Study on the doping technologies of wide bandgap semiconductors for the development of light emitting devices

Study on the doping technologies of wide bandgap semiconductors

for the development of light emitting devices 

ワイドバンドギャップ半導体における発光素子開発のための  ドーピング技術に関する研究 

June 2007

2007 年 6 月

Abstract

The aim of the present study was to improve performance of visible–UV light emitting devices by optimization and characterization of doping techniques for new wide bandgap semiconductors.

For red-green region light emitting devices, InGaAlP quaternary alloys were used to construct double hetero (DH) structure light emitting devices. N-type, p-type and residual doping characteristics and electrical properties were investigated systematically. Effects of substrate misorientation on dopant incorporation and passivation were explored. The results show that by use of off-angle substrates a significant improvement in performance of InGaAlP red-green light emitting diodes (LEDs) and red-yellow laser diodes (LDs) could be achieved, such as enhancement of p-type doping and light emission efficiency. The performances of a green LED and a yellow LD with InAlP cladding layers grown on off-angle substrates was investigated and demonstrated.

For blue-violet region light emitting devices, doping and electrical characteristics were investigated using GaN based materials. The origin of low carrier concentrations in p-type AlGaN was studied. The data show that the hole thermal activation energy increases with increasing Al mole fraction in Mg-doped AlGaN. N2 ambient growth was found to dope p-type GaN and AlGaN without post-growth thermal annealing.

Performances of blue-violet region LDs with cleavage mirror facets or inner stripe structures have been demonstrated.

For UV light emitting devices from diamond, detailed electronic characterization of p- and n-type doped films have been performed. Donor and acceptor impurities showed suitable features for application in semiconductor devices, as boron (p-type) and phosphorus (n-type) do not generate additional compensating defects for optimized diamond growth. Excellent interfaces properties between metals and diamond layers were obtained.

The presented data and results show that significant improvements of light emitting devices from wide bandgap semiconductors could be achieved by application of structural and electronic characterizations in combination with optimized growth

Acknowledgements

Any time one undertakes a work of this magnitude, the person cannot do it alone.

I am indebted to a great number of individuals who have helped me along the way, and I will now do my best to express my gratitude.

I would like to start by thanking Prof. Hiroshi Kawarada of Waseda University for giving me the chance to gain a PhD and for fruitful suggestions and discussions. I share the same sentiments for Prof. Iwao Odomari, Prof. Keiji Horikoshi of Waseda University and Dr. Christoph E. Nebel of AIST.

I would sincerely like to thank the late Prof. Yasutada Uemura Science University.

He has played role in my graduate studies and provided me the foundation and interest for semiconductor physics. I also would like to thank Prof. Nobuo Tsuda of Science University who was a teacher of my master degree’s studies. I respect his attitude towards experimental research on materials science very much, and it has greatly affected my stance on research work.

The present study has been carried out in Toshiba Corporate Research Center partially supported by METI through NEDO and collaborations with several research organizations.

I would like to thank Dr. Shuichi Uchikoga, who is an ex-leader of our laboratory, and Mr. Keiji Takaoka, who is a present leader of my laboratory, for their continuous encouragement.

I gratefully acknowledge Dr. Gen-ichi Hatakoshi, Mr. Masayuki Ishikawa, Dr.

Yasuhiro Kokubun, Mr. Masayuki Okajima, Dr. Masao Mashita and Dr. Kazuhiko Itaya, who were my supervisors and also my co-workers. They have given a lot of useful knowledge and advice, fruitful discussions and encouragement. I can never forget to have discussed about my work with Mr. Ishikawa until just before his leaving to USA for his overseas education. I would also like to thank Dr. Chiharu Nozaki, Dr. Hideto Sugawara, Dr. Nishikawa, Mr. Hironori Ishikawa and Dr. John Rennie for their helpful works and discussions on MOCVD growth, evaluation of materials, and device

Dr. Shinya Nunoue, Mr. Masaaki Onomura, Mr. Yoshiharu Takada, Dr. Mizunori Ezaki, Mr. Kunio Tsuda for their help and discussions on device processing and fabrications.

Particularly, I thank to Mr. Onomura for his transferring of his skill for electrode processing and construction of electrical measurements.

