Effects of Laser Sources on Damage Mechanisms and Reverse-Bias Leakages of Laser Lift-Off GaN-Based LEDs

(1)

Effects of Laser Sources on Damage Mechanisms and

Reverse-Bias Leakages of Laser Lift-Off GaN-Based LEDs

Ji-Hao Cheng,a,

*

YewChung Sermon Wu,a,

**

,zWei Chih Peng,a and

Hao Ouyangb

a

Department of Materials Science and Engineering, National Chiao Tung University, Hsinchu 300, Taiwan

b

Department of Materials Engineering, National Tsing Hua University, Hsinchu 300, Taiwan

The frequency-tripled neodymium-doped yttrium aluminum garnet laser共355 nm兲 and the KrF pulsed excimer laser 共248 nm兲 were employed to separate GaN thin films from sapphire substrates and to transfer the films to bond with other substrates. The different laser lift-off processes would generate the dislocation density on different regions. In this study, the effects of these two laser sources on structural damage mechanisms and reverse-bias leakages of InGaN–GaN light-emitting diodes 共LEDs兲 were studied.

Manuscript submitted April 3, 2009; revised manuscript received May 4, 2009. Published June 12, 2009.

Light-emitting diodes共LEDs兲 are expected to play the most im-portant role in next-generation light source due to their advantages of long life, high efficiency, small size, various colors, and wide applications. In particular, the high brightness GaN-based LEDs have attracted considerable attention for white light solid-state light-ing. GaN-based LEDs usually were grown on sapphire and have improved quickly due to rapid advances in growth techniques. How-ever, owing to the poor thermal and electrical conductivity of sap-phire substrate, which affects the LED efficiency and reliability, the GaN layer usually has a high defect density.

The current progress may be further extended by combining the GaN-based LEDs with other materials such as Si, Cu, and diamond. A less direct integration method involves laser lift-off 共LLO兲 and transfer for bonding with dissimilar materials.1-3 Several laser sources including the frequency-tripled neodymium-doped yttrium aluminum garnet共Nd:YAG兲 laser 共355 nm兲4,5 and the KrF pulsed excimer laser 共248 nm兲 have been demonstrated to successfully separate GaN thin films from sapphire substrates.6 During laser scanning, the GaN layer will decompose into Ga metal and nitrogen gas. Consequently, the freestanding GaN layer is separated from the sapphire. These LLO processes would degrade the performance of InGaN–GaN LEDs. In this study, the effect of these two laser sources on the defect generation mechanisms and reverse-bias leak-ages of InGaN–GaN LEDs was studied.

Experimental

In this study, three types of LEDs were investigated. Samples designated as “CV-LED” were conventional LEDs without any laser treatment. Samples designated as “KrF-LED” were CV-LEDs treated with KrF pulsed excimer laser, while “YAG-LEDs” were CV-LEDs treated with Nd:YAG laser. The basic fabrication pro-cesses of these LEDs were almost the same.7The LED structures were grown by low pressure metallorganic chemical vapor deposi-tion. The structures comprised a 5 nm thick Si-doped n+_-InGaN

layer, a 400 nm thick Mg-doped p-GaN layer, an InGaN–GaN mul-tiple quantum well共MQW兲, a 2 ␮m thick Si-doped n-GaN layer, a 2␮m thick undoped-GaN layer film, and a buffer layer on the sap-phire substrate.

For the CV-LED, the device mesa with a chip size of 350 ⫻ 350 ␮m was defined by an inductively coupled plasma which removed the Mg-doped GaN and MQW until the Si-doped GaN was exposed. Then, the indium tin oxide共ITO兲 layer was deposited on the n+_{-InGaN layer using an E-beam coater to form a p-side contact}

layer and a current spreading layer. The Cr/Au layer was deposited onto the ITO layer to form the p- and n-side electrodes.

