Analysis of the various light extraction efficiency enhancement mechanisms for the GaN-based light emitting diodes (LEDs) was investigated. Experiments utilized the imprinting technique to fabricate pyramid and inverted pyramid microstructures. Roughness treatment was then integrated with these imprinting structures on patterned sapphire substrate (PSS) LEDs. An approximate 33% improvement in light output power was obtained using the pyramid profile when compared with the planar LED. This was nearly 15% higher than that of the inverted pyramid profile. The roughness effect provided an approximate 5% efficiency enhancement. The total light enhanced efficiency increased to 85.9% by integrating the imprinting pyramid structure, PSS, and surface roughness.
© 2010 OSA
Although GaN-based light-emitting diodes (LEDs) have received great interests due to their small size, energy efficiency, longevity, and environmental soundness [1–3], a need for attaining higher external quantum efficiency (EQE) of LEDS continues to exist because of their wide applications. The EQE is decided by the internal quantum efficiency (IQE) and the light extraction efficiency (LEE). The IQE of GaN-based LEDs has greatly improved because of advances in crystal structure and quality in recent years, but the LEE of LED chips is still low because of the Fresnel loss and total internal reflection (TIR) . The poor LEE is because of the large refractive index difference and a smooth interface between air (n = 1) and the GaN (n = 2.5). In order to improve the LEE, various methods were proposed to increase the opportunity for the emitting photons to escape into free space. These methods include changes to the surface roughness on the LED surface layer [5,6], the use of a patterned sapphire substrate (PSS) [7–9] and the direct fabrication of perodical optical structures on a GaN surface [10–13]. These methods build micro- or nano-structures on the LED or the substrate surface. In this study, some factors influencing LEE were investigated. These included the use of the imprinting technique to fabricate micro-pyramid and inverted micro-pyramid arrays on a PSS LED surface that received etching treatment to produce nano scale roughness on the imprinted structures. Each experiment was analyzed to determine the influence of various factors on the LEE.
The GaN-base LED used in this study were grown on c-face (0001), two-inch conventional sapphire substrates (CSS) and PSS using a metal-organic chemical vapor deposition (MOCVD) system. Figure 1 shows the process by which 300 × 300 μm2 LED chips were fabricated.
2.1 PSS wafer preparation
The wet etching technique prepared the PSS. SiO2 film with a two- dimensional triangular array of circles, a 3μm diameter, and a 5 μm periodicity as an etching mask was patterned onto the sapphire via plasma enhanced chemical vapor deposition (PECVD) and standard photolithography. The substrate was then placed in an etching solution of H2SO4:H3PO4 with a mixture ratio of 5:1 at 300 °C. Finally, dilute HF etched the SiO2 mask away to complete the PSS process.
2.2 LED epitaxy
The LED layer structure consisted of a low temperature GaN nucleation layer, a 2.5 μm thick unintentionally doped GaN layer, a 3 μm thick n-type GaN layer, an active region with 10 periods of InGaN/GaN multiple quantum wells (MQWs), a 30 nm thick Mg-doped p-Al0.15Ga0.85N cladding layer (p = 5 × 1017 cm−3), and a 200 nm thick Mg-doped first p+-GaN contact layer (n = 7 × 1017 cm−3). The active layer consists of a 2.3 nm thick InGaN-well layer and a 13 nm thick GaN-barrier layer for the InGaN/GaN MQW LED structures.
2.3 Conventional chip process
The fabricated LED sample had indium tin oxide (ITO) evaporated onto it as a transparent conductive layer. Inductively coupled plasma (ICP) partially etched the LED sample to expose the n-GaN. Optical lithography defined the ITO pattern, and wet etching exposed the p-GaN layer. Thermal evaporation with rapid thermal annealing to create the p- and n-electrodes deposited Cr/Au on the p-GaN and n-GaN surfaces.
2.4 Surface structure imprinting
The imprinting technique built microstructure pyramids and inverted pyramids on the LED surface.
2.4.1 Pyramid structure
An imprinting technique using a silicon mold created the pyramid surface structures. The mold patterns reflected the mask layout: a two- dimensional triangular array of circles with a 3 μm diameter and 5 μm periodicity. The fabrication of the silicon molds used photolithography and the wet etching process. Finally, the mold patterns became an inverted pyramid array.
