Defect selective passivation in GaN epitaxial growth and its application to light emitting diodes

Defect selective passivation in GaN epitaxial growth and its application to light

emitting diodes

M.-H. Lo, P.-M. Tu, C.-H. Wang, Y.-J. Cheng, C.-W. Hung, S.-C. Hsu, H.-C. Kuo, H.-W. Zan, S.-C. Wang, C.-Y. Chang, and C.-M. Liu

Citation: Applied Physics Letters 95, 211103 (2009); doi: 10.1063/1.3266859 View online: http://dx.doi.org/10.1063/1.3266859

View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/95/21?ver=pdfcov

Published by the AIP Publishing

Articles you may be interested in

GaInN light-emitting diodes using separate epitaxial growth for the p-type region to attain polarization-inverted electron-blocking layer, reduced electron leakage, and improved hole injection

Appl. Phys. Lett. 103, 201112 (2013); 10.1063/1.4829576

Near milliwatt power AlGaN-based ultraviolet light emitting diodes based on lateral epitaxial overgrowth of AlN on Si(111)

Appl. Phys. Lett. 102, 011106 (2013); 10.1063/1.4773565

A visualization of threading dislocations formation and dynamics in mosaic growth of GaN-based light emitting diode epitaxial layers on (0001) sapphire

Appl. Phys. Lett. 101, 231911 (2012); 10.1063/1.4769905

Defect reduction and efficiency improvement of near-ultraviolet emitters via laterally overgrown GaN on a GaN/patterned sapphire template

Appl. Phys. Lett. 89, 161105 (2006); 10.1063/1.2363148

Influence of residual oxygen impurity in quaternary InAlGaN multiple-quantum-well active layers on emission efficiency of ultraviolet light-emitting diodes on GaN substrates

J. Appl. Phys. 99, 114509 (2006); 10.1063/1.2200749

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 140.113.38.11 On: Wed, 30 Apr 2014 14:45:54

Defect selective passivation in GaN epitaxial growth and its application

to light emitting diodes

M.-H. Lo,1,2P.-M. Tu,1C.-H. Wang,1Y.-J. Cheng,1,2,a兲C.-W. Hung,1S.-C. Hsu,2 H.-C. Kuo,1H.-W. Zan,1S.-C. Wang,1C.-Y. Chang,3and C.-M. Liu4

1

Department of Photonics and Institute of Electro-Optical Engineering, National Chiao Tung University, 1001 Ta Hsueh Rd., Hsinchu 300, Taiwan

2

Research Center for Applied Sciences, Academia Sinica, Taipei 11529, Taiwan

3

Institute of Electronics, National Chiao Tung University, 1001 Ta Hsueh Rd., Hsinchu 300, Taiwan

4

Sino-American Silicon Products Inc., Hsinchu 300, Taiwan

共Received 4 August 2009; accepted 28 October 2009; published online 24 November 2009兲 A defect selective passivation method to block the propagation of threading dislocations in GaN epitaxial growth is demonstrated. The defect selective passivation is done by using defect selective chemical etching to locate defect sites, followed by silicon oxide passivation of the etched pits, and epitaxial over growth. The threading dislocation density in the regrown epilayer is significantly improved from 1⫻109 to 4⫻107 cm−2. The defect passivated epiwafer is used to grow light emitting diode and the output power of the fabricated chip is enhanced by 45% at 20 mA compared to a reference one without using defect passivation. © 2009 American Institute of Physics. 关doi:10.1063/1.3266859兴

GaN based light emitting devices have attracted great attention in last few years due to its importance in solid state lighting applications. Researchers are actively investigating various approaches to improve device performance. The de-vices are often epitaxially grown on foreign substrates, for example sapphire. The as grown GaN epitaxial layer has high threading dislocation共TD兲 density typically in the range of 108–10 _cm−2_{due to the mismatches in lattice constants and} thermal expansion coefficients between GaN and sapphire. These defects are nonradiative recombination centers and are detrimental to optoelectronic device performance. The reduc-tion of TD is of great importance for the development of GaN based devices.

There are several epitaxial growth methods to improve crystal quality. A very commonly used one is the epitaxial lateral overgrowth technique共ELOG兲.1,2Strips of SiO₂mask along specific crystal direction are deposited on GaN episur-face, followed by epitaxial growth. The growth starts from the window regions and grows vertically as well as laterally to cover the SiO₂ strips until obtaining planar surface over whole wafer. The lateral growth above mask area bends the propagation direction of threading dislocation and results in significantly lower defect density. The defect density is how-ever still high at window regions and coalescent boundaries. Another approach is to use patterned sapphire substrate for epitaxial growth,3,4but the reduction in TD defect density is often not as effective as ELOG method. Other methods use

in situ SiN_xor ex situ TiN_xporous insertion layers,5,6where GaN nucleates from the pores of the inserted layer and lateral overgrowth on top of it. Recently, defect reduction methods using defect selective etching followed by metalorganic chemical vapor deposition 共MOCVD兲7 or hydride vapor phase epitaxy8regrowth have also been reported. In this let-ter, we demonstrate a TD reduction method by self-aligned defect selective passivation without the need of photolithog-raphy and use it to fabricate a high efficiency light emitting

diodes 共LED兲. The defect selective passivation is done by defect selective etching, SiO₂ passivation at etch pits, and epitaxial over growth.

