We demonstrated improved luminous efficacy for GaN-based vertical light emitting diodes (VLEDs) employing a low index layer composed of silicon dioxide (SiO2) on the top surface. Three-dimensional ðnite-difference time-domain simulations for the fabricated VLED chip show that the penetration ratio of the emitted/reflected light into the VLED chip decreased by approximately 20% compared to a normal VLED chip. This result is in good agreement with an empirical study stating that white VLEDs having a SiO2 layer exhibit an 8.1% higher luminous efficacy than white VLEDs with no layer at an injection current of 350 mA. Photons penetrating into the VLED chip, which become extinct in the VLED chip, are reflected from the SiO2 layer due to the index contrast between the SiO2 layer and epoxy resin containing phosphor, with no degradation of the light-extraction efficiency of the VLED chip. As such, this structure can contribute to the enhancement of the luminous efficacy of VLEDs.
© 2013 OSA
Recently, there has been considerable progress made in GaN-based light emitting diodes (LEDs), enabling them to be used in the fields of full-color displays, automotive lighting, traffic lights, LCD backlight units (BLUs), and solid-state lighting owing to their high energy efficiency and reliable lifetime [1–3]. Among various types of LEDs, phosphor-converted white LEDs composed of a blue-emission LED chip and yellow phosphor have been regarded as a promising and efficient technology . Advantages of this method for the generation of a white spectrum are their lower relative cost, simple fabrication process, and higher energy conversion efficiency compared to mixing individual red, green, and blue LEDs . However, the performance of white LEDs remains insufficient for significant penetration into commercial lighting markets; there are two major issues associated with their use: the high cost per lumen of LEDs, and their complex thermal management. Hence, both greater performance and cost improvements of LEDs are required before they can be sufficiently commercialized [6, 7].
It is known that low phosphor-conversion efficiency is mainly due to the low forward efficiency of the phosphor layer on the LED chip . For example, in the Fig. 1 , only 40% of white light emitted/transmitted (region I) at the phosphor propagates in the forward direction, whereas 60% propagates in the backward direction (region II). Also, in the region III, the part of white light in backward direction at the region II is penetrated into a vertical LED (VLED) or is re-reflected on the textured GaN surface. As a result, a significant ratio of light penetrates and then dissipates into a VLED chip [9–13]. To date, several optical structures have been implemented in attempts to improve the forward conversion efficiency in phosphor-converted white LEDs, such as scattered photon extraction (SPE) LEDs , an LED package with a diffuse reflector cup , band pass filters , and a remote phosphor structure with a hemispherical dome [13,14]. However, these methods require both a complicated optical design and several processes to produce high-efficiency LEDs, which subsequently makes them lose their cost competitiveness in LED markets.
In this study, to prevent penetration into the VLED chip by emitted/reflected light at the phosphor, we propose that SiO2 be used as a low index layer on GaN-based VLEDs. The effects of the SiO2 layer on the luminous power of phosphor-converted white LEDs are then theoretically investigated using the three-dimensional ðnite-difference time-domain (3D-FDTD) simulation method.
2. 3D-FDTD simulation model
The 5 × 5 × 5 μm3 simulation structure was constructed as follows: the four sides of the structure were surrounded by periodic boundary conditions (PBCs), whereas the top and bottom were surrounded by perfectly matched layers (PMLs). A periodic hexagonal cone array (1.2-μm-periode, 900-nm-height and 56° sidewall angle) was placed on a 2-μm-thick GaN layer (refractive index, n = 2.46), and the conformally coated SiO2 layer (n = 1.46) was placed on a textured surface having a variable thickness. The reflective mirror, composed of silver, was located on the bottom of the GaN layer and the complex dielectric constant of silver at the optical wavelength was then approximated using the Drude model . Finally, the entire structure was covered in epoxy resin having a variable refractive index (1.40~1.55).
Computations were carried out for the three simulation structures in order to evaluate the effect of SiO2 thickness on both the light extraction efficiency of VLEDs and the penetration ratios of the blue and yellow light. As shown in Fig. 2(a) , to evaluate the effect of the SiO2 thickness on the light extraction efficiency of VLEDs, the detection plane was located on the top of simulation structure, and the point-dipole source (center wavelength of 450 nm and spectral width of 15 nm) were excited at the multiple quantum wells (MQWs). Figures 2(b) and 2(c) show that the detection plane was located on the bottom of simulation structure, and that the point-dipole sources, which have a center wavelength of 450 nm and spectral width of 15 nm (blue) and a center wavelength 560 nm and spectral width of 100 nm (yellow), were excited at a distance of 500 nm from the top, in order to evaluate the effect of the SiO2 thickness on the penetration ratio of the blue and yellow light into the VLED chip. The reflective mirror was removed from the simulation structure to enable a more precise assessment [Figs. 2(b) and 2(c)].