I would also thank to my colleagues in my diamond research, Mr. Tadashi Sakai, Mr.

Tomio Ono, Mr. Naoshi Sakuma, Mr. Hiroaki Yoshida, and Dr. Masayuki Katagiri for their help and useful discussions on diamond work. Special thanks to Mr. Sakai, who has endured patiently my unique and tardy works, continuously encouraged, and given me lots of chances to go out to overseas to join a conference. I would also thank to Dr.

Toshihide Izumiya and Dr. Katsuaki Natori for their useful discussions and their help for my work and life during working on the present study. I would like to express my gratitude to Ms. Ayako Hasegawa for her continuous help and encouragement.

I would really like to thank Dr. Satoshi Koizumi of NIMS for his beautiful samples offer, his help in various situations and lots of fruitful discussions. I would also like to be grateful Dr. Yasuo Koide and Dr. Hisao Kanda of NIMS for their fruitful discussions and continuous encouragement. I am grateful to Dr. Hideyo Okushi of AIST who offered me a chance to start C-V measurements on B-doped diamond, and Dr. Masahiko Ogura of AIST for the growth of B-doped diamond and his help for some electrical measurements in AIST. I would like to thank Dr. Sun-Gi Ri of AIST for his help concerning electrical measurements in AIST. I would like to express my appreciation to Dr. Satoshi Yamazaki, Dr. Daisuke Takeuchi, Dr. Takahiro Yamada, and Dr.

Hiromitsu Kato of AIST for their help and useful discussions on diamond. I would like to thank Prof. Milos Nesladek of CEA for their fruitful discussions and continuous encouragement.

I would like to give thanks not only to my colleagues at work. I extend appreciation to Ms. Junko Kamimura, Ms. Atsuko Tetsuka, Ms. Mayumi Okumura, Ms.

Akiko Ishizaki, Ms. Naomi Aoki, Mr. Nobutoshi Amakatsu, Mr. and Ms. Kimura (Saburo and Eiko) and Mr. Yasuo Morie for their continuous encouragement. I thank to Jiro and Fuko, who have given me a lot of energy to write this dissertation.

I feel grateful to my uncle and aunt Tatsujiro and Miyako Suzuki, and my cousins Hiroshi and Rimi Suzuki, Mikiko Suzuki and Fumiko Suzuki as well as my late parents Takako and Katsusaburo Suzuki, for their support and continuous encouragement.

Finally, I thank most sincerely to my brother Toshitaka Suzuki, and my late grand mother Take Sato, who have supported me extensively.

謝辞  

本研究に関して博士学位取得の機会を与えて下さいまして、論文作成にあたり一方なら ぬ御指導と御鞭撻を賜りました早稲田大学理工学部  川原田洋教授に謹んで感謝の意を表 します。同様に、本論文作成にあたって有益な御助言をいただきました早稲田大学理工学 部  大泊巌教授、堀越佳治教授、産総研  Christoph E. Nebel博士に深く感謝いたします。

東京理科大学理学部並び理学研究科在学中に大変お世話になり、研究を進めていくため の基礎的な知識、実験の進め方、考え方を指導してくださった、故・植村泰忠教授、津田 惟雄教授に深く感謝いたします。特に、津田教授の研究に対する姿勢は、現在の研究生活 において目標となっております。

本研究は、株式会社東芝研究開発センターにて行われたもので、論文にまとめるにあた り御助力いただきました、我々のラボラトリーの前リーダー  内古閑修一博士、現リーダ ー  高岡圭児氏に心より厚く御礼申し上げます。

研究を進めるにあたり一方ならぬ御指導をいただき、また共同研究者として有意義な御 助言をいただきました、波多腰玄一博士、石川正行氏、国分義弘博士（現石巻専修大学教 授）、岡島正季氏、真下正夫博士（現弘前大学教授）、板谷和彦博士に感謝いたします。同 じく共同研究者として特に第2章のInGaAlP系材料の開発にあたって、    有意義な御助 言ならびに実験に協力いただきました野崎千晴博士、菅原秀人博士、西川幸江博士、石川 博規氏、John Rennie博士に心より感謝いたします。

