Enhanced light output power of thin film GaN-based high voltage light-emitting diodes Ching-Ho Tien,1 Ken-Yen Chen,1 Chen-Peng Hsu,2 and Ray-Hua Horng3,4,* 2
1 Department of Materials Science and Engineering, National Chung Hsing University, Taichung 40227, Taiwan Electronics and Opto-electronics Research Laboratories, Industrial Technology Research Institute, Hsinchu 31040, Taiwan 3 Graduate Institute of Precision Engineering, National Chung Hsing University, Taichung 40227, Taiwan 4 Advanced Optoelectronic Technology Center, National Cheng Kung University, Tainan 70101, Taiwan * [email protected]
Abstract: The characteristics of high-voltage light-emitting diodes (HVLEDs) consisting of a 64-cell LED array were investigated by employing various LED structures. Two types of HVLED were examined: a standard HVLED with a single roughened indium tin oxide (ITO) surface grown on a sapphire substrate and a thin-film HVLED (TF-HVLED) with a roughened n-GaN and ITO double side transferred to a mirror/silicon substrate. At an injection current of 24 mA, the output powers of the HVLEDs fabricated using a sapphire substrate and those fabricated using a mirror/silicon substrate were 170 and 216 mW, respectively. Because the TF-HVLED exhibited improved thermal dissipation and light extraction, it produced a greater output power than the HVLED fabricated using the sapphire substrate did. ©2014 Optical Society of America OCIS codes: (230.2090) Electro-optical devices; (230.3670) Light-emitting diodes
References and links 1.
K. C. Shen, W. Y. Lin, D. S. Wuu, S. Y. Huang, K. S. Wen, S. F. Pai, L. W. Wu, and R. H. Horng, “An 83% enhancement in the external quantum efficiency of ultraviolet flip-chip light-emitting diodes with the incorporation of a self-textured oxide mask,” IEEE Electron Device Lett. 34(2), 274–276 (2013). 2. C. T. Lee, U. Z. Yang, C. S. Lee, and P. S. Chen, “White light emission of monolithic Carbon-implanted InGaN–GaN light-emitting diodes,” IEEE Photon. Technol. Lett. 18(19), 2029–2031 (2006). 3. R. H. Horng, K. C. Shen, Y. W. Kuo, and D. S. Wuu, “GaN light emitting diodes with wing-type imbedded contacts,” Opt. Express 21(S1 Suppl 1), A1–A6 (2013). 4. K. J. Vampola, M. Iza, S. Keller, S. P. DenBaars, and S. Nakamura, “Measurement of electron overflow in 450 nm InGaN light-emitting diode structures,” Appl. Phys. Lett. 94(6), 061116 (2009). 5. C. H. Wang, C. C. Ke, C. Y. Lee, S. P. Chang, W. T. Chang, J. C. Li, Z. Y. Li, H. C. Yang, H. C. Kuo, T. C. Lu, and S. C. Wang, “Hole injection and efficiency droop improvement in InGaN/GaN light-emitting diodes by band-engineered electron blocking layer,” Appl. Phys. Lett. 97(26), 261103 (2010). 6. M. F. Schubert, S. Chhajed, J. K. Kim, E. F. Schubert, D. D. Koleske, M. H. Crawford, S. R. Lee, A. J. Fischer, G. Thaler, and M. A. Banas, “Effect of dislocation density on efficiency droop in GaInN/GaN light-emitting diodes,” Appl. Phys. Lett. 91(23), 231114 (2007). 7. K. C. Shen, D. S. Wuu, C. C. Shen, S. L. Ou, and R. H. Horng, “Surface modification on wet-etched patterned sapphire substrates using plasma treatments for improved GaN crystal quality and LED performance,” J. Electrochem. Soc. 158(10), H988–H993 (2011). 8. D. A. Steigerwald, J. C. Bhat, D. Collins, R. M. Fletcher, M. O. Holcomb, M. J. Ludowise, P. S. Martin, and S. Rudaz, “Illumination with solid state lighting technology,” IEEE J. Sel. Top. Quantum Electron. 8(2), 310–320 (2002). 9. M. L. Tsai, J. H. Liao, J. H. Yeh, T. C. Hsu, S. J. Hon, T. Y. Chung, and K. Y. Lai, “High-voltage thin-film GaN LEDs fabricated on ceramic substrates: the alleviated droop effect at 670 W/cm2.,” Opt. Express 21(22), 27102–27110 (2013). 10. C. H. Wang, D. W. Lin, C. Y. Lee, M. A. Tsai, G. L. Chen, H. T. Kuo, W. H. Hsu, H. C. Kuo, T. C. Lu, S. C. Wang, and G. C. Chi, “Efciency and droop improvement in GaN-based high-voltage light-emitting diodes,” IEEE Electron Device Lett. 32(8), 1098–1100 (2011). 11. R. H. Horng, K. C. Shen, Y. W. Kuo, and D. S. Wuu, “Effects of cell distance on the performance of GaN high-voltage light emitting diode,” Electrochem. Solid-State Lett. 1(5), R21–R23 (2012).
