Effects of current crowding on light extraction efficiency of conventional GaN-based lightemitting

Effects of current crowding on light extraction efficiency of conventional GaN-based lightemitting diodes Bin Cao, 1 Shuiming Li, 1 Run Hu, 2 Shengjun Zhou, 3 Yi Sun, 1 Zhiying Gan, 4 and Sheng Liu 4* 1 Wuhan National Laboratory for Optoelectronics, School of Optical and Electronic Information, Huazhong University of Science and Technology, Wuhan 430074, China 2 School of Energy & Power Engineering, Huazhong University of Science & Technology, Wuhan 430074, China 3 School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai, 200240, China 4 Institute of Microsystems, School of Mechanical Science & Engineering, Huazhong University of Science & Technology, Wuhan 430074, China * victor_liu63@126.com Abstract: Current crowding effects (CCEs) on light extraction efficiency (LEE) of conventional GaN-based light-emitting diodes (LEDs) are analyzed through Monte Carlo ray-tracing simulation. The non-uniform radiative power distribution of the active layer of the Monte Carlo model is obtained based on the current spreading theory and rate equation. The simulation results illustrate that CCE around n-pad (n-cce) has little effect on LEE, while CCE around p-pad (p-cce) results in a notable LEE droop due to the significant absorption of photons emitted under p-pad. LEE droop is alleviated by a SiO 2 current blocking layer (CBL) and reflective p- pad. Compared to the conventional LEDs without CBL, the simulated LEE of LEDs with CBL at 20 A/cm 2 and 70 A/cm 2 is enhanced by 7.7% and 19.0%, respectively. It is further enhanced by 7.6% and 11.4% after employing a reflective p-pad due to decreased absorption. These enhancements are in accordance with the experimental results. Output power of LEDs with CBL is enhanced by 8.7% and 18.2% at 20 A/cm 2 and 70 A/cm 2, respectively. And the reflective p-pad results in a further enhancement of 8.9% and 12.7%. 2013 Optical Society of America OCIS codes: (230.0230) Optical devices; (230.3670) Light-emitting diodes. References and links 1. E. F. Schubert, Light-Emitting Diodes (Cambridge University, 2006). 2. S. Liu and X. Luo, LED Packaging for Lighting Applications: Design, Manufacturing, and Testing (John Wiley & Sons, 2011). 3. A. I. Zhmakin, Enhancement of light extraction from light emitting diodes, Phys. Rep. 498(4-5), 189 241 (2011). 4. T. X. Lee, C. Y. Lin, S. H. Ma, and C. C. Sun, Analysis of position-dependent light extraction of GaN-based LEDs, Opt. Express 13(11), 4175 4179 (2005). 5. T. X. Lee, K. F. Gao, W. T. Chien, and C. C. Sun, Light extraction analysis of GaN-based light-emitting diodes with surface texture and/or patterned substrate, Opt. Express 15(11), 6670 6676 (2007). 6. C. C. Sun, C. Y. Chen, H. Y. He, C. C. Chen, W. T. Chien, T. X. Lee, and T. H. Yang, Precise optical modeling for silicate-based white LEDs, Opt. Express 16(24), 20060 20066 (2008). 7. Z. Liu, K. Wang, X. Luo, and S. Liu, Precise optical modeling of blue light-emitting diodes by Monte Carlo ray-tracing, Opt. Express 18(9), 9398 9412 (2010). 8. C. Sommer, F. P. Wenzl, P. Hartmann, P. Pachler, M. Schweighart, S. Tasch, and G. Leising, Tailoring of the color conversion elements in phosphor-converted high-power LEDs by optical simulations, IEEE Photon. Technol. Lett. 20(9), 739 741 (2008). 9. M. V. Bogdanov, K. A. Bulashevich, O. V. Khokhlev, I. Y. Evstratov, M. S. Ramm, and S. Y. Karpov, Current crowding effect on light extraction efficiency of thin-film LEDs, Phys. Status Solidi 7(7 8), 2124 2126 (2010) (c). 10. B. Cao, S. Zhou, and S. Liu, Effects of ITO pattern on the electrical and optical characteristics of LEDs, ESC J. Solid State Sci. Technol. 2(1), R24 R28 (2013). (C) 2013 OSA 21 October 2013 Vol. 21, No. 21 DOI:10.1364/OE.21.025381 OPTICS EXPRESS 25381