GaN の研究に関しまして特に結晶成長、評価において御助言、御助力いただきました西 尾譲司博士、杉浦理沙博士、齋藤真司氏、金子桂氏に深く感謝します。また、布上真也博 士、小野村正明氏、高田賢治氏、江崎瑞仙博士、津田邦夫氏にはGaNからダイヤモンドの 開発にわたり、特に電極プロセスに関して一方ならぬご助言、御協力をいただきましたこ とに心より感謝いたします。

ダイヤモンドの研究におきまして、共同研究者としてご助言、御助力いただきました酒 井忠司氏、小野富男氏、佐久間尚志氏、吉田博昭氏、片桐雅之博士に深く感謝いたします。

特に酒井忠司氏には、海外の会議での発表等、研究者として非常に貴重な機会を何度も与 えていただき、また、ダイヤモンド半導体の研究を続けさせていただいたことに心より感 謝いたします。

だだきました。物材機構  小泉聡博士には高水準の n 型ダイヤモンド試料を提供していた くと共に、研究に関しまして御指導・御助力いただき心より感謝いたします。物材機構の 小出康夫博士には一方ならぬ御助言をいただき、研究を続けていただくにあたり、大いな る励みとなりましたことに感謝いたします。物材機構との共同研究をするにあたりまして、

大変お世話になりました神田久生博士に心より感謝いたします。

また、仏CEAのMilos Nesladek 博士にも御助言、御助力いただきましたことに感謝い

たします。

産総研  大串秀世博士にはダイヤモンドにおいて C-V測定およびショットキー電極に関 する実験を始めるきっかけを作っていただいたこと、また有意義な御助言をいただきまし たことに心より感謝いたします。また、同じく産総研  小倉正彦博士、李成奇博士にはダ イヤモンドの電気的評価について御助力いただき、また、有意義な討論をしていただきま したことについて感謝いたします。産総研  山崎聡博士、竹内大輔博士、山田貴壽博士、

加藤宙光博士にはダイヤモンドに関しまして有意義なご助言をいただきましたことを感謝 いたします。

良き友人としてあるいは先輩として励まし続けていただいた上村順子氏、手塚温子氏、

金野まゆみ氏、石崎聡子氏、青木直美氏、天勝延寿氏、木村三郎氏・栄子氏夫妻、森江康 雄氏、金田和博氏に深く感謝いたします。

亡き父母に感謝すると共に、父母亡き後、一方ならぬお世話になりました伯父および伯 母  鈴木辰次郎氏・美弥子氏夫妻、従妹兄弟の鈴木弘志氏・理美氏夫妻、鈴木美紀子氏、

鈴木芙美子氏に厚くお礼申し上げます。

最後に、継続的に経済的あるいは生活全般においてご助力いただきました、兄  鈴木俊 孝氏ならびに亡き祖母  佐藤たけ氏に心より感謝いたします。

Table of Contents

Chapter 1. General introduction

1.1 Wide bandgap semiconductors for light emitting devices 1.2 Importance of doping technologies in light emitting devices 1.3 Topical overview of the present study

Chapter 2. Doping technologies for InGaAlP light emitting devices 2.1 Introduction to InGaAlP alloys

2.2 MOCVD growth of InGaAlP alloys 2.3 Background of p-type and n-type doing in InGaAlP alloys

2.4 Si doping characteristics in InGaAlP alloys 2.4.1 DX centers in III-V semiconductor alloys

2.4.2 Experiments

2.4.3 Doping characteristics 2.4.4 Deep levels

2.4.5 Shallow and deep donor states

2.4.6 Electrical properties of Si-doped InGaAlP 2.4.7 Summary of Section 2.4

2.5 Residual impurities in InGaAlP alloys

2.5.1 Residual impurities in III-V semiconductors

2.5.2 Experiments

2.5.3 Residual hydrogen doping in InGaAlP alloy system 2.5.4 Residual oxygen doping in InGaAlP

2.5.5 Summary of Section 2.5

2.6 Effects of substrate misorientation on Zn and Si doping characteristics in InGaAlP alloys

2.6.1 Off-angle substrate

2.6.2 Experiments

2.6.3 Effects of substrate misorientation on Zn doping characteristics in Zn-doped In

(Ga

Al

)