#217213 - $15.00 USD Received 17 Jul 2014; revised 22 Aug 2014; accepted 25 Aug 2014; published 5 Sep 2014 (C) 2014 OSA 20 October 2014 | Vol. 22, No. S6 | DOI:10.1364/OE.22.0A1462 | OPTICS EXPRESS A1462
12. R. H. Horng, X. Zheng, C. Y. Hsieh, and D. S. Wuu, “Light extraction enhancement of InGaN light-emitting diode by roughening both undoped micropillar-structure GaN and p-GaN as well as employing an omnidirectional reflector,” Appl. Phys. Lett. 93(2), 021125 (2008). 13. R. H. Horng, S. H. Huang, C. Y. Hsieh, X. Zheng, and D. S. Wuu, “Enhanced luminance efficiency of wafer-bonded InGaN-GaN LEDs with double-side textured surfaces and omnidirectional reflectors,” IEEE J. Quantum Electron. 44(11), 1116–1123 (2008).
1. Introduction Gallium nitride (GaN) LEDs are recently developed, attractive solutions for visible and near-ultraviolet light generation and can be applied in solid-state lighting, large full-color displays, and the back lighting of liquid crystal displays [1–3]. In GaN LEDs, the chip size is typically increased to enable higher-current operation and, thus, increase the output power of the LED. However, an “efficiency droop” easily occurs when employing high injection currents; this problem has been attributed partially to poor thermal dissipation of the sapphire substrate. The thermal effect results in a decrease in LED efficiency. Several factors including carrier leakage, carrier overflow, defects (or dislocations), and delocalization from InN fluctuations are considered responsible for the efficiency droop. However, the main underlying factors remain uncertain [4–7]. A series-connection cell LED configuration was proposed to ameliorate the efficiency droop. Compared with conventional LEDs, this series-connection cell LED features a relatively low driving current, high forward voltage operation, a simplified power-supply design, a more favorable current-spreading effect, and excellent reliability at an identical output power. However, the gap between each cell LED in high-voltage light-emitting diodes (HVLEDs) is usually too small, resulting in a substantial loss of light [8]. In a previous study [9], thin-film HVLEDs (TF-HVLEDs) were manufactured on ceramic substrates and exhibited excellent current spreading and thermal dissipation capabilities. However, performance comparisons between HVLEDs fabricated using a sapphire substrate and thin-film HVLEDs fabricated using the same epilayer and processing techniques have not yet been conducted. Currently, HVLEDs are manufactured by using microchips connected in series in one large chip [10]. The distance between the microchips is not optimal, and current crowding easily occurs in this type of electrode design. Therefore, we proposed a novel HVLED structure comprising a cell LED array (series-by-parallel connected configuration) and a replacement substrate to meet the requirements of high-voltage operation and high thermal dissipation. The optimal gap in HVLEDs of 80 μm identified in our previous study [11] was applied in this research. In addition, this study indicated the advantages of using TF-HVLEDs; namely, high thermal dissipation can be achieved through HVLEDs bonding to the Si substrate and high light extraction can be attained by incorporating a roughened n-GaN and indium tin oxide (ITO) surface into the structure. Details on the electrical and optical properties of HVLEDs fabricated using a sapphire substrate and HVLEDs fabricated using a mirror/silicon substrate are presented in the Results and Discussion section. 2. Experimental The fabrication process for a single cell LED is shown in Fig. 1. All LED structures were grown on a c-plane sapphire substrate consisting of a 30-nm-thick GaN nucleation layer, a 2-μm-thick undoped GaN layer, a 2-μm-thick n-GaN layer, undoped four-period GaN/InGaN multiple-quantum wells (MQWs), a 0.3-μm-thick p-GaN layer, and a 0.1-μm-thick p-GaN contact layer by using a metal-organic chemical vapor deposition system. A cell LED with a chip size of 10 × 23 mil2 and a roughened ITO layer was fabricated using photolithography and etching processes, as shown in Fig. 1(a). A Cr/Au (25/300 nm) metal layer serving as the n- and p-electrode pad was deposited on the surface of the n-GaN and ITO layer through thermal evaporation. Subsequently, the cell LED was mounted on a temporary glass substrate and the sapphire substrate was removed using a laser lift-off technique, as shown in Figs. 1(b)–1(e). In this step, the back surface of the sapphire substrate was irradiated using an excimer pulse laser (λ = 248 nm). The GaN layer was locally heated and dissociated close to the sapphire–GaN
#217213 - $15.00 USD Received 17 Jul 2014; revised 22 Aug 2014; accepted 25 Aug 2014; published 5 Sep 2014 (C) 2014 OSA 20 October 2014 | Vol. 22, No. S6 | DOI:10.1364/OE.22.0A1462 | OPTICS EXPRESS A1463
interface. After the entire LED wafer was illuminated with the laser beam, the sapphire substrate was separated from the LED structure. To enhance the light-extraction efficiency, a 4 M NaOH solution was used to roughen the n-GaN surface of the LED at 60 °C for 4 min. By using glue bonding, the top side (n side) of the inverted cell LED was then bonded to a thin-film-coated silicon substrate consisting of a 900-nm-thick glue layer (refractive index: 1.4) and a highly reflective Ag/Ti mirror layer (200/20 nm thickness) [12]. The layer of glue was thin and did not affect thermal dissipation, as demonstrated in our previous study [13]. Finally, the adhesive was etched away using an HCl solution to separate the glass substrate (see Fig. 1(f)). Two types of HVLED structure and their images, which were obtained using optical microscopy, are depicted in Figs. 2 and 3. A polymer was used to passivate and flatten the trenches in the cells (Fig. 2(a)). Subsequently, an 8 × 8 cell LED array with double-sided textured surfaces on a mirror/silicon substrate, designated as the TF-HVLED, was connected by employing a Ti/Al/Ti/Au (20/1500/20/200 nm) multilayer. For comparison, the roughened ITO layer of the HVLED grown on a sapphire substrate, designated as the C-HVLED, was fabricated (see Fig. 2(b)). Light loss resulting from light crosstalk, which occurs in emitting cell LEDs, can be reduced by optimizing the gap between the cells. In our previous study, we identified an optimal gap in HVLEDs of 80 μm. Applying this gap reduced the effect of light absorption by the surrounding cells [11]. The HVLED samples were encapsulated in epoxy before the optoelectronic measurements, and the photograph of the encapsulated HVLED was shown in Fig. 2(c). The current–voltage (I–V) characteristics of the two HVLEDs were measured at room temperature by using an Agilent 4155B semiconductor parameter analyzer. The light output power and electroluminescence (EL) spectra of the HVLEDs were measured using an integrating sphere detector (CAS 140B, Instrument Systems). Surface temperatures of TF‐HVLED and C‐HVLED were evaluated by using the thermal infrared image in thermal equilibrium at the injection currents of 24 and 80 mA, respectively.