11. X. Guo and E. F. Schubert, Current crowding and optical saturation effects in GaInN/GaN light-emitting diodes grown on insulating substrates, Appl. Phys. Lett. 78(21), 3337 3339 (2001). 12. C. Huh, J. M. Lee, D. J. Kim, and S. J. Park, Improvement in light-output efficiency of InGaN/GaN multiplequantum well light-emitting diodes by current blocking layer, J. Appl. Phys. 92(5), 2248 2250 (2002). 13. P. Wang, W. Wei, B. Cao, Z. Gan, and S. Liu, Simulation of current spreading for GaN-based light-emitting diodes, Opt. Laser Technol. 42(5), 737 740 (2010). 14. S. J. Chang, C. F. Shen, M. H. Hsieh, C. T. Kuo, T. K. Ko, W. S. Chen, and S. C. Shei, Nitride-based LEDs with a hybrid Al mirror+tio 2 /SiO 2 DBR backside reflector, J. Lightwave Technol. 26(17), 3131 3136 (2008). 15. S. Zhou, F. Fang, B. Cao, S. Liu, and H. Ding, Enhancement in light output power of LEDs with reflective current blocking layer and backside hybrid reflector, Sci. China Technol. Sc. 56(6), 1544 1549 (2013). 16. H. Y. Ryu and J. I. Shim, Effect of current spreading on the efficiency droop of InGaN light-emitting diodes, Opt. Express 19(4), 2886 2894 (2011). 17. W. B. Joyce and S. H. Wemple, Steady-state junction-current distributions in thin resistive films on semiconductor junctions (solutions of =± e, J. Appl. Phys. 41(9), 3818 (1970). 2 v v 18. H. Kim and S. N. Lee, Theoretical considerations on current spreading in GaN-based light emitting diodes fabricated with top-emission geometry, J. Electrochem. Soc. 157(5), H562 H564 (2010). 19. J. M. Shah, Y. L. Li, T. Gessmann, and E. F. Schubert, Experimental analysis and theoretical model for anomalously high ideality factors (n>2.0) in AlGaN/GaN p-n junction diodes, J. Appl. Phys. 94(4), 2627 2631 (2003). 20. H. Y. Ryu, H. S. Kim, and J. I. Shim, Rate equation analysis of efficiency droop in InGaN light-emitting diodes, Appl. Phys. Lett. 95(8), 081114 (2009). 21. P. Kivisaari, L. Riuttanen, J. Oksanen, S. Suihkonen, M. Ali, H. Lipsanen, and J. Tulkki, Electrical measurement of internal quantum efficiency and extraction efficiency of III-N light-emitting diodes, Appl. Phys. Lett. 101(2), 021113 (2012). 22. C. C. Sun, T. X. Lee, Y. C. Lo, C. C. Chen, and S. Y. Tsai, Light extraction enhancement of GaN-based LEDs through passive/active photon recycling, Opt. Commun. 284(20), 4862 4868 (2011). 23. S. J. Chang, C. F. Shen, W. S. Chen, T. K. Ko, C. T. Kuo, K. H. Yu, S. C. Shei, and Y. Z. Chiou, Nitride-based LEDs with an insulating SiO 2 layer underneath p-pad electrodes, Electrochem. Solid-State Lett. 10(6), H175 H177 (2007). 24. J. C. Chen, G. J. Shen, F. S. Hwu, H. I. Chen, J. K. Shen, T. X. Lee, and C. C. Sun, Electrical-optical analysis of a GaN/sapphire LED chip by considering the resistivity of the current-spreading layer, Opt. Rev. 16(2), 213 215 (2009). 25. N. M. Lin, S. C. Shei, S. J. Chang, and X. F. Zeng, GaN-based LEDs with omnidirectional metal underneath an insulating SiO 2 Layer, IEEE Photon. Technol. Lett. 24(10), 815 817 (2012). 1. Introduction Regarded as the next generation lighting sources, gallium nitride (GaN)-based light-emitting diodes (LEDs) have been widely used in many illumination areas including traffic signals, liquid crystal display (LCD) backlighting sources and solid-state lighting sources due to high efficiency, long lifetime, high reliability, and environmental protection [1, 2]. Light extraction efficiency (LEE) of LEDs is an important parameter to characterize the ratio of total photons escaping into free space to the total photons emitted from active layer [1, 2]. LEE of conventional GaN-based LEDs is low due to the small light escape cone caused by the large reflective index difference between GaN and air [3, 4]. In order to analyze the LEE of LEDs, Monte Carlo ray-tracing method was widely used to simulate the light propagation processes [4 8]. However, the non-uniform distribution of photon caused by current crowding effect (CCE) was ignored and replaced by the uniform distribution in these models [4 8]. As a result, the effect of current crowding on LEE was rarely considered and reported. Bogdanov et al. studied this effect of thin-film LEDs based on a three-dimensional modeling coupling electrical and optical properties [9]. They just considered the thin-film LEDs and CCE in conventional GaN-based LEDs has not been studied so far. Actually, the phenomenon of current crowding in conventional GaN-based LEDs is different from the one in thin-film LEDs because current spreads laterally in the epitaxial layers of conventional LEDs due to the insulating sapphire, as shown in Fig. 1(a). Figure 1(b) shows current density distribution in the active layer at injected current density of 35 A/cm 2, which is represented in our previous work [10]. The current density decrease from about 99 A/cm 2 around n-pad to about 11 A/cm 2 around p-pad, illustrating current crowds around the n-pad significantly. Current crowding around n-pad or p-pad (n-cce or p-cce) is dependent on the device parameters [11, 12]. In this paper, we try to analyze in detail the effects of different current crowding phenomena on LEE of conventional GaN-based LEDs by the (C) 2013 OSA 21 October 2013 Vol. 21, No. 21 DOI:10.1364/OE.21.025381 OPTICS EXPRESS 25382