The fabrication processes of KrF-LED and YAG-LED are shown in Fig.1. Before the LLO process, the CV-LED wafer was bonded to a host substrate covered with an adhesive/glue layer, and the sapphire substrate of KrF-LED was polished using a diamond paper. The KrF laser beam spot size of 700⫻ 700 ␮m 共four chips were lifted off in etch pulse兲 was scanned without overlap from pulse to pulse. The 355 nm Nd:YAG LLO system was constructed by a se-ries of lens, the laser beam was near circle shape, and the lateral energy of the converged laser beam was a Gaussian distribution. The Nd:YAG laser beam spot size of 500␮m was scanned with a 20% overlap from pulse to pulse.

The pulse length of the KrF laser was 35 ns, which was longer than that of the YAG laser共5 ns兲. The energy densities of the KrF laser were in the range between 700 and 1000 mJ/cm2_{, while those}

of the YAG laser were in the range between 100 and 400 mJ. These wafers were then bonded to a sapphire substrate with an adhesive layer at 200°C for 60 min with a comprehensive load of 10 kg/cm2. The host substrate and glue layer were subsequently removed.

Results and Discussion

After LLO, for KrF-LED, when the laser energy was below 700 mJ/cm2_{, no visible alteration of the GaN layer was observed.}

*Electrochemical Society Student Member. **Electrochemical Society Active Member.

z

E-mail: [email protected]

Figure 1. 共Color online兲 Schematic diagram of YAG-LED and KrF-LED transfer process:共a兲 CV-LED bonding to host substrate, 共b兲 LLO process, 共c兲 bonding to sapphire substrate, and共d兲 removal of host substrate and glue layer.

Journal of The Electrochemical Society, 156共8兲 H640-H643 共2009兲

0013-4651/2009/156共8兲/H640/4/$25.00 © The Electrochemical Society

H640

) unless CC License in place (see abstract). ecsdl.org/site/terms_use

address. Redistribution subject to ECS terms of use (see 140.113.38.11

(2)

The GaN layers were lifted off when the laser intensity was above this critical energy. However, when the laser energy exceeded 900 mJ/cm2_{, the absorbed photon energy led to local heating of the}

layer above the critical sublimation temperature of Ga, causing the destruction of the GaN layer. As a result, most LED devices failed after such high energy laser treatment.7Only those devices with a median energy density共800 mJ/cm2_{兲 had good yields 共 ⬎ 90 % 兲.}

As for YAG-LED, the median energy density of the YAG laser was 200 mJ/cm2.7 These two laser energy intensities were utilized to study the effects of various laser sources on the damage mechanisms and reverse-bias leakages of LEDs.

Figure2shows the current–voltage共I-V兲 characteristics of the LEDs. The forward voltages of both KrF-LED and YAG-LED were about 3.2 V at 20 mA, which was similar to that of CV-LED 共3.2 V兲, indicating that the transfer method did not change the LED performance much. The light intensity of CV-LED was 96 mcd共at 20 mA兲, which was larger than the KrF-LED 共84 mcd兲 and YAG-LED共81 mcd兲. This is because, when the light passes through the bonded interface, the Fresnel losses resulting from the GaN/ adhesive layer and adhesive layer/sapphire interfaces might have a negative effect on the luminance intensity. Besides, during the bond-ing process, we might create interfacial defects at the bonded inter-face. These defects have a negative effect on the optical properties. However, the reverse-bias leakage currents of three LEDs were quite different. Figure3shows the reverse-bias leakage currents of LEDs up to −5 V. Leakage currents increased with laser energy densities. Under a reverse bias of −5 V, the leakage current of YAG-LED was 1.65⫻ 103_{nA, which was 10,000 times higher than that}

of the KrF-LED共0.17 nA兲 and 33,000 times higher than that of the CV-LED 共0.05 nA兲. Because these LED devices were fabricated from the same wafer, these degradations must be caused by the LLO processes, which generated screw dislocations. These dislocations had been demonstrated as the primary culprit in reverse-bias leakage paths in GaN.8-12