Fig. 1. Fabrication flow of a single cell LED.
#217213 - $15.00 USD Received 17 Jul 2014; revised 22 Aug 2014; accepted 25 Aug 2014; published 5 Sep 2014 (C) 2014 OSA 20 October 2014 | Vol. 22, No. S6 | DOI:10.1364/OE.22.0A1462 | OPTICS EXPRESS A1464
Fig. 2. Schematic structure of the (a) TF-HVLEDs, (b) C-HVLEDs, and (c) encapsulated HVLEDs.
Fig. 3. Optical microscope images of the (a) C-HVLEDs and (b) TF-HVLEDs.
3. Results and discussion Typical I–V characteristics of the TF-HVLED and C-HVLED are shown in Fig. 4. The forward voltages of the single cell LED fabricated using a sapphire substrate and the single cell LED fabricated using a mirror/silicon substrate were almost identical, as shown in the inset of Fig. 4. The performance of the cell LED was apparently unaffected by the additional substrate replacement process, as shown in the inset of Fig. 4. Thus, the electrode design of the cell was optimal. The series and parallel connections obviated the problem of current crowding, which is a consistent problem of HVLEDs fabricated with microchips containging series electrodes [10]. The forward voltages of the TF-HVLED and C-HVLED measured at an injection current of 24 mA were 21.6 and 21.88 V, respectively.
#217213 - $15.00 USD Received 17 Jul 2014; revised 22 Aug 2014; accepted 25 Aug 2014; published 5 Sep 2014 (C) 2014 OSA 20 October 2014 | Vol. 22, No. S6 | DOI:10.1364/OE.22.0A1462 | OPTICS EXPRESS A1465
26
TF-HVLED C-HVLED 4
22
Voltage (V)
Voltage (V)
24
20
3
LED on mirror/Si LED on sapphire 2
18
30
40
0
50
10 20 30 40 50 60 70
Current (mA) 60
Current (mA)
70
80
Fig. 4. I–V curves of the C-HVLED and TF-HVLED. The inset shows I–V curves of the single cell LED on a sapphire and a mirror/silicon substrate.
Figure 5 shows the light output power and wall-plug efficiency (WPE) against the injection current of the HVLED grown on a sapphire substrate and that grown on a mirror/silicon substrate. The light output powers at 24 mA of the TF-HVLED and C-HVLED were 216 and 170 mW, respectively. The TF-HVLED exhibited a 27% enhancement in output power compared with the C-HVLED. This enhancement was attributed to increased light scattering by the glue/Ag/Ti/Si omnidirectional reflector structure, double-side roughening, and improved thermal dissipation by the Si substrate. The difference between the light output powers of the two HVLEDs increased with the injection current. The output power enhancement ratio increased from 27% to 30% when the injected current was increased from 24 to 80 mA. The structure of the rough double side in the TF-HVLED might be a factor in the improvement of the output power relative to that of the C-HVLED, which had only one rough side. The WPE values of the TF-HVLED and C-HVLED at an injection current of 24 mA were estimated to be 41.8% and 32.43%, respectively, representing a 28.89% enhancement in the WPE of the TF-HVLED compared with that of the C-HVLED. This enhancement occurred because of double-sided roughening, which generates photons with a greater probability of escaping from the active region, thus improving the light-extraction efficiency. Although the WPE values of the TF-HVLED and C-HVLED at 80 mA decreased to 33.8% and 25.79%, the droop levels were 19.1% and 20.5%, respectively. This difference arises from the improved effect of thermal dissipation attributable to bonding on a mirror/silicon substrate. Thus, the droop levels were improved noticeably compared with that of a conventional power chip driven at 350 mA because the driving current of the two types of LED was low. On the other hand, it is well known that the external quantum efficiency (EQE) is the internal quantum efficiency (IQE) multiplied by the light extraction efficiency (LEE). The relationship between the WPE, EQE, IQE and LEE can be expressed in the Eq. (1):
ηWPE =
η EQE ⋅ E ph η IQE ⋅η LEE ⋅ E ph PL = = × 100% Vf I f Vf Vf
(1)
where If denotes the injection current; Vf denotes the forward voltage; PL is the light output power; Eph is the emission light energy. To clarify the improvement of the light extraction efficiency in the TF-HVLED at an injection current of 24 mA, we defined both the WPE and LEE ratios between TF‐HVLED and C-HVLED, as shown in the Eq. (2).