The schematic of process flow is shown in Figs. 1共a兲–1共d兲. A 30 nm of low temperature GaN nucleation layer followed by a 2 ␮m GaN buffer layer is grown on 共0001兲 sapphire template by low pressure MOCVD. The GaN wafer is immersed in high temperature molten KOH at 280 ° C for 10 min. The molten KOH selectively etches defect sites and forms hexagonal pits on GaN surface9 as illustrated in Fig. 1共a兲. There can be one or several TDs under an etch pit and the KOH etching may not etch all defects, which will be explained later. A 0.5 um SiO₂ film is deposited on the etched surface by plasma enhanced chemical vapor deposi-tion as shown in Fig. 1共b兲. The SiO₂ thin film on the flat surface area is then removed by chemical mechanical polish-ing, which leaves only SiO₂ at the etched pits as shown in Fig.1共c兲. The SiO2fillings at etch pits act as defect passiva-tion to block the continuous propagapassiva-tion of TDs in the sub-sequent epitaxial growth. The exposed GaN flat surface pro-vides the seed layer for epitaxial regrowth, which grows in

a兲_{Electronic mail: [email protected].}

Sapphire TDs EPs Sapphire TDs EPs

SiO

Sapphire

TDs

EPs

SiO

Sapphire

TDs

EPs

Sapphire

TDs

EPs

Sapphire

TDs

EPs

FIG. 1.共Color online兲 关共a兲–共d兲兴 Defect selective passivation process flow.

APPLIED PHYSICS LETTERS 95, 211103共2009兲

0003-6951/2009/95_{共21兲/211103/3/$25.00} 95, 211103-1 © 2009 American Institute of Physics This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

both vertical and lateral direction to cover over the SiO₂ passivated pits, as shown in Fig.1共d兲. This defect passivation and regrowth process can be viewed as a variation of con-ventional ELOG method but with some unique advantages. First, photolithography patterning is not required. Second, micro SiO₂ masks are self-aligned to the TD pits created by defect selective etching. Third, the size of each mask is ex-actly matched to individual etch pit size. This method, com-pared to ELOG, provides defect selective passivation rather than random blocking of TD defects.

Figures2共a兲–2共c兲are the surface scanning electron mi-croscope 共SEM兲 images corresponding to processing steps Figs.1共a兲–1共c兲. Figure2共a兲shows the random distribution of hexagonal pits after molten KOH etching. The etched crystal planes are mostly 兵11¯01其 facets. Aside from individual hex-agonal pits, there are often several pits clustered together. The etch pit density counting all the individual pits is about 5⫻108 cm−2. The deposited SiO2 thin film follows the etched pit topography as shown in Fig.2共b兲. After chemical mechanical polishing 共CMP兲 process, only SiO2 in etched pits are left as shown in Fig. 2共c兲. Small pits are filled with SiO₂. For large pits, SiO₂only fills the side walls leaving a void in the center. The subsequent MOCVD epitaxial over growth covers the whole wafer with flat surface. A LED structure with 2 ␮m Si-doped n-GaN, ten pairs of InGaN/ GaN multiple quantum wells共QWs兲, and a 30 nm Mg-doped p-GaN were grown on the template. The QW emission wave-length is at 425 nm.

To assess the TD reduction, a tunneling electron micro-scope共TEM兲 image was taken as shown in Fig.3共a兲. The TD density can be estimated by directly counting the TD lines in the plane-view micrograph. The TD density estimated at the dashed line right below defect passivation layer is about 1 ⫻109 _cm−2_{. This number is slightly larger than the KOH} etch pit density 5⫻108 _cm−2_{. The discrepancy is due to the} fact that there can be multiple TDs under an etch pit as can be seen in Fig.3共a兲. The TD density is significantly reduced to 4⫻107 cm−2at the dashed line near the QW region. SiO2 passivations at etched pits do effectively block the propaga-tion of TDs. The SEM image however also shows that SiO₂ fillings do not occur on top of all TDs. In other words, no all

the TDs are etched by molten KOH. As a result, some TDs are not blocked and propagate all the way to the top surface. Those TDs not etched by KOH are presumably specific types of TDs resistive to molten KOH etching. It has been reported that KOH preferentially etches screw type TD and is less effective in etching edge and mixed type TDs.10To verify if those TDs propagating all the way to the top surface are resistive to KOH etching, the regrown sample was put in molten KOH under the same etching conditions again. The SEM image of the etched sample is shown in Fig. 3共b兲. There are much less etch pits, with density of only 1 ⫻105 _cm−2_{. The great reduction in etch pit density from} 5⫻108 to 1⫻105 cm−2 shows that the TDs that can be etched by KOH are mostly blocked by the SiO2 fillings in the defect selective passivation process. The etch pit density

