed on the CWE-PSS with the etching time of 120 s. Figure 1c shows the evolution of CWE-PSS with the in-crease of sapphire etching time. The diameter of the sapphire pattern also increases with the increase of etching time due to the side-etching effect; however, the period of the sapphire pattern keeps the same as 6m. Additionally, the high slope 共57°兲 crystallography-etched facet of CWE-PSS is hard to fabricate by dry etching, and it has been demonstrated in our previous work that this inclined facet was crucial for improving light extraction efficiency.15
To compare the LED performances with different crystallography-etched facet patterns, the sapphire substrates of etching times of 0, 30, 60, 90, and 120 s were employed into this paper. All of these CWE-PSSs were then grown and processed at the same time, eliminating any artificial issue during LED fabrication. The LED chips were packaged into TO-18 without epoxy resin for the subsequent measurement. The typical current–voltage 共I–V兲 measurements were performed using a high-current measure unit 共Keithley 240兲. The light output power of the LEDs was measured
tional共sapphire etching time of 0 s兲 and all CWE-PSS LEDs indi-cates that the LED composition and growth rate were not associated with the CWE-PSS.
Figure 4a shows the measurement results of room temperature output power共L–I curve兲 of conventional and CWE-PSS LEDs as a function of the forward-bias current. In this figure, all the CWE-PSS LEDs demonstrate a significant improvement in output power as compared to the conventional LED under our measurement condi-tion up to 200 mA. The enhanced factor of output power of CWE-PSS LEDs compared to the conventional LEDs at a driving current of 20 mA is shown in Fig. 4b. According to this figure, the opti-mized CWE-PSS condition was achieved at an etching time of 90 s, corresponding to an enhanced factor of 1.4. Figure 4c shows the external quantum efficiency共EQE兲 of the conventional and CWE-PSS LEDs with the forward injection currents up to 100 mA. It was found that the EQE of the CWE-PSS LED with etching time of 90 s reached a maximum value of about 25% at an injection current of 5 mA and then decreased significantly with a further increase in the forward bias current. Nearly the same trend was also obtained for the conventional LED sample except for a lower EQE value of 17.8%. The degradation at the higher current might be due to over-flow of injection carriers and the joule heating effect. Even though the CWE-PSS LED performance in absolute terms of external quan-tum efficiency does not exceed state-of-the-art devices using other approaches, comparison is being made on the overall intensity en-hancement using the CWE-PSS scheme.
In order to study the fundamentals of enhancement of light out-put with different etching times of CWE-PSS LEDs, we used a commercial ray-tracing software, Tracepro, employing the Monte Carlo algorithm for forward ray-tracing, various outputs including efficiency value, and spatial distributions of radiometric and photo-metric data. Shape and size of the solid model for the ray-tracing calculation was determined and was exactly the same as the SEM images and microscopic measurements of the geometry of CWE-PSS LEDs, as shown in Fig. 1 and 2. The solid model was built up as a combination of simple solid objects, each semiconductor layer adjacent to the other. According to the recombination process,25 light rays were generated in the active layer with a uniform random distribution. Monochromatic radiation representing the peak wave-length of the measured spectral emission共450 nm兲 was used in the simulation. Figure 5 shows the calculated radiation patterns of共a兲 the conventional and共b兲 CWE-PSS LEDs with etching time of 30 s. A stronger axial radiation on the CWE-PSS LED than that of the conventional LED was observed in this figure. The same epitaxial models were also built on the other CWE-PSS LEDs with the etch-ing time of sapphire substrate of 60, 90, and 120 s. The comparison of overall light extraction efficiency was plotted and shown in Fig. 5c. According to this calculation, the light extraction efficiency is dramatically enhanced with the increasing of sapphire etching time and over twofold of magnitude of light output was observed on the CWE-PSS LED with the etching time of 120 s. Therefore, the crystallography-etched patterns that evolving with the increasing of etching time of sapphire substrate affect the light extraction effi-ciency profoundly. With the increase of etching time, the triangle-shaped flat surface of the 兵0001其 C-plane in the pattern center finally vanishes due to its relative fast etching rate than that of the 兵1-102其 R-plane. The sustained 兵1-102其 R-plane has an inclined Figure 2.共a, b兲 SEM images of the CWE-PSS having etching times of 90
and 120 s, respectively.共c兲 A top-view drawing depicts the evolution of CWE-PSS with the increase of etching time.
