High-efficiency InGaN-based LEDs grown on patterned sapphire substrates using nanoimprinting technology

(1)

High-efﬁciency InGaN-based LEDs grown on patterned sapphire substrates using

nanoimprinting technology

Yeeu-Chang Lee

a,⇑

_{, Shiang-Chih Yeh}

a

_{, Yen-Yu Chou}

a

_{, Pei-Jung Tsai}

b

_{, Jui-Wen Pan}

b

_{, Hsiu-Mei Chou}

c

_,

Chia-Hung Hou

c

, Yung-Yuan Chang

d

, Min-Sheng Chu

d

, Cheng-Hui Wu

d

, Chun-Hsien Ho

d

a

Dept. of Mechanical Eng., Chung Yuan Christian Univ., Chung Li 32023, Taiwan

b_{Institute of Photonic Systems, National Chiao Tung Univ., Tainan 71150, Taiwan} c

Lextar Electronics Corp., Hsinchu 30075, Taiwan

d

Procrystal Technology Co. Ltd., Yilan 27049, Taiwan

a r t i c l e

i n f o

Article history:

Received 28 November 2012 Accepted 17 January 2013 Available online 29 January 2013 Keywords:

Roller imprint lithography Light-emitting diodes Light extraction efﬁciency

a b s t r a c t

This study employed roller imprint lithography and dry etching to fabricate patterned sapphire sub-strates (PSSs) of convex-shape with features of various heights. A soft polymer, polydimethylsiloxane (PDMS), was used as a mold to duplicate the pattern of a hard silicon template. The imprinted material was spin deposited onto a PDMS mold and transferred to the sapphire substrate using roller imprinting equipment. Inductive coupled plasma (ICP) etching was then used to fabricate the PSS. After epitaxial growth and chip processing, the current–voltage characteristics and light output of various LEDs were measured. The results demonstrate that the PSS process did not detract from the electrical properties of the LEDs; in fact, the output power of the proposed PSS LEDs was 25–30% greater than that of conven-tional LEDs. Simulation results show that PSS LEDs with structures of various heights would enhance optical efﬁciency in a manner similar to that demonstrated in these experiments.

1. Introduction

GaN-based light-emitting diodes (LEDs) have attracted consid-erable attention due to their small size, energy efficiency, longev-ity, and environmental friendliness [1–3]. However, the large difference in refractive index through the smooth interface be-tween air (n = 1) and GaN (n = 2.5) can lead to considerable Fresnel loss and total internal reflection (TIR)[4], with a subsequent de-crease in light extraction efficiency (LEE). The external quantum efficiency (EQE) of LEDs is determined by the internal quantum efficiency (IQE) and LEE. The IQE of GaN-based LEDs has been greatly improved because of advances in the crystal quality in re-cent years; however, increasing the LEE of LED chips remains a key factor in attaining higher EQE. In efforts to improve the LEE, various methods have been proposed to make it easier for emitted photons to escape into free space. These methods include altering the shape geometry of LEDs[5], surface roughening[6–7]and the use of a patterned sapphire substrate (PSS)[8–10]. Among these, the use of PSSs has been widely adopted in the industry because epitaxial quality and LEE can be improved simultaneously [11– 13]. Previous research has indicated that LEE increases with a de-crease in structural spacing [14]; however, decreasing structural

spacing to the sub-micrometer scale is difﬁcult to achieve using conventional photolithographic techniques. This study fabricated convex-shaped PSSs with sub-micron spacing and features of var-ious heights using a rapid, low cost process, involving roller imprinting with dry etching. GaN-base LEDs were then prepared on the PSS using epitaxial growth and conventional chip fabrica-tion processes. Finally, we investigated how the height of struc-tures on the PSS inﬂuences light output.

2. Experiments

2.1. Fabrication of PDMS mold

Fig. 1(a) illustrates the process used in the preparation of the Si template. A SiO2layer (300 nm thickness) and photoresist (Shipley

1818) were spin deposited onto a two-inch Si substrate, respec-tively. The patterns were deﬁned as an etching mask during induc-tively coupled plasma (ICP) etching via photolithography. The width of the cylinderical structures was 1.9

l

m with spacing of 0.6

l

m, as shown inFig. 1(b). The heights of structures on the fab-ricated Si templates were 2

l

m, 2.4

l

m, and 3

l

m, respectively.

A ﬂexible mold was used for the non-planar imprinting process to ensure high imprint quality.Fig. 2illustrates the procedure of fabricating the PDMS (Polydimethylsiloxane) mold and the results. The PDMS solvent and curing agent were mixed (10:1 weight ratio)

http://dx.doi.org/10.1016/j.mee.2013.01.027 ⇑Corresponding author. Tel.: +886 32654310.