Monte Carlo ray-tracing simulation. The non-uniform distribution of photons emitted from the LED model caused by CCE is calculated based on the rate equation and current spreading theory. And then the LEEs of LEDs with p-cce and n-cce are analyzed, respectively. The simulated results are compared to the experimental results through the LEDs without a SiO 2 current blocking layer (CBL), with a CBL, and with a CBL and a reflective p-pad (RCBL). Fig. 1. (a) Schematic diagram of the conventional LED chip (b) contour of current density distribution in the active layer of conventional GaN-based LEDs with n-cce [10]. 2 Experimental details The epitaxial layers of the conventional GaN-based LED chips are grown on c-plane patterned sapphire substrate by a metal-organic chemical vapor deposition (MOCVD) system [13], as shown in Fig. 1(a). They consists of a 2.5-μm-thick undoped GaN layer, a 3-μm-thick Si-doped n-gan layer, a six-period In 0.2 Ga 0.8 N/GaN multiple quantum well (MQW) active layer, a 20-nm-thick Mg-doped p-al 0.1 Ga 0.9 N electron blocking layer, and a 100-nm-thick Mg-doped p-gan layer. After the epitaxial growth processes, the wafer is cleaned and an inductively coupled plasma (ICP) etching process is implemented to etch the mesh structure of the LEDs to expose the n-gan. A 130-nm-thick indium-tin oxide (ITO) layer is deposited on p-gan as the current spreading layer by thermal evaporator and then the epitaxial wafer is rapidly thermal annealed at 540 C for 5 mins in nitrogen ambient to improve ITO ohmic contact to GaN layer. A Cr/Pt/Au (50/50/1500 nm) layer is then deposited on ITO and n-gan as n-pad and p-pad, respectively. The epitaxial wafer is then thinned down to about 100 μm. Then, a hybrid reflector combining distributed Bragg reflector (DBR) and a 500-nm-thick Al metal mirror is evaporated on the backside of sapphire to improve the reflectivity [14]. DBR consists of 2-pair TiO 2 /SiO 2 with optical thickness of quarter of 460 nm for each layer. The reflectivity of hybrid reflector is nearly 100%. Finally, the LED wafer is diced into chips with size of 350 350 μm 2. As comparison, LEDs with the CBL and RCBL are fabricated, respectively. All three kinds of LEDs are equipped with the hybrid reflector structure. For LEDs with CBL, a film of SiO 2 layer with thickness of 200 nm is deposited on the p-gan surface by plasma enhanced chemical vapor deposition (PECVD) before the ITO deposition and then is patterned by photolithography and wet etching to from the CBL. A film of Al layer is inserted underneath the p-pad of LEDs with CBL to form the LEDs with RCBL structure. The reflectivity of Cr/Pt/Au and Al/Cr/Pt/Au are about 30% and 85% at wavelength of 460 nm, respectively, through the measurement of ellipsometry [15]. The output powercurrent-voltage (L-I-V) characteristics of the LED chips without silicone are measured in an integrating sphere under pulse current operation at room temperature to eliminate the thermal effect. 3. Non-uniform Radiation Modeling In order to simulate the LEE of LEDs with CCE by Monte Carlo ray-tracing method, the model with the active layer emitting photons non-uniformly should be built. Thus, we construct the relationship between the radiant power distribution of the active layer and CCE (C) 2013 OSA 21 October 2013 Vol. 21, No. 21 DOI:10.1364/OE.21.025381 OPTICS EXPRESS 25383