3. Results and discussion
3.1. 3D-FDTD simulation results
The results of the 3D-FDTD simulation are summarized in Fig. 3 . First, we calculated the effect of SiO2 thickness on the light extraction efficiency of the VLED chip (Fig. 3(a)). When the refractive index of the epoxy resin was lower or higher than that of SiO2 (n = 1.46), the light extraction efficiency somewhat decreased with an increasing SiO2 thickness. This degradation in light extraction efficiency can be attributed to the insertion of SiO2 layer between the textured GaN and epoxy resin, as it has a lower index. On the other hand, when the refractive index of the epoxy resin was equal to that of SiO2, the light extraction efficiency did not change with an increase in the SiO2 thickness.
Figures 3(b) and 3(c) show the penetration ratios of blue and yellow light into the VLED chip; results of the 3D-FDTD simulation for each were similar. In the case that the refractive index of epoxy resin is equal to that of SiO2 (n = 1.46), the penetration ratio did not change with the SiO2 thickness for either the blue or yellow light. When the refractive index of epoxy resin (n = 1.40) was lower than that of SiO2 (n = 1.46), the penetration ratio slightly increased due to the insertion of an index layer lower than the epoxy resin. However, when the refractive index of epoxy resin (n = 1.50, 1.55) was higher than SiO2 (n = 1.46), the penetration ratios of the blue and yellow light continuously decreased, ultimately becoming saturated at an SiO2 thickness above 900 nm. The light propagating from the epoxy resin to the GaN layer is reflected on the textured SiO2 layer covered surface due to the index difference between SiO2 (n = 1.46) and epoxy resin . In addition, the saturated penetration ratio at a SiO2 thickness above 900 nm is because the layer has become thicker than the hexagonal cone height of the simulation condition (900 nm). That is, when the thickness of the SiO2 layer is thinner than the height of the hexagonal cone, the penetration ratio was gradually decreased until 900 nm, because the thickness of the SiO2 layer is not enough covering the textured GaN surface. However, with thickness of the SiO2 layer being over the 900 nm, the penetration ratio was saturated due to perfectly covered the GaN surface. In addition, the decrease of the penetration ratio can be explained as follows Figs. 3(d) and 4(a) . To quantify the effect of SiO2 thickness on the light diffuse transmittance of a textured GaN layer covered with an SiO2 layer, we measured the diffuse transmittance for different SiO2 thicknesses using an integrating sphere [inset of Fig. 3(d)]. During these measurements, the direction of incident light was from the epoxy resin to the GaN layer. The diffuse transmittance of the textured GaN layer that was covered with an SiO2 layer significantly decreased for increased SiO2 thicknesses, for the entire wavelength, as shown in Fig. 3(d). These results indicate that the SiO2 layer plays a role in preventing light from penetrating into the VLED chip. Therefore, the reduction of penetration ratio with increasing the SiO2 thickness implies that the SiO2 layer reflects light propagating toward the VLED chip. That is, the white luminous power of VLED is enhanced, because the emitted/reflected light by phosphor to VLED chip is obtained the probability of transmission via phosphor by re-reflection on SiO2 layer.
Figure 4(a) shows the electric ðeld intensity proðles for each SiO2 thickness (0 nm, 450 nm, and 900 nm). The figure clearly shows that photons propagating to the GaN layer are reflected on the textured GaN layer that is covered by the SiO2 layer, which prevents the absorption of photons toward the VLED chip. At 450 nm, the thickness of the SiO2 layer did not perfectly cover the height of the hexagonal cone. Accordingly, when the light at dipole source propagated to the VLED chip, penetration was more dominant than reflection due to the SiO2 layer being thinner than the height of the hexagonal cone. However, in the 900-nm-thick SiO2 layer, light reflection on the SiO2 surface was dominant, because the SiO2 layer perfectly covered the height of the hexagonal cone. Therefore, when the thickness of the SiO2 layer is increased to above the height of the hexagonal cone, the penetration ratio became saturated, as shown in Figs. 3(b) and 3(c). Figure 4(b) present a scanning electron microscopy (SEM) images showing that the 900-nm-thick SiO2 layer is deposited conformally on the textured surface of the GaN-based VLED. The inset is tilted SEM image. The random textured surface is formed of the hexagonal cone-shaped and the cone height was about 0.75–0.95 μm (average 0.87 μm) [17,18]. As the shape of the SiO2 layer is deposited on the textured GaN surface, it follows the GaN surface geometry.