#217213 - $15.00 USD Received 17 Jul 2014; revised 22 Aug 2014; accepted 25 Aug 2014; published 5 Sep 2014 (C) 2014 OSA 20 October 2014 | Vol. 22, No. S6 | DOI:10.1364/OE.22.0A1462 | OPTICS EXPRESS A1466
(ηWPE )TF − HVLED (ηWPE )C − HVLED
=
η IQE ⋅η LEE ⋅ E ph Vf TF − HVLED
(2)
η IQE ⋅η LEE ⋅ E ph Vf C − HVLED
Obviously, there was a 27% enhancement in the light extraction efficiency (@24 mA) of TF-HVLED compared with that of C-HVLED. This can be attributed to the introductions of the double-side textured surfaces and a mirror/silicon substrate.
Output power (mW)
TF-HVLED
42
600
39
500
36
400
33
300
30
200 100 20
27
C-HVLED 30
40
50
60
70
80
Wall-plug efficiency (%)
45 700
24
Current (mA) Fig. 5. Plots of light output power and WPE against the injection current for the C-HVLED and TF-HVLED.
The emission peak wavelengths of the C-HVLED and TF-HVLED at various driving currents are shown in Fig. 6(a). The EL emission peak shifted to a shorter wavelength (blue shift) as the injected current increased from 24 to 48 mA because of the band filling effect. Red shifting (a shift to a longer wavelength) occurred in the spectra of the C-HVLED when the current exceeded 48 mA because of the thermal effect of Joule heating at a greater current. However, the high thermal conductivity of Si in the TF-HVLED provided greater heat dissipation and enabled the peak position to remain at approximately 456.3 nm, even at an injection current of over 48 mA. The EL spectra of the two HVLEDs at an injection current of 80 mA are shown in Fig. 6(b). The peak wavelengths of the C-HVLED and TF-HVLED were approximately 458 and 456.3 nm, respectively. The TF-HVLED exhibited a blue shift in the emission peak relative to the C-HVLED at an injection current of 80 mA. This behavior may have been caused by high thermal dissipation by the Si substrate (148 W/mK), which has six times the thermal conductivity of sapphire (25 W/mK) and thus provides an effective heat sink. These properties enhanced the light output power of the HVLED at a high current injection because the Si substrate alleviates the Joule heating problem, stabilizing the peak-wavelength shift. To confirm the thermal dissipation performance, the HVLEDs were analyzed using thermal infrared imaging, as shown in Fig. 7. At 24 mA, the surface temperature distributions of the C-HVLED and TF-HVLED were 35.4–40.8 °C and 35.7–38.2 °C, with accompanying temperature differences (∆T) at 5.4 °C and 2.5 °C, respectively. These results indicated that the TF-HVLED exhibited a lower surface temperature and a more uniform temperature distribution. At a driving current of 80 mA, the C-HVLED exhibited surface temperatures in a higher range (42.6–51.3 °C) and a substantial ∆T (8.7 °C). By contrast, the surface temperature and ∆T of the TF-HVLED were greatly improved (40.7–41.2 °C and 0.5 °C, respectively);
#217213 - $15.00 USD Received 17 Jul 2014; revised 22 Aug 2014; accepted 25 Aug 2014; published 5 Sep 2014 (C) 2014 OSA 20 October 2014 | Vol. 22, No. S6 | DOI:10.1364/OE.22.0A1462 | OPTICS EXPRESS A1467
these improvements were attributed to improvement of the Si substrate. The TF-HVLED exhibited greater thermal dissipation and a more uniform temperature distribution at a higher driving current. The thin glue, as a low refractive index material, did not affect thermal dissipation in the TF-HVLEDs.
Fig. 6. (a) EL emission peak wavelength as a function of the injection current for the C-HVLED and TF-HVLED. (b) EL spectra of both HVLEDs measured at 80 mA.
Fig. 7. Surface temperature distributions of the C-HVLED and TF-HVLED at injection currents of (a and c) 24 mA and (b and d) 80 mA.
4. Conclusion In this study, a TF-HVLED was fabricated by performing substrate bonding, removal, and surface roughening. When operating at a high voltage and low injection current, the TF-HVLED exhibited advantages in optical and electrical properties. A 30% enhancement in the light output power of the TF-HVLED compared with that of the C-HVLED was observed and attributed to the introduction of a roughened double side and the mirror/silicon substrate. The results have implications regarding manufacturing technology and the design of HVLED configurations. Roughening the n-GaN and ITO layers substantially improved fabrication because surface treatment enhances the light-emission efficiency. Acknowledgments The authors would like to thank the Ministry of Economic Affairs under Grant No.102-E0605, and Ministry of Science and Technology, Taiwan, under Grant Nos. 100-2221-E005-092-MY3, 102-2221-E-005-071MY3, 103-2218-E005-004 and 102-2622-E-005-006.