based on the rate equation and current spreading theory, and then the ray-tracing simulation is carried out. The procedure is illustrated in the following: 1) Defining the current density distribution J(x) of the LED chip under injected current density J inj based on the current spreading theory; 2) Defining the radiant power distribution R(x) based on the rate equation; 3) Defining the top and bottom surface of the active layer as the light source surfaces, and inputting R(x) into the meshed light source surfaces; 4) Defining the material properties of each layer; 5) Ray tracing simulation and collecting the simulation data. Current spreads laterally in conventional GaN-based LEDs grown on an insulating sapphire substrate. The current density J(x) extends away from the contact based on the current spreading theory, which can be expressed as [11, 16] J x J0 x L s ( ) = exp( / ) (1) where J 0 is the maximum current density at the electrode edge, x is the distance from a certain location to electrode edge and L s denotes the current spreading length, which is related to the sheet resistance of epitaxial layers and J 0. For the LEDs with n-cce, current density reaches the maximum at the mesh edge. For the LEDs with p-cce, current density under p-pad is the same and reaches the maximum. If the CBL is inserted under p-pad, the maximum current density is presented at the electrode edge and current density decays along the edge to the center of p-pad as Eq. (1). L s, always used to evaluate the CCE, can be expressed as [17, 18] L s = 2 nidealkt / q J ( ρ + ρ ) 0 ITO n (2) where ρ ITO and ρ n are sheet resistances of ITO and n-gan, respectively, k, q and T are the Boltzmann constant, element charge and temperature, n ideal is the ideality factor of the LEDs. The n ideal of GaN-based LEDs is always between 2 and 7 [19] and can be derived from the diode equation [1, 10]. In this paper, the n ideal is assumed to be 5. The thermal effect is eliminated when the GaN-based LEDs are operated under pulse current at room temperature and thus T is the room temperature. An integral of J(x) with respect to the meshed light source surface S(x) gives injected current I, which is the product of average injected current density J inj and S I = J S = J( x ) S( x ) (3) inj i i i Thus, L s and J 0 are derived from each J inj based on Eqs. (1)-(3). Next, the radiant power distribution R(x) is calculated based on the rate equation. When the current density J(x) flows over the active layer, the rate equation for carriers under steady state can be written as [20, 21]: J x η qd = An x + Bn x + Cn x (4) 2 3 ( ) inj /( ) ( ) ( ) ( ) where η inj, d and n are the injection efficiency, thickness of active layer and carrier concentration, respectively. A, B and C are the coefficients of Shockley-Read-Hall (SRH) recombination, radiative recombination and Auger recombination, respectively. According to the definition, internal quantum efficiency η IQE and radiative recombination efficiency η rad are expressed as η = η η (5) IQE inj rad (C) 2013 OSA 21 October 2013 Vol. 21, No. 21 DOI:10.1364/OE.21.025381 OPTICS EXPRESS 25384

rad Bn An Bn Cn Based on Eqs. (4)-(6), we can obtain 2 2 3 η = /( + + ) (6) 2 Bn = η ( )/( ) IQE J x qd (7) Besides, R(x) of active layer can be expressed as follows [10, 21]: Rx ( ) 2 = γ dbn (8) where γ is the average energy of a photon emitted from active layer, which is approximate to a constant. Therefore, substituting Eq. (7) into Eq. (8), we can obtain: Rx ( ) = γ Jx ( ) η / q (9) Equation (9) implies that R(x) depends on the current density. The non-uniform current density flowing over the active layer results in non-uniform radiate power distribution. In order to obtain the radiant power distribution R(x) of the active layer, the dependence of η IQE on J(x) is obtained from theoretical analysis based on the Ref [20]. The exponential expression of η IQE versus J(x) is obtained by curve fitting as IQE η IQE = a+ bexp( J( x) / c) (10) where a, b and c are fitting parameters. As demonstrated in Ref [20], η IQE versus J inj is only dependent on the maximum η IQE and the current density where η IQE reaches the maximum, therefore, a, b, and c are only dependent on these two parameters. Considering that the maximum η IQE of above 80% is obtained when the current density is always under 10 A/cm 2 for GaN-based LEDs, the fitting parameters of a, b and c with value of 0.50, 0.30 and 145.4 A/cm 2, respectively, are employed in the following simulation. In fact, the LEE obtained based on the different fitting parameters shows little variation due to the independence of LEE on η IQE. Thus, the radiant power distribution of the active layer is obtained according