A transmission electron microscopy共TEM兲 image with a two-beam condition was employed to identify screw dislocations. The low loss region of transmission electron-energy-loss spectroscopy was utilized to estimate the sample thickness to obtain accurate dis-location density.13The distribution of dislocation densities of these two lift-off LEDs was quite different. Figure 4 shows the cross-sectional TEM images of LEDs. Figure4shows the cross-sectional TEM images of LEDs with a共0002兲 two-beam condition to observe the screw dislocation. The dislocation densities were obtained by counting the number of dislocations and then dividing it by the

thickness and width of TEM sampling regions. In this case, those dislocations that penetrated through the MQW region were counted because they could affect the leakage current of the LED. The screw dislocation density in the MQW region of YAG-LED 共2.9 ⫻ 109_cm−2_{兲 was much higher than that of KrF-LED 共3.75}

⫻ 108_cm−2_{兲. Therefore, the reverse-bias leakage current of}

YAG-LED was higher than that of KrF-YAG-LED.

Figures5aand6ashow the enlarged superficial region of YAG-LED and KrF-YAG-LED, respectively. The dislocation density near the LLO surface of KrF-LED was higher than that of YAG-LED. Their relative dislocation densities were schematically illustrated in Fig.7. The MQW dislocation density of KrF-LED was lower than that of YAG-LED. The LLO surface dislocation density of KrF-LED was higher than that of YAG-LED.

The difference in dislocation densities of the two LEDs in the MQW region is because the absorption coefficient of GaN at 248 nm共for KrF-LED兲 is 2 ⫻ 105cm−1, which is 3.33 times higher than that at 355 nm 共for YAG-LED兲共6 ⫻ 104_cm−1_兲.14

In other words, the absorption depth of the YAG laser was much thicker than that of the KrF excimer laser. The absorption of laser energy induces a highly localized, rapid, thermal decomposition of GaN. This de-composition which introduces a biaxial compressive stress can cause deformation共formation of dislocations兲 of the GaN layer. Compared to the KrF excimer laser, the YAG laser had a thicker absorption depth, which causes more dislocations to be formed in the MQW region, resulting in an increase in reverse-bias leakage current.

Besides dislocations, serious stacking faults were observed within approximately 40 nm of the LLO interface, as shown in Fig.

5band6b. The detailed analysis was displayed in Fig.5candd,6c, andd, which show the computed fast Fourier transforms taken from different parts of Fig.5band6bindicated by the arrows. Figures5d

and6dshow regular diffraction spots indicating good crystal quality. In Fig.5cand6c, the共0001兲 and 共0001¯兲 spots were pulled into lines toward the center spot, meaning the existence of a huge amount of stacking faults. Besides, some twin spots were also observed in this region. The presence of the stacking faults at the same distance indicates the occurrence of LLO at the GaN/sapphire interface for both KrF and YAG laser sources.15,16

Figures5eand6eshow the enlargement of the part of the lattice image approximately 200 nm below the LLO surface. Figure 5e

共YAG-LED兲 shows a regular lattice image, while Fig.6e共KrF-LED兲

shows a distorted image. There are obviously different microstruc-tures at 200 nm below the LLO surface; therefore, the inverse fast Fourier transform共IFFT兲 was adopted to further analyze this region. They were analyzed by IFFT. Figure6gdisplays the processed IFFT image of Fig. 6f 共200 nm below the LLO surface of KrF-LED兲

Figure 2. 共Color online兲 I-V curve of forward voltage characteristics of KrF-LED, YAG-LED, and CV-LED.

Figure 3.共Color online兲 Leakage currents of LEDs under a reverse bias of −5 V.

H641

Journal of The Electrochemical Society, 156共8兲 H640-H643 共2009兲 H641

(3)

using the diffraction spots along the关0002兴 direction. KrF-LED has sequential series dislocations, which have the same center, indicated by the dashed white arc lines. The processed IFFT image of YAG-LED did not show such series dislocations.