(2)

and degassed in a vacuum chamber to remove air bubbles from the mixture. After curing at 150 °C for 30 min, the solidiﬁed PDMS mold was peeled away from the Si template.

2.2. Roller imprinting process

Fig. 3 outlines the roller imprinting process. Rather than imprinting the entire surface at one time, the roller and substrate were imprinted progressively. This approach is applicable to the fabrication of structures with a large surface area[15–16]. Polytet-rafluoroethylene (PTFE) was first sprayed onto the PDMS soft mold at 2000 rpm for 30 s and then cured at 200 °C for 10 min to obtain a lower surface energy. Hexamethyldisilazane (HMDS) was then sprayed onto the LED substrate at 6000 rpm for 30 s before being cured at 110 °C for 1 min to increase adhesion. Photoresist (AZ601) was sprayed onto the PDMS mold at room-temperature. The two-inch sapphire wafer was fixed to the translation stage using a vacuum holder in preparation for roller imprinting. During the imprinting process, a UV lamp was used as a curing source

be-cause the optical energy was sufﬁcient to penetrate the quartz cyl-inder and PDMS mold and focus on the contact area.

2.3. Convex-shape PSS wafer preparation

In this study, convex-shaped PSSs were produced by dry etch-ing. A mixture of BCl3, Cl2, and Ar gasses was introduced for ICP

at ﬂow rates of 40 sccm, 10 sccm, and 30 sccm, respectively. The ICP power and the RF power were set at 400 W and 166 W. Follow-ing the etchFollow-ing process, residual photoresist was removed usFollow-ing acetone, isopropyl alcohol, and DI water.

2.4. PSS LED epitaxy and chip process

The GaN-based PSS LED structure in this study was grown using metal organic chemical vapor deposition (MOCVD). The LED layer structure comprised a low-temperature GaN nucleation layer, a thick unintentionally doped GaN layer, a n-type GaN layer, an ac-tive region with 10 periods of InGaN/GaN multiple quantum wells (MQWs), and a p-type GaN layer. Indium tin oxide (ITO) was

evap-Fig. 1. (a) Schematic illustration of Si template; (b) SEM image of Si template.

Fig. 2. (a) Schematic illustration of PDMS molding process; (b) SEM image of PDMS mold.

(3)

out the PSS.

2.5. Light extraction analysis using the ray tracing method

To calculate extraction efﬁciency, we used Monte Carlo ray trac-ing to analyze the light propagation of LED chips with PSSs[17]. The simulation model of an LED chip with a sapphire of 430

l

m, an n-GaN of 4

l

m, an MQW of 460 nm, a p-GaN of 200 nm, and an ITO of 240 nm, is shown inFig. 4. A two dimensional convex-shaped hexagonal close-packed array was arranged across the en-tire sapphire surface, as in the experiment. To reduce calculation time, the chip size was assumed to be 123

l

m 232

l

m. The total emission power from MQW was set at 0.1 W. Mesa etching and electrodes on p- and n-GaN (as on the real chip) were disregarded. Light extraction was calculated according to the fraction of light collected by a spherical detector located 5 cm from the LED. The absorption coefﬁcient of the MQWs was investigated; however, the effects of the other layers were considered inconsequential and therefore disregarded[18].

3. Results and discussion

Fig. 5presents the imprinted results made using a PDMS mold with 2

l

m deep structures. Our results revealed a residual layer (approximately 200 nm in depth) at the bottom of the structures, which decreased the height of the PSS structures following the ICP etching process.

Fig. 4. Diagram of the PSS LED model for ray tracing simulation.

Fig. 5. SEM image of imprinted results.

(4)

The residual layer was eliminated by applying O2plasma

etch-ing for 50 s, as shown inFig. 6.Fig. 6(a) presents hexagonal close-packed structures with long-range order. The heights of the

struc-tures in Figs. 6(b)–(d) are approximately 1.6

l

m, 2.2

l

m, and 2.7

l

m, respectively.

Fig. 7presents SEM images of the convex-shaped PSS after ICP etching. The imprinted cylinders inFig. 6(b)–(d) were used as an etching mask during the ICP etching process (22, 24, and 28 min, respectively). The structures in Fig. 7(a)–(c) are approximately 0.9

l

m, 1

l

m, and 1.4

l

m in height.

Room temperature EL measurements of the InGaN LED with and without PSS as a function of forward current are presented in

Fig. 8(a). Nearly all of the curves overlap, indicating that the elec-trical properties were not adversely affected by the PSS fabrication process. To investigate the inﬂuence of the PSS on the LEDs, the

Fig. 7. SEM images of convex-shape PSSs with a structure height of (a) 0.9lm, (b) 1lm, (c) 1.4lm.

Fig. 8. (a) Current–voltage relationship of LEDs with and without the PSSs; (b) Output intensity of PSS LEDs and conventional LEDs as a function of driving current.