to Eqs. (1), (9) and (10). The top and bottom surfaces of the active layer are meshed and set as luminescent surfaces with the calculated non-uniform radiant power distribution at each injected current density J inj. When Monte Carlo ray-tracing simulation is carried out, photons are emitted from the active layer randomly and the power of each photon is dependent on its location. The total power of all the photons escaping into the air is collected by the detector that surrounds the LED chips. The refractive indices and absorption coefficients of sapphire, n-gan, active layer, p-gan and ITO are set to be 1.77, 2.45, 2.54, 2.45, and 2.0 and 0, 5 mm 1, 8 mm 1, 5 mm 1, and 0, respectively [7]. The reflectivity of bottom surfaces of the three kinds of LEDs is assumed to be 100% considering the reflectivity of the hybrid reflector, that is to say, when the photons transport to backside of sapphire substrate, they will be reflected back without any energy loss. The effect of photon recycling is not considered in this model [22]. 4. Results and discussion The sheet resistances of n-gan and ITO of our fabricated LEDs are 16.7Ω/ and 45.1Ω/ according to the four-point probe measurement, respectively. Current prefers to flow laterally in n-gan layer rather than the ITO layer due to the smaller sheet resistance of n-gan. As a result, the current crowds under the p-pad. The current spreading lengths L s versus injected current density J inj of LEDs without and with CBL calculated from Eqs. (1)-(3) are shown in Figs. 2(a) and 2(b), respectively. In order to compare the effects of different current crowding phenomena on the LEE, the current spreading lengths of LEDs with ITO sheet resistances of 30Ω/ and 60Ω/ are also calculated, as shown in Fig. 2. It can be seen that L s decreases significantly with the increase of injected current density, which implies that CCE becomes serious under high current [9, 15]. The decrease of ITO sheet resistance is helpful to alleviating the p-cce [15, 16, 23, 24], as shown in Fig. 2. The difference of sheet resistances (C) 2013 OSA 21 October 2013 Vol. 21, No. 21 DOI:10.1364/OE.21.025381 OPTICS EXPRESS 25385

between n-gan and ITO decreases with the decrease of ITO sheet resistance, resulting in a more uniform current distribution and longer L s [1]. Meanwhile, p-cce is alleviated by employing the CBL. Compared to the LEDs without CBL, L s of LEDs with CBL increases by 17.8%, 20.8%, and 18.2% for ITO sheet resistance of 30Ω/, 45.1Ω/, and 60Ω/ at 20 A/cm 2, respectively. Fig. 2. Current spreading length L s versus injected current density η inj of LED (a) without CBL and (b) with CBL. Figure 3(a) shows the current crowding effects on LEE of conventional LEDs with different ITO sheet resistances. For comparison, the n-cce is analyzed with the assumption that the ITO sheet resistance is 16.7Ω/ and n-gan sheet resistance varies from 30Ω/ to 60Ω/. The LEE of LEDs without CCE (J inj = 0) is 39.7%. It can be seen that LEE of LEDs with p-cce drops significantly. When current density increases from 0 to 70 A/cm 2, it is decreased by 18.8%, 22.0%, and 24.2% for ITO sheet resistances of 30Ω/, 45.1Ω/, and 60Ω/, respectively, as shown in Fig. 3(a). Lots of photons are emitted under p-pad due to p- CCE, and they are absorbed significantly by the metallic p-pad when they transport upwards for the first time, as illustrated in Fig. 1(a). The number of photon emitted under p-pad increases with degeneration of p-cce and consequently, the absorption by the p-pad increases. Thus, LEE drops remarkably with degeneration of p-cce caused by high current density as well as large ITO sheet resistance. However, n-cce shows little influence on LEE, as show in Fig. 3(b). This is due to the photons crowding around mesh edge are easy to escape into the air without less absorption [4], as illustrated in Fig. 1(a). As a result, LEDs with n-cce have a superior LEE. The radiant power under p-pad decreases due to the reduction of p-cce achieved by the CBL and improved ITO performance [15, 16, 23, 24]. As a result, the absorption of photons transporting upwards for the first time is decreased and subsequently, the LEE of LEDs with p-cce is increased. Figure 3(a) illustrates that the LEE droop of LEDs with p-cce is alleviated significantly due to CBL. Compared to LEDs without CBL at 20 A/cm 2, LEE of LEDs with CBL is increased by 5.5%, 7.7%, and 8.7% for ITO sheet resistances of 30Ω/, 45.1Ω/, and 60Ω/, respectively. At 70A/cm 2, LEE is found to be increased by 15.9%, 19.0%, and 21.5%, respectively. (C) 2013 OSA 21 October 2013 Vol. 21, No. 21 DOI:10.1364/OE.21.025381 OPTICS EXPRESS 25386