3.2. Experimental results
For a practical demonstration, a 1 mm x 1 mm size of VLED chip was fabricated using laser lift-off (LLO) and wafer bonding processes, as shown in Fig. 5(a) . After completion of the VLED fabrication, the SiO2 layer was deposited via plasma enhanced chemical vapor deposition method such that it covered the entire textured n-GaN surface of the VLEDs, except for the wire-bonding electrode region; the thicknesses of the SiO2 layer were 300 nm, 600 nm, and 900 nm. In the LED package process, the VLED chips were mounted on the center of 9080 lead-frame with size of 9 mm x 8 mm using Ag paste, as shown in Fig. 5(b). Finally, the VLED chips were molded by the transparent epoxy-resin (n = 1.52) for blue-emission and by epoxy-resin containing the yellow phosphors, respectively. In this study, the phosphor typically used the Cerium-doped Yttrium Aluminum Oxide (Y3Al5O12:Ce, Nemoto model).
The respective forward voltages of the blue-white emission VLEDs with 0 nm, 300 nm, 600 nm, and 900 nm-thick-SiO2 layers were 3.48, 3.44, 3.45, 3.50 V at 350 mA, respectively, as shown in Fig. 6(a) . The forward voltages of VLEDs are nearly similar, but the observed little variation is due to the difference of location in same epi-wafer. Therefore, the forward voltages of the blue-emission VLED are seen to be independent of SiO2 thickness on the textured surface of the VLEDs. Figure 6(b) shows the emission spectra of the white-emission VLED packages. The emission intensities for the blue and yellow regions improved with a thicker SiO2 layer, due to re-reflection on the SiO2 layer. The dominant wavelengths of the blue-emission VLEDs were about 450 nm and the correlated color temperatures (CCT) of white-emission VLEDs were about 4030 K at 350 mA. And the color rendering index (CRI) of VLEDs with 0 nm, 300 nm, 600 nm, and 900 nm-thick-SiO2 layers were 61.1, 60.5, 60.5, 60.9, respectively. Figure 6(c) shows the optical output power (blue line) of the blue-emission VLED and the luminous efficacy (red line) of a white-emission VLED operating at 350 mA in an integrating sphere. By increasing the SiO2 layer thickness, the optical output power of the blue-emission VLEDs slightly decrease from 364 mW to 351 mW. This result was attributed to an increase in the total internal reflection of photons emitted in MQWs due to the insert of a low index layer (n = 1.46) between the GaN (n = 2.46) and epoxy-resin (n = 1.52) layers. Furthermore, as shown in Fig. 4 (a), the photons generated in the MQWs increasingly have difficulty escaping the VLED chip due to the thick SiO2 layer that eventually covers the hexagonal cone. However, the luminous efficacy of the white-emission VLED increased from 80.8 lm/W to 87.4 lm/W when increasing the thickness of the SiO2 layer due to re-reflection on the SiO2 layer. Notably, enhancement in the luminous efficacy of the VLED with the 900-nm-thick SiO2 layer compared to that of the VLED with no SiO2 layer was 8.1% at 350 mA.
These experimental results were in good agreement with the 3D-FDTD simulations. Such improvements could be ascribed to the reduced penetration ratio of the backward emitted/reflected light into the VLED chip, which prevented the extinction of light in the VLED structure. However, the improvement of luminous efficacy remained somewhat lower than was expected in the 3D-FDTD simulation. It is posited here that this low improvement is caused from the geometry of the dichromatic white LED packages, which include yellow phosphors. In a conventional structure, the phosphor particles are not near the VLED chip so that photons emitted/reflected at the phosphor propagate toward not only to the VLED chip but also the submount of the LED package.
We demonstrated an improvement in the luminous efficacy for GaN-based white VLEDs covered with SiO2 as a low index layer. Even though the output power of the blue-emission VLEDs were somewhat decreased, the luminous efficacy of the white-emission VLED increased by about 8.1% compared to normal VLEDs. The enhanced luminous efficacy was subsequently attributed to the re-reflection of backward emitted/reflected light (blue and yellow) onto the surface of the VLED that was covered with SiO2 layer.
This work was supported by Enhancement in Technical Capability of Photonics Industry funded by the Ministry of Knowledge Economy (12210-12-1001). The work was also partially supported by Green-manufacturing process technology development funded by The Small & Medium Business Administration (S2023367).