#217213 - $15.00 USD Received 17 Jul 2014; revised 22 Aug 2014; accepted 25 Aug 2014; published 5 Sep 2014 (C) 2014 OSA 20 October 2014 | Vol. 22, No. S6 | DOI:10.1364/OE.22.0A1462 | OPTICS EXPRESS A1468
1 Department of Materials Science and Engineering, National Chung Hsing University, Taichung 40227, Taiwan Electronics and Opto-electronics Research Laboratories, Industrial Technology Research Institute, Hsinchu 31040, Taiwan 3 Graduate Institute of Precision Engineering, National Chung Hsing University, Taichung 40227, Taiwan 4 Advanced Optoelectronic Technology Center, National Cheng Kung University, Tainan 70101, Taiwan * [email protected]
Abstract: The characteristics of high-voltage light-emitting diodes (HVLEDs) consisting of a 64-cell LED array were investigated by employing various LED structures. Two types of HVLED were examined: a standard HVLED with a single roughened indium tin oxide (ITO) surface grown on a sapphire substrate and a thin-film HVLED (TF-HVLED) with a roughened n-GaN and ITO double side transferred to a mirror/silicon substrate. At an injection current of 24 mA, the output powers of the HVLEDs fabricated using a sapphire substrate and those fabricated using a mirror/silicon substrate were 170 and 216 mW, respectively. Because the TF-HVLED exhibited improved thermal dissipation and light extraction, it produced a greater output power than the HVLED fabricated using the sapphire substrate did. ©2014 Optical Society of America OCIS codes: (230.2090) Electro-optical devices; (230.3670) Light-emitting diodes
References and links 1.
K. C. Shen, W. Y. Lin, D. S. Wuu, S. Y. Huang, K. S. Wen, S. F. Pai, L. W. Wu, and R. H. Horng, “An 83% enhancement in the external quantum efficiency of ultraviolet flip-chip light-emitting diodes with the incorporation of a self-textured oxide mask,” IEEE Electron Device Lett. 34(2), 274–276 (2013). 2. C. T. Lee, U. Z. Yang, C. S. Lee, and P. S. Chen, “White light emission of monolithic Carbon-implanted InGaN–GaN light-emitting diodes,” IEEE Photon. Technol. Lett. 18(19), 2029–2031 (2006). 3. R. H. Horng, K. C. Shen, Y. W. Kuo, and D. S. Wuu, “GaN light emitting diodes with wing-type imbedded contacts,” Opt. Express 21(S1 Suppl 1), A1–A6 (2013). 4. K. J. Vampola, M. Iza, S. Keller, S. P. DenBaars, and S. Nakamura, “Measurement of electron overflow in 450 nm InGaN light-emitting diode structures,” Appl. Phys. Lett. 94(6), 061116 (2009). 5. C. H. Wang, C. C. Ke, C. Y. Lee, S. P. Chang, W. T. Chang, J. C. Li, Z. Y. Li, H. C. Yang, H. C. Kuo, T. C. Lu, and S. C. Wang, “Hole injection and efficiency droop improvement in InGaN/GaN light-emitting diodes by band-engineered electron blocking layer,” Appl. Phys. Lett. 97(26), 261103 (2010). 6. M. F. Schubert, S. Chhajed, J. K. Kim, E. F. Schubert, D. D. Koleske, M. H. Crawford, S. R. Lee, A. J. Fischer, G. Thaler, and M. A. Banas, “Effect of dislocation density on efficiency droop in GaInN/GaN light-emitting diodes,” Appl. Phys. Lett. 91(23), 231114 (2007). 7. K. C. Shen, D. S. Wuu, C. C. Shen, S. L. Ou, and R. H. Horng, “Surface modification on wet-etched patterned sapphire substrates using plasma treatments for improved GaN crystal quality and LED performance,” J. Electrochem. Soc. 158(10), H988–H993 (2011). 8. D. A. Steigerwald, J. C. Bhat, D. Collins, R. M. Fletcher, M. O. Holcomb, M. J. Ludowise, P. S. Martin, and S. Rudaz, “Illumination with solid state lighting technology,” IEEE J. Sel. Top. Quantum Electron. 8(2), 310–320 (2002). 9. M. L. Tsai, J. H. Liao, J. H. Yeh, T. C. Hsu, S. J. Hon, T. Y. Chung, and K. Y. Lai, “High-voltage thin-film GaN LEDs fabricated on ceramic substrates: the alleviated droop effect at 670 W/cm2.,” Opt. Express 21(22), 27102–27110 (2013). 10. C. H. Wang, D. W. Lin, C. Y. Lee, M. A. Tsai, G. L. Chen, H. T. Kuo, W. H. Hsu, H. C. Kuo, T. C. Lu, S. C. Wang, and G. C. Chi, “Efciency and droop improvement in GaN-based high-voltage light-emitting diodes,” IEEE Electron Device Lett. 32(8), 1098–1100 (2011). 11. R. H. Horng, K. C. Shen, Y. W. Kuo, and D. S. Wuu, “Effects of cell distance on the performance of GaN high-voltage light emitting diode,” Electrochem. Solid-State Lett. 1(5), R21–R23 (2012).