Pyramid microstructures were then built on the LED samples. The polymethylmethacrylate (PMMA, (950 PMMA A4, Micro. Chem.)) was spun onto the substrate to produce a uniformly thick, 0.7 μm film across the substrate. To reduce the friction between the PMMA and the mold and improve the replica quality during demolding, the polytetrafluoroethylene (PTFE, (601S1-100-6, Dupont)) was coated onto the mold as an anti-adhesive film. The mold and substrate were then sent to press at a temperature of 150 °C and a pressure of 1500 Nt/cm2. After imprinting, a partial etching process was used to remove the imprinting structures on the p- and n-contacts.
2.4.2 Inverted pyramid structure
A reversal imprinting technique with a polydimethylsiloxane (PDMS, ((SylgardTM 184, Dow Corning)) mold created the inverted pyramid surface structures. A PDMS elastomeric mold with a pyramid pattern was cast on the silicon mold as mentioned above. PDMS is flexible, thermally stable, and conforms in such a manner as to allow complete conformal contact with the substrate (Normally, sapphire wafers warp 0 ~10 μm). This mold could be used for large area imprinting, as it is cost effective and a commonly used soft mold material. Instead of spinning the PMMA onto the substrates, the PMMA was spun onto the PDMS mold after spraying the PTFE. Much less pressing force was required (27 NT/cm2 in this experiment) to fabricate microstructures on the LED surface since the structures had already been formed by the mold pattern. The necessary requirement of the reversal imprinting is the mold had a lower surface energy than the substrate, which allowed the polymer material to adhere better to the substrate and detach successfully from the mold.
High-density plasma was used to generate ion bombardment and create nano roughness on the surface of the imprinting microstructure. This gave the photons generated in the LEDs increased opportunities for escape into free space.
To demonstrate the light extraction enhancement effect of the PSS, imprinting surface structure, and roughness, conventional LED chips and LED chips with a planar PMMA covered layer were prepared.
Finally, a high current measure unit (HP4156C) measured the current–voltage (I–V) by injecting different amounts of DC current into the LEDs. An integrated sphere with a calibrated power meter measured the light output power of the LEDs.
3. Results and discussions
3.1 SEM morphology
Two groups of LED chips with different extraction mechanisms were prepared. The C group of Table 1 shows the LEDs that were grown on the CSS, and the P group of Table 2 shows the LEDs grown on the PSS. In Table 1, C1 is a conventional LED, C2 is a single sprayed lasyer of PMMA on the LED, C3 is same as C2 but has extra roughness treatment. C4 represents the imprinting pyramid array applied onto the conventional LED, and C5 has roughness added to the pyramids. The imprinting of the inverted pyramid array with and without roughness onto the conventional LEDs are labeled as C6 and C7 respectively.
Figure 2 (a) and 2(b) show the imprinting molds of silicon and PDMS. The depth and height of the pyramids and inverted pyramids are 2.2~2.3 μm. Figure 3 (a) and 3(b) are the imprinted patterns made by the molds of Fig. 2(a) and 2(b). In Fig. 3(a), the thickness of the residual layer is 0.23 μm, and the height of the imprinting pyramid is 1.91 μm shorter than the depth of the mold because the PMMA did not completely fill the silicon mold. In Fig. 3(b), the depth of the imprinting inverted pyramidis 1.26 μm, which is much less than the height of PDMS mold. This is because PDMS is an elastomeric material and the mold deforms during embossing even under light pressure.
Figure 4 shows the SEM pictures of the imprinting structure after the HDP process. The figure reveals no evident roughness on the pyramid and roughness on the planar surface only (4(b)), but clear textures are found on both of the concave and planar surfaces (4(a)). The mean roughness, as measured by AFM, was 2.3 nm and 21.3 nm, before and after the HDP process.
3.2 Light extraction mechanisms analysis
3.2.1 Conventional LEDs
Figure 5 shows the light output power versus the injection current of the LED chips mentioned in Table 1. Table 3 shows the output power for C group LEDs under a 20 mA current injection and the relative enhanced optical efficiency as compared to C1.
The C2 chip has a slightly higher efficiency than C1. The refractive index of PMMA is 1.5, therefore, the critical angle of total internal reflection for PMMA/GaN is 36.9° larger than the critical angle of 23.6° for air/GaN. But TIR exists between air/PMMA at the critical angle of 41.8°. This indicates that some portions of the light in the PMMA layer are still trapped in the semiconductor (Fig. 6 (a) ). The slight 3.83% improvement in efficiency is because of a reduction of Fresnel loss for the buffer layer of PMMA that exists between air and the GaN. Furthermore, a roughened surface can ruin TIR of a planar surface and scatter the light out to free space (Fig. 6 (b)). Fujii et al.  noted that the size of roughened features influences the performance of extraction efficiency. The efficiency of C3 is nearly 4.4% higher than that of C2, which means that the surface texture in C3 is useful to extract trapped light but not strong enough to provide an intense scattering effect.