crystallography-etched facet with a high slope aslarge as 57°, adding the opportunity of the guided light to meet the escape cone on the top of the chip surface. Figure 6 is a simple schematic ray-tracing of the CWE-PSS LEDs with the increase of sapphire etching time. In the case of the CWE-PSS LED with the large兵0001其 C-plane pat-tern, i.e., a short period of sapphire etching time, the light emitting from the LED active region multiple quantum well 共MQW兲 was
much easier to be guided inside the LED chip, as compared to that of the longer period of etching time, corresponding to the larger surface of high-slope crystallography-etched facets of 兵1-102其 R-plane. As shown in Fig. 6, more guided light can be extracted from the LED top surface, enhancing the total light output power. This is the reason why in Fig. 5b we can observe strong illumination in the axial direction on the CWE-PSS LED.
Figure 3.共Color online兲 High-resolution X-ray measurement 共Bede D1 HR-XRD兲.
Figure 4.共Color online兲 Measurement results of room-temperature output power 共L–I curves兲 of the conventional and CWE-PSS LEDs.共b兲 The enhancement factor on output power while compar-ing the CWE-PSS LEDs to the conventional LED under a drivcompar-ing current of 20 mA.共c兲 EQE of the conventional and CWE-PSS LEDs with the forward injection currents up to 100 mA.
Figure 5. 共Color online兲 Monte Carlo ray-tracing calculated results of radiation patterns of 共a兲 the con-ventional and 共b兲 CWE-PSS LEDs with etching time of 30 s.共c兲 Calcu-lated enhancement of the light ex-traction efficiency with increasing sapphire etching time.
In comparing the ray-tracing calculation and real device mea-surement, enhancement of LEDs with etching time of 120 s was degraded probably due to the nonoptimized metallorganic chemical vapor deposited共MOCVD兲 growth condition. Cross-sectional side-views of SEM images of CWE-PSS LEDs with different etching times are shown in Fig. 7. The crystallography-etched sapphire pat-terns can be buried completely by the GaN epitaxy in all CWE-PSS LEDs, except for the sample of the etching time of 120 s. According to the ray-tracing calculation, the light extraction efficiency was significantly improved with the increasing of sapphire etching time and could be contributed to the high-slope crystallography-etched facet of the 兵1-102其 R-plane. However, the large inclined crystallography-etched surface also indicates the deep depth of the sapphire pattern, and it also takes more effort for adjusting the growth condition to obtain a high-quality GaN film. As shown in Fig. 7d, a void locating inside the sapphire pattern can be observed due to the relative difficulty for MOCVD to grow on this deep and inclined crystallography-etched facet. Therefore, in the Fig. 4b, a drop of light extraction efficiency was observed on the CWE-PSS LED of sapphire etching time of 120 s. In addition, we did not consider the surface morphology while building the ray-tracing cal-culation model, where surfaces of LED chips were assumed as per-fectly flat. However, as can be seen in Fig. 7, surfaces of LED chips were quite rough; the contribution to output power by adopting the CWE-PSS scheme would be somehow eliminated as compared to the actual device performance in Fig. 4b to the calculation result in Fig. 5c. Nevertheless, by ignoring these epitaxial issues as
men-tioned above, both the ray-tracing calculation and real device mea-surement depict the same trend on the enhancement of light extrac-tion efficiency, indicating that the modeling of the LED chip by ray-tracing calculation can be a powerful tool in predicting the effi-ciency of LED optics designs.
An aging test was performed on the conventional and CWE-PSS LEDs under a driving current of 50 mA at 55°C. In Fig. 8, the EL intensity to the initial EL intensity is shown as a function of aging time. According to this figure, all the aging samples exhibit the same degradation trend. However, all the CWE-PSS LEDs present a gradual degradation in the EL intensity under our measurement con-dition up to 600 h. In general, the EL intensity of conventional and CWE-PSS LEDs were decayed by about 20 and 10%, respectively; indicating that improvement on the epitaxial quality could be achieved via growth on the CWE-PSS scheme. The better aging behavior of CWE-PSS LEDs could be due to the reduction of threading dislocations of the epitaxial layers. Typically, by employ-ing a patterned sapphire scheme, the epitaxial mode would be shifted from the initial 3D-dominated mode to 2D mode, enforcing threading dislocations bending toward lateral directions, thus im-proving overall device performances.16-22 To further identify whether the CWE-PSS scheme could benefit from reduction of dis-location density or the enhancement of internal quantum efficiency, detailed experiments, such as transmission electron microscope or direct measurement of internal quantum efficiency will be investi-gated in the future.