#217213 - $15.00 USD Received 17 Jul 2014; revised 22 Aug 2014; accepted 25 Aug 2014; published 5 Sep 2014 (C) 2014 OSA 20 October 2014 | Vol. 22, No. S6 | DOI:10.1364/OE.22.0A1462 | OPTICS EXPRESS A1462
12. R. H. Horng, X. Zheng, C. Y. Hsieh, and D. S. Wuu, “Light extraction enhancement of InGaN light-emitting diode by roughening both undoped micropillar-structure GaN and p-GaN as well as employing an omnidirectional reflector,” Appl. Phys. Lett. 93(2), 021125 (2008). 13. R. H. Horng, S. H. Huang, C. Y. Hsieh, X. Zheng, and D. S. Wuu, “Enhanced luminance efficiency of wafer-bonded InGaN-GaN LEDs with double-side textured surfaces and omnidirectional reflectors,” IEEE J. Quantum Electron. 44(11), 1116–1123 (2008).
1. Introduction Gallium nitride (GaN) LEDs are recently developed, attractive solutions for visible and near-ultraviolet light generation and can be applied in solid-state lighting, large full-color displays, and the back lighting of liquid crystal displays [1–3]. In GaN LEDs, the chip size is typically increased to enable higher-current operation and, thus, increase the output power of the LED. However, an “efficiency droop” easily occurs when employing high injection currents; this problem has been attributed partially to poor thermal dissipation of the sapphire substrate. The thermal effect results in a decrease in LED efficiency. Several factors including carrier leakage, carrier overflow, defects (or dislocations), and delocalization from InN fluctuations are considered responsible for the efficiency droop. However, the main underlying factors remain uncertain [4–7]. A series-connection cell LED configuration was proposed to ameliorate the efficiency droop. Compared with conventional LEDs, this series-connection cell LED features a relatively low driving current, high forward voltage operation, a simplified power-supply design, a more favorable current-spreading effect, and excellent reliability at an identical output power. However, the gap between each cell LED in high-voltage light-emitting diodes (HVLEDs) is usually too small, resulting in a substantial loss of light [8]. In a previous study [9], thin-film HVLEDs (TF-HVLEDs) were manufactured on ceramic substrates and exhibited excellent current spreading and thermal dissipation capabilities. However, performance comparisons between HVLEDs fabricated using a sapphire substrate and thin-film HVLEDs fabricated using the same epilayer and processing techniques have not yet been conducted. Currently, HVLEDs are manufactured by using microchips connected in series in one large chip [10]. The distance between the microchips is not optimal, and current crowding easily occurs in this type of electrode design. Therefore, we proposed a novel HVLED structure comprising a cell LED array (series-by-parallel connected configuration) and a replacement substrate to meet the requirements of high-voltage operation and high thermal dissipation. The optimal gap in HVLEDs of 80 μm identified in our previous study [11] was applied in this research. In addition, this study indicated the advantages of using TF-HVLEDs; namely, high thermal dissipation can be achieved through HVLEDs bonding to the Si substrate and high light extraction can be attained by incorporating a roughened n-GaN and indium tin oxide (ITO) surface into the structure. Details on the electrical and optical properties of HVLEDs fabricated using a sapphire substrate and HVLEDs fabricated using a mirror/silicon substrate are presented in the Results and Discussion section. 2. Experimental The fabrication process for a single cell LED is shown in Fig. 1. All LED structures were grown on a c-plane sapphire substrate consisting of a 30-nm-thick GaN nucleation layer, a 2-μm-thick undoped GaN layer, a 2-μm-thick n-GaN layer, undoped four-period GaN/InGaN multiple-quantum wells (MQWs), a 0.3-μm-thick p-GaN layer, and a 0.1-μm-thick p-GaN contact layer by using a metal-organic chemical vapor deposition system. A cell LED with a chip size of 10 × 23 mil2 and a roughened ITO layer was fabricated using photolithography and etching processes, as shown in Fig. 1(a). A Cr/Au (25/300 nm) metal layer serving as the n- and p-electrode pad was deposited on the surface of the n-GaN and ITO layer through thermal evaporation. Subsequently, the cell LED was mounted on a temporary glass substrate and the sapphire substrate was removed using a laser lift-off technique, as shown in Figs. 1(b)–1(e). In this step, the back surface of the sapphire substrate was irradiated using an excimer pulse laser (λ = 248 nm). The GaN layer was locally heated and dissociated close to the sapphire–GaN
#217213 - $15.00 USD Received 17 Jul 2014; revised 22 Aug 2014; accepted 25 Aug 2014; published 5 Sep 2014 (C) 2014 OSA 20 October 2014 | Vol. 22, No. S6 | DOI:10.1364/OE.22.0A1462 | OPTICS EXPRESS A1463
interface. After the entire LED wafer was illuminated with the laser beam, the sapphire substrate was separated from the LED structure. To enhance the light-extraction efficiency, a 4 M NaOH solution was used to roughen the n-GaN surface of the LED at 60 °C for 4 min. By using glue bonding, the top side (n side) of the inverted cell LED was then bonded to a thin-film-coated silicon substrate consisting of a 900-nm-thick glue layer (refractive index: 1.4) and a highly reflective Ag/Ti mirror layer (200/20 nm thickness) [12]. The layer of glue was thin and did not affect thermal dissipation, as demonstrated in our previous study [13]. Finally, the adhesive was etched away using an HCl solution to separate the glass substrate (see Fig. 1(f)). Two types of HVLED structure and their images, which were obtained using optical microscopy, are depicted in Figs. 2 and 3. A polymer was used to passivate and flatten the trenches in the cells (Fig. 2(a)). Subsequently, an 8 × 8 cell LED array with double-sided textured surfaces on a mirror/silicon substrate, designated as the TF-HVLED, was connected by employing a Ti/Al/Ti/Au (20/1500/20/200 nm) multilayer. For comparison, the roughened ITO layer of the HVLED grown on a sapphire substrate, designated as the C-HVLED, was fabricated (see Fig. 2(b)). Light loss resulting from light crosstalk, which occurs in emitting cell LEDs, can be reduced by optimizing the gap between the cells. In our previous study, we identified an optimal gap in HVLEDs of 80 μm. Applying this gap reduced the effect of light absorption by the surrounding cells [11]. The HVLED samples were encapsulated in epoxy before the optoelectronic measurements, and the photograph of the encapsulated HVLED was shown in Fig. 2(c). The current–voltage (I–V) characteristics of the two HVLEDs were measured at room temperature by using an Agilent 4155B semiconductor parameter analyzer. The light output power and electroluminescence (EL) spectra of the HVLEDs were measured using an integrating sphere detector (CAS 140B, Instrument Systems). Surface temperatures of TF‐HVLED and C‐HVLED were evaluated by using the thermal infrared image in thermal equilibrium at the injection currents of 24 and 80 mA, respectively.