Similar results have been reported by Chen et al.16who analyzed GaN LLO surfaces using a single KrF excimer laser pulse with an energy density of 600 mJ/cm2_{. GaN surface regions were}

character-ized by cross-sectional high resolution transmission electron micros-copy 共HRTEM兲 and fitted using the model of stress waves. They found, in addition to the superficial damages caused by laser absorp-tion, a cluster of half-loops共sequential series dislocations兲 near the LLO interface, which were generated by a laser-induced shock wave. The stress wave model analysis revealed that when the pulse duration exceeded the critical time共10 ns兲, plastic waves were

gen-erated and converged at a position accompanied by stress saltation. The shock-wave-generating position was about 200 nm away from the LLO surface.

In our study, because the KrF-laser pulse time共35 ns兲 was longer than the critical time, a cluster of half-loops was also found at about 200 nm away from the LLO surface共Fig.5c兲. However, no loops

were found near the LLO surface of YAG-LED because the pulse time共5 ns兲 was shorter than the critical time. As a result, the screw dislocation density near the LLO surface of KrF-LED was higher than that of YAG-LED.

The TEM observation and the effect of absorption constant are summarized in Fig.7. Compared to YAG laser, KrF excimer laser had a higher absorption coefficient of GaN共shallower penetration depth兲. Most KrF laser energy was absorbed around the GaN/ sapphire interface. However, the longer pulse time brought about plastic and shock wave, which caused the dense dislocation and deformed the superficial structure. As for YAG-LED, owing to the lower absorption coefficient, the Nd:YAG laser had deeper penetra-tion depth, which caused more dislocapenetra-tions in the bulk region and penetrated the MQW region 共Fig. 4b兲 and led to an increase in

reverse-bias leakage current.

Conclusion

In this study, the effects of the frequency-tripled Nd:YAG laser 共355 nm兲 and the KrF pulsed excimer laser 共248 nm兲 on the struc-tural damage mechanisms and reverse-bias leakages of GaN-based LEDs were investigated. The absorption depth of the YAG laser was much thicker than that of the KrF excimer laser because the absorp-tion coefficient of GaN at 248 nm is 3.33 times higher than that at 355 nm. As a result, the MQW screw dislocation density and the reverse-bias leakage current of YAG-LED were much higher than those of KrF-LED. Most KrF laser energy was absorbed around the GaN/sapphire interface. Consequently, the LLO surface dislocation density of KrF-LED was higher than that of YAG-LED. Moreover, the longer pulse time of the KrF laser共35 ns兲 brought about plastic and shock waves, which caused dense dislocation and deformed the superficial structure.

Figure 4.共Color online兲 TEM images of the whole LED structure: 共a兲 KrF-LED and共b兲 YAG-LED.

Figure 5. TEM images of YAG-LED:共a兲 bright-field cross-sectional image of the YAG-LED LLO interface,共b兲 magnified part of the superficial region from共a兲, 关共c兲 and 共d兲兴 fast Fourier transforms of different regions indicated by arrows, and共e兲 HRTEM lattice image of the region 200 nm below the LLO interface.

H642 Journal of The Electrochemical Society, 156共8兲 H640-H643 共2009兲

H642

(4)

Acknowledgment

This project was funded by Epistar Corporation, Sino American Silicon Products Incorporation, and the National Science Council of the Republic of China under grant no. 95-2221-E009-087-MY3. Technical support from the National Nano Device Laboratory, Cen-ter for Nano Science and Technology, Nano Facility CenCen-ter, and Semiconductor Laser Technology Laboratory of the National Chiao Tung University is also acknowledged. The authors thank H. C. Kuo for valuable discussions.

National Chaio Tung University assisted in meeting the publication costs of this article.