1⫻105 cm−2 is much smaller than the TD density 4

⫻107 _cm−2 _{estimated from the cross section TEM image} near surface region. Most of TDs not blocked by defect pas-sivation process are those resistive to KOH etching. The sample was subsequently immersed in 195 ° C H₃PO₄ solu-tion for 20 min. It has been reported that KOH and H3PO4 have different preferential crystallographic etch planes.11The surface SEM image is shown in Fig. 3共c兲. The H₃PO₄ etch pit morphology is very different from that of KOH etching. The etch pit density is about 1.2⫻107 _cm−2_{. This number is} close to the TD density estimated from the TEM image near surface region. This implies that KOH and H3PO4somehow etches different defect types of TDs and both should prob-ably be used to etch TDs in the defect selective etching pro-cess to increase defect passivation coverage.

The optical characteristic is investigated by cathodlumi-nescent 共CL兲 and SEM cross section images as shown in Figs. 4共a兲 and4共b兲. These two images are taken by simply switching detection mode from cathodoluminescent detec-tion to scattering electron detecdetec-tion under the same magnifi-cation condition and thus have one to one lomagnifi-cation corre-spondence. The CL intensity changes dramatically across SiO2 passivation boundary. The bright CL spots are mostly located at the regrown GaN right on top of SiO2passivation masks. The intensity of bright CL spot is so high that when it

1 μm

1 μm

FIG. 2. 关共a兲–共c兲兴 SEM image of the sample surface corresponding to the process step in Figs.1共a兲–1共c兲, respectively.

MQW n-GaN

GaN buffer layer SiO2

MQW

n-GaN

GaN buffer layer

SiO

4x10

_/cm

1x10

_/cm

SiO

GaN buffer layer SiO2

MQW

n-GaN

GaN buffer layer

SiO

4x10

_/cm

1x10

_/cm

SiO

FIG. 3.共a兲 TEM image of LED sample with defect selective passivation. 共b兲 SEM image of the defect passivated sample after molten KOH etching.共c兲 SEM image of the defect passivated sample after high temperature H3PO4 etching.

211103-2 Lo et al. Appl. Phys. Lett. 95, 211103共2009兲

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 140.113.38.11 On: Wed, 30 Apr 2014 14:45:54

is adjusted not to saturate detector, the CL intensity differ-ence between lower GaN layer and sapphire become visually hard to distinguish. By analyzing the gray scale levels of image pixels, we can still identify GaN-sapphire interface, which is also confirmed by comparing Figs. 4共a兲and 4共b兲. The CL intensity averaged over the cross section area of the regrown layer is measured to be about 23 times of that of the layer below passivation boundary. The threading dislocation defects are strong nonradiative recombination centers.12The significant increase in CL intensity demonstrates that the loss of excited carriers due to nonradiative recombination is greatly reduced in the defect passivated layer as a result of the reduction in TD density.

LED chips with size of 300⫻300 ␮m2 _{were fabricated} from the defect passivated epiwafer. The optical and electri-cal characteristics are compared with a reference LED going through the same fabrication process except for the defect passivation layer. The light current共L-I兲 and voltage current 共V-I兲 curves of the defect passivated LED 共DP-LED兲 and reference LED 共R-LED兲 are shown in Fig.4共c兲. The driving voltage of DP-LED共red solid line兲 is slightly lower than that 共black dotted line兲 of R-LED. We remark that this slightly lower driving voltage may be due to the lower defect density. When the defect density is lower, electrons are easier to move across material without scattering thus leading to lower resistance and driving voltage. The optical output power is however significantly enhanced by 45% compared to that of the reference sample at 20 mA. The SiO₂masks not only block the propagation of TD but also can act as light scattering sites to improve LED light extraction efficiency, similar to the use of patterned GaN-sapphire interface3,13 to reduce light trapped by total internal reflection. A Monte Carlo ray tracing simulation is used to check if light extrac-tion enhancement is a significant part of total output power

enhancement. A simplified two-dimensional array of inverted hexagonal pyramid SiO2 masks is used to model the defect passivation. Various geometries with 0.7– 1 ␮m lateral size, 0.5– 1 ␮m height, and 3 – 4 ␮m center to center spacing are calculated and show a variation of light extraction ment from 10%–25%. It shows that light extraction enhance-ment cannot be totally neglected in the total 45% output power enhancement. The output power enhancement from TD defect reduction is not as large as the above measured CL intensity enhancement. It has been reported that InGaN quantum well emission is less sensitive to TD defects. LEDs using GaN or AlGaN active layer is however very sensitive to TD defects and really requires the reduction of TD.14,15