Fig. 1. Fabrication flow of a single cell LED.
#217213 - $15.00 USD Received 17 Jul 2014; revised 22 Aug 2014; accepted 25 Aug 2014; published 5 Sep 2014 (C) 2014 OSA 20 October 2014 | Vol. 22, No. S6 | DOI:10.1364/OE.22.0A1462 | OPTICS EXPRESS A1464
Fig. 2. Schematic structure of the (a) TF-HVLEDs, (b) C-HVLEDs, and (c) encapsulated HVLEDs.
Fig. 3. Optical microscope images of the (a) C-HVLEDs and (b) TF-HVLEDs.
3. Results and discussion Typical I–V characteristics of the TF-HVLED and C-HVLED are shown in Fig. 4. The forward voltages of the single cell LED fabricated using a sapphire substrate and the single cell LED fabricated using a mirror/silicon substrate were almost identical, as shown in the inset of Fig. 4. The performance of the cell LED was apparently unaffected by the additional substrate replacement process, as shown in the inset of Fig. 4. Thus, the electrode design of the cell was optimal. The series and parallel connections obviated the problem of current crowding, which is a consistent problem of HVLEDs fabricated with microchips containging series electrodes [10]. The forward voltages of the TF-HVLED and C-HVLED measured at an injection current of 24 mA were 21.6 and 21.88 V, respectively.
#217213 - $15.00 USD Received 17 Jul 2014; revised 22 Aug 2014; accepted 25 Aug 2014; published 5 Sep 2014 (C) 2014 OSA 20 October 2014 | Vol. 22, No. S6 | DOI:10.1364/OE.22.0A1462 | OPTICS EXPRESS A1465
26
TF-HVLED C-HVLED 4
22
Voltage (V)
Voltage (V)
24
20
3
LED on mirror/Si LED on sapphire 2
18
30
40
0
50
10 20 30 40 50 60 70
Current (mA) 60
Current (mA)
70
80
Fig. 4. I–V curves of the C-HVLED and TF-HVLED. The inset shows I–V curves of the single cell LED on a sapphire and a mirror/silicon substrate.
Figure 5 shows the light output power and wall-plug efficiency (WPE) against the injection current of the HVLED grown on a sapphire substrate and that grown on a mirror/silicon substrate. The light output powers at 24 mA of the TF-HVLED and C-HVLED were 216 and 170 mW, respectively. The TF-HVLED exhibited a 27% enhancement in output power compared with the C-HVLED. This enhancement was attributed to increased light scattering by the glue/Ag/Ti/Si omnidirectional reflector structure, double-side roughening, and improved thermal dissipation by the Si substrate. The difference between the light output powers of the two HVLEDs increased with the injection current. The output power enhancement ratio increased from 27% to 30% when the injected current was increased from 24 to 80 mA. The structure of the rough double side in the TF-HVLED might be a factor in the improvement of the output power relative to that of the C-HVLED, which had only one rough side. The WPE values of the TF-HVLED and C-HVLED at an injection current of 24 mA were estimated to be 41.8% and 32.43%, respectively, representing a 28.89% enhancement in the WPE of the TF-HVLED compared with that of the C-HVLED. This enhancement occurred because of double-sided roughening, which generates photons with a greater probability of escaping from the active region, thus improving the light-extraction efficiency. Although the WPE values of the TF-HVLED and C-HVLED at 80 mA decreased to 33.8% and 25.79%, the droop levels were 19.1% and 20.5%, respectively. This difference arises from the improved effect of thermal dissipation attributable to bonding on a mirror/silicon substrate. Thus, the droop levels were improved noticeably compared with that of a conventional power chip driven at 350 mA because the driving current of the two types of LED was low. On the other hand, it is well known that the external quantum efficiency (EQE) is the internal quantum efficiency (IQE) multiplied by the light extraction efficiency (LEE). The relationship between the WPE, EQE, IQE and LEE can be expressed in the Eq. (1):
ηWPE =
η EQE ⋅ E ph η IQE ⋅η LEE ⋅ E ph PL = = × 100% Vf I f Vf Vf
(1)
where If denotes the injection current; Vf denotes the forward voltage; PL is the light output power; Eph is the emission light energy. To clarify the improvement of the light extraction efficiency in the TF-HVLED at an injection current of 24 mA, we defined both the WPE and LEE ratios between TF‐HVLED and C-HVLED, as shown in the Eq. (2).