For the C4 sample, the output power was greatly enhanced to 32.92%. In addition to reduce the Fresnel loss, the imprinting pyramid arrays created more opportunities for photons to escape into free space. The greater extraction area and the peculiar geometry to destroy TIR are two major factors in enhancing optical efficiency (Fig. 6 (c)). For the inverted pyramids, C6, the light extraction efficiency was lower than that of the pyramid, C4. This may be because the inverted structure was shallow due to the deformation caused by the imprinting PDMS mold, which reduced the extraction area. However, the major reason for the lower extraction efficiency was that the photons in C6 traveled a longer distance to free space than those of C4 (Fig. 6 (d)). This means that more light was absorbed by using the inverted pyramid arrays. We suggest that even if the depth of the inverted pyramids were the same as the height of the pyramids, the extraction efficiency of pyramid LEDs would still be superior to that of the inverted pyramid LEDs.
Figure 4 (b) shows no roughened effect on the pyramid of C5. But from Table 3, the increased efficiencies of C4 to C5 is nearly the same as C6 to C7. Even though the inverted pyramid surface was roughened, the optical efficiency did not increase. The contribution of the convex and concave structures (large size feature) is same as the roughened effect (small szie feature) to help trapped photons out of free sapce. The increased efficiencies of C5 and C7 compared to C4 and C6 mainly come from the roughness of the planar surface, not from the pyramid or inverted pyramids. As mentioned above, the roughened surface is still not strong enough to enhance a great amount of extraction. Increasing the intensity or processing time of the HDP could produce a more roughened surface, however, this could damage the imprinting structures.
3.2.2 PSS LEDs
Figure 7 shows the light output power versus the injection current of the LED chips mentioned in Table 2. Table 4 shows the output power for P group LEDs under 20 mA current injection and relative enhanced optical efficiency compared to P1. Similar to the results obverved in Fig. 5, the light extraction efficiency enhancements of the PSS chips with imprinting structures and roughened are higher than that of the PSS chip alone. The dominant wavelengths of both C1 and P1 are 458 nm. The output power of P1 was 39.3% higher than that of C1 at an injection current of 20 mA. The enhancement of light extraction efficiency obtained by the PSS growth technique can be attributed to the random and multiple reflections at the interface between the GaN and the sapphire, where photons could find escape cones. Moreover, the PSS growth technique could also decrease the dislocation density, which resulted from the lattice mismatch between the GaN and sapphire substrate, to increase internal quantum efficiency . When compared to the same condition to spray a PMMA layer on LEDs, the degree of enhancement for P2 as compared with P1 was greater than that of C4 as compared with C1. This result is attributed to the inclined angle of the substrate texture which provided photons trapped inside the LED to find escape cones more effectively.
Likewise, C4 benifitted from the pyramids on the LED surface. Significant light extraction efficiency enhancement was observed from P4. However, the degree of enhancement for C4 as compared with C1 was greater than that of P4 as compared with P1. Since the conventional LED traps more emitting photons than that of the PSS LED, once the surface texture was integrated into the devices, the relative ratio of free photons of C4 to free photons of C1 was higher than that of free photons of P4 to free photons of P1. When comparing P4 with C1, the overall enhancement of pyramids on the PSS LED jumps to 85.9%. The simulation results of Lee et al.  show that no matter what the PSS or surface texture, both are useful in improving the light extraction efficiency. However, once both schemes are integrated into the LED, the efficiecny was nearly 10% higher than any other individual technique. The main difference is that our microstructures were made of PMMA formed onto the LED surface instead of directly fabricating the GaN texture as in Lee’s study. The PMMA buffer between the GaN and air is obvious to reduce Fresnel loss and TIR as mentioned above.
The slanted angle of the microstructures also influenced the light extraction efficiency. Figure 8 shows a pyramid array in which the angle is only 11.3° with a poor light extraction enhancement of only 12.5%. This experimantal result conforms to the numerical simulation which demonstrated that pyramid angles from 20° to 70 o have similar enhanced efficiency, but that the efficiency drops down abruptly once the angles are out of this range .