Fig. 3. LEE of LEDs with current crowding around (a) p-pad and (b) n-pad. On the other hand, a reflector underneath the electrode pad is introduced to decrease the absorption to improve LEE of LEDs with p-cce [15, 25]. Figure 4 depicts the effects of reflectivity of p-pad on the LEE of LEDs with ITO sheet resistance of 45.1Ω/. It can be seen that LEE increases with the increase of reflectivity. With the reflectivity of p-pad increasing from 30% to 85%, the LEE of LEDs without and with CBL increases by 13.4% and 7.6% at 20 A/cm 2, and 26.3% and 11.4% at 70 A/cm 2. That is to say, the LEEs of LEDs with RCBL are increased by 7.6% and 11.4% compared to these of LEDs with CBL at 20 A/cm 2 and 70 A/cm 2, respectively. Although the number of photons emitted under p-pad increases with current density, the absorption is decreased by increasing the reflectivity of p-pad. Fig. 4. LEEs of LEDs with ITO sheet resistance of 45.1Ω/ when reflectivity of p-pad varies. Figure 5 shows the L-I-V curves of LEDs without CBL, with CBL and with RCBL, respectively. The inset shows the cross-sectional TEM image of the electrode pad of LEDs with RCBL. It can be seen that the SiO 2 CBL, ITO layer, and Al metal reflector are subsequently deposited on epitaxial layer. Current density is found to crowd under the p-pad according to the measurement of the sheet resistances of ITO and n-gan. At injection current density of 20 A/cm 2, the forward voltages are 3.20 V, 3.27 V and 3.31 V for LEDs without CBL, with CBL, and with RCBL, respectively. The higher voltages of the LEDs with CBL and RCBL are attributed to larger series resistance caused by the less contact area of ITO and p-gan [23]. Compared to LEDs without CBL, the output power of LEDs with CBL is enhanced by 8.7% and 18.2% at 20 A/cm 2 and 70 A/cm 2, respectively. The output power of LEDs with RCBL is further enhanced by 8.9% and 12.7% compared to these of LEDs with CBL at 20 A/cm 2 and 70 A/cm 2, respectively. These results, which are accordance with the (C) 2013 OSA 21 October 2013 Vol. 21, No. 21 DOI:10.1364/OE.21.025381 OPTICS EXPRESS 25387

simulation results, illustrate that the CBL and RCBL are beneficial to preventing photons being emitted under p-pad and being absorbed, especially under high current density. Fig. 5. L-I-V relationships of LEDs without CBL, with CBL and with RCBL. The inset is TEM image of the deposited SiO 2 /ITO/Al layers underneath the p-pad of the LED with RCBL. 5. Conclusions In this paper, LEEs of the conventional GaN-based LEDs with current crowding around p-pad and n-pad are analyzed respectively by Monte Carlo ray-tracing method to optimize the design of LEDs. The improved LED model with non-uniform distribution of photons emitted from the active layer is constructed based on the rate equation and current spreading theory. The results show that current crowding around n-pad has little effect on the LEE, while current crowding around p-pad has a significant effect on the LEE due to the remarkably absorption of photons emitted under p-pad. Thus, high current density and/or inferior device parameters including low reflectivity of p-pad and large sheet resistance difference between n-gan and TCL lead to degeneration of LEE droop. In order to increase LEE, we can decrease the emission of photons under p-pad by CBL and/or decrease the absorption by reflective p-pad. The simulated results show that LEE of LEDs with CBL is enhanced by 7.7% and 19.0% and further enhanced by 7.6% and 11.4% with a reflective p-pad at 20 A/cm 2 and 70 A/cm 2, respectively. The enhancements are in accordance with the experimental results. Acknowledgments This work was supported by the State Key Development Program for Basic Research of China under Grant 2011CB013103 and National High Technology Research and Development Program of China under Grant 2011AA03A106. (C) 2013 OSA 21 October 2013 Vol. 21, No. 21 DOI:10.1364/OE.21.025381 OPTICS EXPRESS 25388