#217213 - $15.00 USD Received 17 Jul 2014; revised 22 Aug 2014; accepted 25 Aug 2014; published 5 Sep 2014 (C) 2014 OSA 20 October 2014 | Vol. 22, No. S6 | DOI:10.1364/OE.22.0A1462 | OPTICS EXPRESS A1466
(ηWPE )TF − HVLED (ηWPE )C − HVLED
=
η IQE ⋅η LEE ⋅ E ph Vf TF − HVLED
(2)
η IQE ⋅η LEE ⋅ E ph Vf C − HVLED
Obviously, there was a 27% enhancement in the light extraction efficiency (@24 mA) of TF-HVLED compared with that of C-HVLED. This can be attributed to the introductions of the double-side textured surfaces and a mirror/silicon substrate.
Output power (mW)
TF-HVLED
42
600
39
500
36
400
33
300
30
200 100 20
27
C-HVLED 30
40
50
60
70
80
Wall-plug efficiency (%)
45 700
24
Current (mA) Fig. 5. Plots of light output power and WPE against the injection current for the C-HVLED and TF-HVLED.
The emission peak wavelengths of the C-HVLED and TF-HVLED at various driving currents are shown in Fig. 6(a). The EL emission peak shifted to a shorter wavelength (blue shift) as the injected current increased from 24 to 48 mA because of the band filling effect. Red shifting (a shift to a longer wavelength) occurred in the spectra of the C-HVLED when the current exceeded 48 mA because of the thermal effect of Joule heating at a greater current. However, the high thermal conductivity of Si in the TF-HVLED provided greater heat dissipation and enabled the peak position to remain at approximately 456.3 nm, even at an injection current of over 48 mA. The EL spectra of the two HVLEDs at an injection current of 80 mA are shown in Fig. 6(b). The peak wavelengths of the C-HVLED and TF-HVLED were approximately 458 and 456.3 nm, respectively. The TF-HVLED exhibited a blue shift in the emission peak relative to the C-HVLED at an injection current of 80 mA. This behavior may have been caused by high thermal dissipation by the Si substrate (148 W/mK), which has six times the thermal conductivity of sapphire (25 W/mK) and thus provides an effective heat sink. These properties enhanced the light output power of the HVLED at a high current injection because the Si substrate alleviates the Joule heating problem, stabilizing the peak-wavelength shift. To confirm the thermal dissipation performance, the HVLEDs were analyzed using thermal infrared imaging, as shown in Fig. 7. At 24 mA, the surface temperature distributions of the C-HVLED and TF-HVLED were 35.4–40.8 °C and 35.7–38.2 °C, with accompanying temperature differences (∆T) at 5.4 °C and 2.5 °C, respectively. These results indicated that the TF-HVLED exhibited a lower surface temperature and a more uniform temperature distribution. At a driving current of 80 mA, the C-HVLED exhibited surface temperatures in a higher range (42.6–51.3 °C) and a substantial ∆T (8.7 °C). By contrast, the surface temperature and ∆T of the TF-HVLED were greatly improved (40.7–41.2 °C and 0.5 °C, respectively);
#217213 - $15.00 USD Received 17 Jul 2014; revised 22 Aug 2014; accepted 25 Aug 2014; published 5 Sep 2014 (C) 2014 OSA 20 October 2014 | Vol. 22, No. S6 | DOI:10.1364/OE.22.0A1462 | OPTICS EXPRESS A1467
these improvements were attributed to improvement of the Si substrate. The TF-HVLED exhibited greater thermal dissipation and a more uniform temperature distribution at a higher driving current. The thin glue, as a low refractive index material, did not affect thermal dissipation in the TF-HVLEDs.
Fig. 6. (a) EL emission peak wavelength as a function of the injection current for the C-HVLED and TF-HVLED. (b) EL spectra of both HVLEDs measured at 80 mA.
Fig. 7. Surface temperature distributions of the C-HVLED and TF-HVLED at injection currents of (a and c) 24 mA and (b and d) 80 mA.
4. Conclusion In this study, a TF-HVLED was fabricated by performing substrate bonding, removal, and surface roughening. When operating at a high voltage and low injection current, the TF-HVLED exhibited advantages in optical and electrical properties. A 30% enhancement in the light output power of the TF-HVLED compared with that of the C-HVLED was observed and attributed to the introduction of a roughened double side and the mirror/silicon substrate. The results have implications regarding manufacturing technology and the design of HVLED configurations. Roughening the n-GaN and ITO layers substantially improved fabrication because surface treatment enhances the light-emission efficiency. Acknowledgments The authors would like to thank the Ministry of Economic Affairs under Grant No.102-E0605, and Ministry of Science and Technology, Taiwan, under Grant Nos. 100-2221-E005-092-MY3, 102-2221-E-005-071MY3, 103-2218-E005-004 and 102-2622-E-005-006.
#217213 - $15.00 USD Received 17 Jul 2014; revised 22 Aug 2014; accepted 25 Aug 2014; published 5 Sep 2014 (C) 2014 OSA 20 October 2014 | Vol. 22, No. S6 | DOI:10.1364/OE.22.0A1462 | OPTICS EXPRESS A1468