References and links
1. S. Noda and M. Fujita, “Light-emitting diodes: Photonic crystal efficiency boost,” Nat. Photonics 3(3), 129–130 (2009). [CrossRef]
2. J. J. Wierer Jr, A. David, and M. M. Megens, “III-nitride photonic-crystal light-emitting diodes with high extraction efficiency,” Nat. Photonics 3(3), 163–169 (2009). [CrossRef]
3. S. Pimputkar, J. S. Speck, S. P. DenBaars, and S. Nakamura, “Prospects for LED lighting,” Nat. Photonics 3(4), 180–182 (2009). [CrossRef]
4. N. Narendran, N. Maliyagoda, L. Deng, and R. M. Pysar, “Characterizing LEDs for general illumination applications: mixed-color and phosphor-based white sources,” Proc. SPIE 4445, 137–147 (2001). [CrossRef]
5. T. Nishida, T. Ban, and N. Kobayashi, “High-color-rendering light sources consisting of a 350-nm ultraviolet light-emitting diode and three-basal-color phosphors,” Appl. Phys. Lett. 82(22), 3817–3819 (2003). [CrossRef]
6. H. C. Chen, K. J. Chen, C. H. Wang, C. C. Lin, C. C. Yeh, H. H. Tsai, M. H. Shih, H. C. Kuo, and T. C. Lu, “A novel randomly textured phosphor structure for highly efficient white light-emitting diodes,” Nanoscale Res. Lett. 7(1), 188–192 (2012). [CrossRef]
7. T. Doan, C. Tran, C. Chu, C. Chen, W. H. Liu, J. Chu, K. Yen, H. Chen, and F. Fan, “Vertical GaN based light emitting diodes on metal alloy substrate boosts high power LED performance,” Proc. SPIE 6669, 666903, 666903-8 (2007). [CrossRef]
8. N. Narendran, Y. Gu, J. P. Freyssinier-Nova, and Y. Zhu, “Extracting phosphor-scattered photons to improve white LED efficiency,” Phys. Status Solidi A 202(6), R60–R62 (2005). [CrossRef]
9. J. K. Kim, H. Luo, E. F. Schubert, J. Cho, C. Sone, and Y. Park, “Strongly enhanced phosphor eﬃciency in GaInN white light-emitting diodes using remote phosphor conﬁguration and diﬀuse reﬂector cup,” Jpn. J. Appl. Phys. 44(21), L649–L651, L651 (2005). [CrossRef]
10. K. Yamada, Y. Imai, and K. Ishii, “Optical simulation of light source devices composed of Blue LEDs and YAG phosphor,” J. Light Vis. Environ. 27(2), 70–74 (2003). [CrossRef]
11. J. R. Oh, S. H. Cho, Y. H. Lee, and Y. R. Do, “Enhanced forward efficiency of Y3Al5O12:Ce3+ phosphor from white light-emitting diodes using blue-pass yellow-reflection filter,” Opt. Express 17(9), 7450–7457 (2009). [CrossRef]
12. N. T. Tran, J. P. You, and F. G. Shi, “Effect of phosphor particle size on luminous efficacy of phosphor-converted white LED,” J. Lightwave Technol. 27(22), 5145–5150 (2009). [CrossRef]
13. S. C. Allen and A. J. Steckl, “A nearly ideal phosphor-converted white light-emitting diode,” Appl. Phys. Lett. 92(14), 143309 (2008). [CrossRef]
14. H. Luo, J. K. Kim, Y. A. Xi, E. F. Schubert, J. Cho, C. Sone, and Y. Park, “Trapped whispering-gallery optical modes in white light-emitting diode lamps with remote phosphor,” Appl. Phys. Lett. 89(4), 041125 (2006). [CrossRef]
15. S. K. Kim, H. K. Cho, D. K. Bae, J. S. Lee, H.-G. Park, and Y.-H. Lee, “Efficient GaN Slab vertical light-emitting diode covered with a patterned high-index layer,” Appl. Phys. Lett. 92(24), 241118 (2008). [CrossRef]
16. E. F. Schubert, Light-Emitting Diode, 2nd ed. (Cambridge University, 2006).
17. H.-S. Kwack, H. S. Lim, H.-D. Song, S.-H. Jung, H. K. Cho, H.-K. Kwon, and M. S. Oh, “Experimental study of light output power for vertical GaN-based light-emitting diodes with various textured surface and thickness of GaN layer,” AIP Advances 2(2), 022127 (2012). [CrossRef]
18. H. M. Ng, N. G. Weimann, and A. Chowdhury, “GaN nanotip pyramids formed by anisotropic etching,” J. Appl. Phys. 94(1), 650–652 (2003). [CrossRef]