References

1. W. Y. Lin, D. S. Wuu, K. F. Pan, S. H. Huang, C. E. Lee, W. K. Wang, S. C. Hsu, Y. Y. Su, S. Y. Huang, and R. H. Horng, IEEE Photonics Technol. Lett., 17, 1809 共2005兲.

2. O. B. Shchekin, J. E. Epler, T. A. Trottier, T. Margalith, D. A. Steigerwald, M. O. Holcomb, P. S. Martin, and M. R. Krames, Appl. Phys. Lett., 89, 071109共2006兲. 3. M. R. Krames, O. B. Shchekin, R. Mueller-Mach, G. O. Mueller, L. Zhou, G.

Harbers, and M. G. Craford, J. Disp. Technol., 3, 160共2007兲. 4. W. C. Peng and Y. S. Wu, Appl. Phys. Lett., 88, 181117共2006兲.

5. M. K. Kelly, O. Ambacher, R. Dimitrov, R. Handschuh, and M. Stutzmann, Phys. Status Solidi A, 159, R3共1997兲.

6. W. S. Wong, T. Sands, and N. W. Cheung, Appl. Phys. Lett., 72, 599共1998兲. 7. Y. S. Wu, J. H. Cheng, W. C. Peng, and H. Ouyang, Appl. Phys. Lett., 90, 251110

共2007兲.

8. P. Kozodoy, J. P. Ibbetson, H. Marchand, P. T. Fini, S. Keller, J. S. Speck, S. P. DenBaars, and U. K. Mishra, Appl. Phys. Lett., 73, 975共1998兲.

9. J. W. P. Hsu, M. J. Manfra, D. V. Lang, S. Richter, S. N. G. Chu, A. M. Sergent, R. N. Kleiman, L. N. Pfeiffer, and R. J. Molnar, Appl. Phys. Lett., 78, 1685共2001兲. 10. D. S. Li, H. Chen, H. B. Yu, H. Q. Jia, Q. Huang, and J. M. Zhou, J. Appl. Phys.,

96, 1111共2004兲.

11. J. W. P. Hsu, M. J. Manfra, R. J. Molnar, B. Heying, and J. S. Speck, Appl. Phys. Lett., 81, 79共2002兲.

12. B. S. Simpkins, E. T. Yu, P. Waltereit, and J. S. Speck, J. Appl. Phys., 946, 1448 共2003兲.

13. R. F. Egerton, Electron Energy-Loss Spectroscopy in the Electron Microscope, 2nd ed., Plenum, New York共1996兲.

14. J. F. Muth, J. H. Lee, I. K. Shmagin, R. M. Kolbas, H. C. Casey, Jr., B. P. Keller, U. K. Mishra, and S. P. DenBaars, Appl. Phys. Lett., 71, 18共1997兲.

15. E. A. Stach, M. Kelsch, E. C. Nelson, W. S. Wong, T. Sands, and N. W. Cheung, Appl. Phys. Lett., 77, 1819共2000兲.

16. W. H. Chen, X. N. Kang, X. D. Hu, R. Lee, Y. J. Wang, T. J. Yu, Z. J. Yang, G. Y. Zhang, L. Shan, K. X. Liu, et al., Appl. Phys. Lett., 91, 121114共2007兲. Figure 6. TEM images of the KrF-LED:共a兲 bright-field cross-sectional

im-age of the KrF-LED LLO interface,共b兲 magnified part of the superficial region from共a兲, 关共c兲 and 共d兲兴 fast Fourier transforms of different regions indicated by arrows,共e兲 HRTEM lattice image of the region 200 nm below the LLO interface,共f兲 TEM image of the region 200 nm from 共a兲, 共g兲 in-versed fast Fourier transform image of共f兲 using diffraction spots along the 关0002兴 direction, and 共h兲 HRTEM lattice image of the region indicated by arrow.

Figure 7.共Color online兲 Schematic diagram of screw dislocation generation after the lift-off process.

H643

Journal of The Electrochemical Society, 156共8兲 H640-H643 共2009兲 H643