In summary, we have demonstrated a defect selective passivation method to reduce TD density and used it to fab-ricate a LED. The defect passivation SiO2 masks are self-aligned to the TD defect pits created by KOH defect selective etching without photolithography patterning and can significantly reduce TD density from 1⫻109 to 4 ⫻107 _cm−2_{. The SiO}

2 masks also improve the light extrac-tion efficiency in LED applicaextrac-tion. TEM image shows that some defects are resistive to KOH etching and propagate all the way to the top surface. Further improvement can be made by exploring additional etching chemicals that are comple-mentary to KOH in defect selective etching to increase the coverage rate of defect selective passivation.

This work was financially supported by Sinica Nano-program and in part by the National Science Council of Re-public of China共ROC兲 Taiwan under Contract NSC97-2112-M-001-027-MY3.

1_{T. Mukai, K. Takekawa, and S. Nakamura,Jpn. J. Appl. Phys., Part 2 37,} L839共1998兲.

2_{O.-H. Nam, M. D. Bremser, T. S. Zheleva, and R. F. Davis,Appl. Phys.}

Lett. 71, 2638共1997兲.

3_{E.-H. Park, J. Jang, S. Gupta, I. Ferguson, C.-H. Kim, S.-K. Jeon, and J.-S.} Park,Appl. Phys. Lett. 93, 191103共2008兲.

4_{Y. J. Lee, H. C. Kuo, T. C. Lu, B. J. Su, and S. C. Wang,J. Electrochem.}

Soc. 153, G1106共2006兲.

5_{J. Xie, Ü. Özgür, Y. Fu, X. Ni, H. Morkoç, C. K. Inoki, T. S. Kuan, J. V.} Foreman, and H. O. Everitt,Appl. Phys. Lett. 90, 041107共2007兲.

6_{Ü. Özgür, Y. Fu, Y. T. Moon, F. Yun, H. Morkoç, H. O. Everitt, S. S. Park,} and K. Y. Lee,Appl. Phys. Lett. 86, 232106共2005兲.

7_{J. W. Lee, C. Sone, Y. Park, S.-N. Lee, J.-H. Ryou, R. D. Dupuis, C.-H.} Hong, and H. Kim,Appl. Phys. Lett. 95, 011108共2009兲.

8_{J. L. Weyher, H. Ashraf, and P. R. Hageman,Appl. Phys. Lett.95, 031913} 共2009兲.

9_{T. Kozawa, T. Kachi, T. Ohwaki, Y. Taga, N. Koide, and M. Koike,J.}

Electrochem. Soc. 143, L17共1996兲.

10_{L. Min, C. Xin, F. Hui-Zhi, Y. Zhi-Jian, Y. Hua, R. Q. Li Zi-Lan, Z.} Guo-Yi, and Z. Bei,Phys. Status Solidi C 1, 2438共2004兲.

11_{D. A. Stocker, E. F. Schubert, and J. M. Redwing,Appl. Phys. Lett. 73,} 2654共1998兲.

12_{M. Albrecht, J. L. Weyher, B. Lucznik, I. Grzegory, and S. Porowski,}

Appl. Phys. Lett. 92, 231909共2008兲.

13_{K. Tadatomo, H. Okagawa, Y. Ohuchi, T. Tsunekawa, Y. Imada, M. Kato,} and T. Taguchi,Jpn. J. Appl. Phys., Part 2 40, L583共2001兲.

14_{T. Mukai, D. Morita, and S. Nakamura,J. Cryst. Growth189, 778共1998兲.} 15_{S. Nakamura,Science 281, 956共1998兲.} Defect passivation layer 1μm GaN-sapphire_interface Defect passivation layer 1μm Defect passivation layer 1μm GaN-sapphire_interface

SiO2Defect passivation

GaN-sapphire interface 1μm

SiO2Defect passivation

GaN-sapphire interface 1μm

(a) (b)

FIG. 4.共Color online兲 关共a兲 and 共b兲兴 Cross section CL and SEM image of the defect passivated epi-wafer under same magnification.共c兲 L-I and V-I curve of the DP-LED and R-LED.

211103-3 Lo et al. Appl. Phys. Lett. 95, 211103共2009兲

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 140.113.38.11 On: Wed, 30 Apr 2014 14:45:54