Table 1

Light output power and relative enhancement of various LEDs.

LED types (structure height) Output power Enhancement (%)

Conventional LED 122.03

PSS LED (0.9lm) 153.4 25.7

PSS LED (1lm) 154.29 26.4

PSS LED (1.4lm) 159.15 30.4

Fig. 9. Efﬁciency enhancement of the experimental and simulation results (compared to conventional LEDs) using structures of various heights: (a) PSS LEDs with paraboloidal structure; (b) PSS LEDs with conical structure.

(5)

were constructed for ray tracing analysis: one with paraboloidal structures and the other with conical structures.Fig. 9compares the simulation results with those from the experiment using struc-tures of various heights. Because the absorption coefficient of the MQWs was unknown, we conducted several optical simulations using a range of absorption coefficients. Although the chip size was reduced to shorten simulation time and the structures were simplified representations of the actual PSSs, the trend of enhance-ment was very similar to that observed in the experienhance-ment. These results prove the reliability of optical simulations as analytical instruments prior to experimentation. Our experimental results indicate that the profiles of structure on the fabricated PSS tended toward paraboloidal and conical shapes. Results of simulated paraboloidal structures with an absorption coefficient of 2.5 104/cm were nearly identical to the devices used in the experiments. With structures of the same height and absorption coefficient; the simulated paraboloidal structure presented higher light extraction efficiency.

4. Conclusions

This study fabricated Si templates using photolithography and dry etching, followed by duplication of the patterns in a soft poly-mer PDMS mold. The imprinted material was transferred from the

[1] M.R. Krames, O.B. Shchekin, R. Mueller-Mach, G.O. Mueller, L. Zhou, G. Harbers, M.G. Craford, J. Display Technol. 3 (2007) 160–175.

[2] T. Gessmann, E.F. Schubert, J. Appl. Phys. 95 (2004) 2203–2216. [3] E.F. Schubert, J.K. Kim, Science 308 (2005) 1274–1278.

[4] A. David, T. Fuji, R. Sharma, K. McGroddy, S. Nakaruma, S. DenBarrss, E.L. Hu, C. Weisbuch, H. Benisty, Appl. Phys. Lett. 88 (2006) 133514-1–133514-3. [5] M.R. Krames, M. Ochiai-Holcomb, G.E. Höﬂer, C. Carter-Coman, E.I. Chen, I.H.

Tan, P. Grillot, H.C. Chui, J.W. Huang, S.A. Stockman, F.A. Kish, M.G. Craford, Appl. Phys. Lett. 75 (1999) 2365–2368.

[6] C. Huh, K.S. Lee, E.J. Kang, S.J. Park, Appl. Phys. Lett. 93 (2003) 9383–9385. [7] T. Fujii, Y. Gao, R. Sharma, E.L. Hu, S.P. DenBaars, S. Nakamura, Appl. Phys. Lett.

84 (2004) 855–857.

[8] J.H. Lee, J.T. Oh, J.H. Lee, IEEE Photonics Technol. Lett. 20 (2008) 1563–1565. [9] N. Okada, T. Murata, K. Tadatomo, H.G. Chang, K. Watanabe, Jpn. J. Appl. Phys.

48 (2009) 221031–1221034.

[10] J.B. Kim, S.M. Lim, Y.W. Kim, S.K. Kang, S.R. Jeon, Jpn. J. Appl. Phys. 49 (2010) 0421021–0421024.

[11] K. Tadatomo, H. Okagawa, Y. Ohuchi, Phys. Status Solidi A-Appl. Mat. 188 (2001) 121–125.

[12] D.S. Wuu, W.K. Wang, K.S. Wen, S.C. Huang, S.H. Lin, R.H. Horng, Y.S. Yu, M.H. Pan, J. Electrochem. Soc. 153 (2006) G765–G770.

[13] D.S. Wuu, W.K. Wang, K.S. Wen, S.C. Huang, S.H. Lin, S.Y. Huang, C.F. Lin, R.H. Horng, Appl. Phys. Lett. 89 (2006) 161–163.

[14] J.H. Lee, D.Y. Lee, B.W. Oh, J.H. Lee, IEEE Trans. Electron Devices 57 (2010) 157– 163.

[15] J.J. Lee, S.Y. Park, K.B. Choi, G.H. Kim, Microelectron. Eng. 85 (2008) 861–865. [16] H. Tan, A. Gilbertson, S.Y. Chou, J. Vac. Sci. Technol. B. 16 (1998) 3926–3928. [17] T.X. Lee, K.F. Gao, W.T. Chien, C.C. Sun, Opt. Express 15 (2007) 6670–6676. [18] J.F. Muth, J.D. Brown, M.A.L. Johnson, Z. Yu, R.M. Kolbas, J.W. Cook, J.F.