Conclusion
In summary, the characteristics of GaN-based LEDs grown on patterned sapphire substrate fabricated by chemical wet etching were specifically analyzed. By chemical wet etching, the sapphire substrate exhibited a particular crystallography-etched facet of the 兵1-102其 R-plane with an inclined slope as large as 57°, facilitating a significant enhancement of light extraction efficiency. A Monte Carlo ray-tracing calculation was employed to further investigate the geometric patterns of this novel CWE-PSS LED, and a similar trend was observed between theoretical calculation and the device mea-surement. Moreover, an improvement of epitaxial quality was also observed on CWE-PSS LEDs, according to device aging results. Therefore, by using this novel CWE-PSS scheme, an overall en-hancement of 40% on the quantum efficiency can be achieved that could contribute not only to the improvement of epitaxial quality, but also to the geometrical shape of the inclined crystallography-etched facets, efficiently scattering the guided light to enter the es-cape cone.
Acknowledgment
The authors thank Professor K. M. Lau of Hong Kong University of Science and Technology and Dr. T. C. Hsu and Dr. M. H. Hsieh of Epistar for useful discussion. This work was supported by the Na-tional Science Council of Republic of China 共R.O.C.兲 in Taiwan Figure 6. 共Color online兲 A schematic ray-tracing of the CWE-PSS LEDs
with increasing sapphire etching time.
Figure 7. Cross-sectional side-view SEM images of CWE-PSS LEDs with different etching times of共a兲 30, 共b兲 60, 共c兲 90, and 共d兲 120 s.
Figure 8.共Color online兲 Reliability test of the conventional and CWE-PSS LEDs under stress conditions of 55°C and 50 mA.
6. J. Jiao and B. Wang, Proc. SPIE, 5187, 234共2004兲.
7. S. D. Lester, F. A. Ponce, M. G. Craford, and D. A. Steigerwald, Appl. Phys. Lett.,
66, 1249共1995兲.
8. M. Broditsky and E. Yablonovitch, Proc. SPIE, 3002, 119共1997兲.
9. C. C. Kao, H. C. Kuo, H. W. Huang, J. T. Chu, Y. C. Peng, Y. L. Hsieh, C. Y. Luo, and S. C. Wang, IEEE Photon. Technol. Lett., 17, 19共2005兲.
10. H. W. Huang, C. C. Kao, J. T. Chu, H. C. Kuo, S. C. Wang, and C. C. Yu, IEEE
Photon. Technol. Lett., 17, 983共2005兲.
11. Y. J. Lee, H. C. Kuo, S. C. Wang, T. C. Hsu, M. H. Hsieh, M. J. Jou, and B. J. Lee,
IEEE Photon. Technol. Lett., 17, 2289共2005兲.
12. Y. J. Lee, H. C. Tseng, H. C. Kuo, S. C. Wang, C. W. Chang, T. C. Hsu, Y. L. Yang,
20. Y. J. Lee, T. C. Hsu, H. C. Kuo, S. C. Wang, Y. L. Yang, S. N. Yen, Y. T. Chu, Y. J. Shen, M. H. Hsieh, M. J. Jou, and B. J. Lee, Mater. Sci. Eng., B, 122, 184 共2005兲.
21. Y. J. Lee, J. M. Hwang, T. C. Hsu, M. H. Hsieh, M. J. Jou, B. J. Lee, T. C. Lu, H. C. Kuo, and S. C. Wang, IEEE Photon. Technol. Lett., 18, 1152共2006兲. 22. J. Wang, L. W. Guo, H. Q. Jia, Z. G. Xing, Y. Wang, H. Chen, and J. M. Zhou, Jpn.
J. Appl. Phys., Part 2, 44, L982共2005兲.
23. S. J. Kim, Jpn. J. Appl. Phys., Part 1, 44, 2921共2005兲. 24. S. J. Kim, IEEE Photon. Technol. Lett., 17, 1617共2005兲.
25. B. A. E. Saleh and M. C. Teich, Fundamental of Photonics, p. 602, John Wiley, New York共1991兲.