The light extraction efficiency enhancements attributed to different mechanisms were analyzed in this paper.We fabricated a series of LED chips with the imprinting structures, the PSS, and surface roughness. The microtexture on the LED surface or the substrate greatly improved the light extraction efficiency to 30~40%. The nanoscale roughness on the surface caused an additional gain of 4~5%. Stronger ion bombardment might create a more roughened surface to obtain higher efficiency, but it might also damage the imprinting structure. The maximum output power efficiency in the present study was increased to 85.9% by integrating the imprinting pyramid array, the PSS, and the roughness effect into the LEDs.
The authors acknowledge funding supports from National Science Council of Taiwan under grant no. NSC 98-2221-E-033-045-MY2, as well as by the project under specific research fields in the Chung Yuan Christian University of Taiwan under grant no. CYCU-98-CR-ME.
References and links
1. M. R. Krames, O. B. Shchekin, R. Mueller-Mach, G. O. Mueller, L. Zhou, G. Harbers, and M. G. Craford, “Status and Future of High-Power Light-Emitting Diodes for Solid-State Lighting,” J. Display Technol. 3(2), 160–175 (2007). [CrossRef]
2. T. Gessmann and E. F. Schubert, “High-efficiency AlGaInP light-emitting diodes for solid-state lighting applications,” J. Appl. Phys. 95(5), 2203–2216 (2004). [CrossRef]
4. A. David, T. Fujii, B. Moran, S. Nakamura, S. P. DenBaars, C. Weisbuch, and H. Benisty, “Photonic crystal laser lift-off GaN light-emitting diodes,” Appl. Phys. Lett. 88(13), 133514 (2006). [CrossRef]
5. T. Fujii, Y. Gao, R. Sharma, E. L. Hu, S. P. DenBaars, and S. Nakamura, “Increase in the extraction efficiency of GaN-based light-emitting diodes via surface roughening,” Appl. Phys. Lett. 84(6), 855–857 (2004). [CrossRef]
6. C. F. Lin, Z. J. Yang, J. H. Zheng, and J. J. Dai, “Enhanced light output in nitride-based light-emitting diodes by roughening the mesa sidewall,” IEEE Photon. Technol. Lett. 17(10), 2038–2040 (2005). [CrossRef]
7. Y. J. Lee, H. C. Kuo, T. C. Lu, B. J. Su, and S. C. Wang, “Fabrication and Characterization of GaN-Based LEDs Grown on Chemical Wet-Etched Patterned Sapphire Substrates,” J. Electrochem. Soc. 153(12), G1106–G1111 (2006). [CrossRef]
8. S. J. Chang, Y. C. Lin, Y. K. Su, C. S. Chang, T. C. Wen, S. C. Shei, J. C. Ke, C. W. Kuo, S. C. Chen, and C. H. Liu, “Nitride-based LEDs fabricated on patterned sapphire substrates,” Solid-State Electron. 47(9), 1539–1542 (2003). [CrossRef]
9. R. H. Horng, W. K. Wang, S. C. Huang, S. Y. Huang, S. H. Lin, C. F. Lin, and D. S. Wuu, “Growth and characterization of 380-nm InGaN/AlGaN LEDs grown on patterned sapphire substrates,” J. Cryst. Growth 298, 219–222 (2007). [CrossRef]
10. Y. C. Lee, M. J. Ciou, and J. S. Huang, “Light output enhancement for nitride-based light emitting diodes via imprinting lithography using spin-on glass,” Microelectron. Eng. in press.
11. Y. K. Ee, P. Kumnorkaew, R. A. Arif, H. Tong, J. F. Gilchrist, and N. Tansu, “Light extraction efficiency enhancement of InGaN quantum wells light-emitting diodes with polydimethylsiloxane concave microstructures,” Opt. Express 17(16), 13747–13757 (2009). [CrossRef]
12. K. T. Lee, Y. C. Lee, and J. Y. Chang, “Light Extraction Enhancement of Gallium Nitride Epilayers with Stripe Pattern Transferred from Patterned Sapphire Substrate,” J. Electrochem. Soc. 155(9), H638–H641 (2008). [CrossRef]
13. H. Ichikawa and T. Baba, “Efficiency enhancement in a light-emitting diode with a two-dimensional surface grating photonic crystal,” Appl. Phys. Lett. 84(4), 457–459 (2004). [CrossRef]
15. 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). [CrossRef]