P

15 24 27 27 27 28 30 35 38 38 38 42 42 37 42 44 48 56 60

60

65

65

1

4

8

2.7 Effects of substrate misorientation on reduction of residual oxygen doping and improvement of luminescence efficiency in undoped In

(Ga

Al

)

P

2.7.1 Influence of residual oxygen on luminescence efficiency 2.7.2 Experiments

2.7.2 Deep levels

2.7.3 Oxygen concentration

2.7.4 Effects of substrate misorientation on light emission 2.7.5 Summary of Section 2.7

2.8 Light emitting devices using InAlP cladding layers 2.9 Summary of Chapter 2

References

Chapter 3. Doping technologies for GaN based light emitting devices 3.1 Introduction to GaN based materials

3.2 MOCVD growth of GaN based materials

3.3 Doping issues for development of GaN based lasers 3.3.1 Carrier overflow in GaN based lasers

3.3.2 Background of n-type and p-type doping in GaN and related materials

3.4 Residual doping level in undoped GaN, AlGaN and InGaN 3.4.1 Experiments

3.4.2 Residual donor concentration 3.4.3 Summary of Section 3.3

3.5 N-type doping in GaN, AlGaN and InGaN

3.5.1 Experiments

3.5.2 Si doping characteristics in Si-doped GaN 3.5.3 Si doping characteristics in Si-doped AlGaN 3.5.4 Si doping characteristics in Si-doped InGaN 3.5.5 Summary of Section 3.4

3.6 P-type doping in GaN, AlGaN and InGaN

3.6.1 Experiments

81

81 82 82 89 89 93 93 96

106 99

106 110

113 113 115

116

116

116

117

117

117

118

118

122

122

128

128

3.6.2 Mg doping characteristics in Mg-doped GaN 3.6.3 Mg doping characteristics in Mg-doped AlGaN 3.6.4 Mg doping characteristics in Mg-doped InGaN 3.6.5 Summary of Section 3.5

3.7 Effects of N

-ambient growth on p-type conduction in Mg-doped GaN, and AlGaN

3.7.1 Electrical conduction in as grown Mg-doped GaN (Introduction)

3.7.2 Experiments

3.7.3 Comparison of growth characteristics in H

-rich and N

-rich ambient in MOCVD

3.7.4 P-type conduction in as-grown Mg-doped GaN

3.7.5 Comparison of Mg doping characteristics in Mg-doped GaN grown by H

-rich and N

-rich ambient MOCVD

~ [Cp

Mg]/[III] ratio dependence

3.7.6 Comparison of Mg doping characteristics in Mg-doped GaN grown by H

-rich and N

-rich ambient MOCVD

~ Growth temperature dependence

3.7.7 As-grown p-type AlGaN grown by N

-ambient MOCVD 3.7.8 Summary of Section 3.6

3.8 GaN-based laser diodes

3.8.1 InGaN-based multi-quantum-well laser diodes with cleaved facets on conventional C-face sapphire substrates 3.8.2 InGaN inner stripe laser diodes

3.8.3 Summary of Section 3.7

3.9 Summary of Chapter 3 References

Chapter 4. Doping studies for diamond light emitting devices 4.1 Introduction to diamond

137 141 145 145 145 145

146 146

148 151

155

155 158 160 160

165 167

170 172

175

p-type diamond

4.4.2 Background of doping and electrical characteristics for n-type diamond

4.5 Electrical properties of B-related acceptor in B-doped p-type homoepitaxial diamond and metal/p-type diamond interfaces 4.5.1 Experiments

4.5.2 Schottky junction characteristics (I–V characteristics) 4.5.3 C-V characteristics

4.5.4 Electrical characteristics of B acceptor 4.5.5 Summary of Section 4.5

4.6 Electrical properties of P-related donor in P-doped n-type homoepitaxial diamond and metal/n-type diamond interfaces

4.6.1 Experiments

4.6.2 Schottky diode characteristics 4.6.3 P-related donor characteristics

4.6.4 Electrical properties of metal/n-type diamond interfaces 4.6.4.1 Schottky barrier height dependence on metal work Function

4.6.4.2 Comparison between n-type and p-type diamond 4.6.5 Summary of Section 4.6

4.7 Summary of Chapter 4 References

Chapter 5. Conclusion

Publications for the present study

International Conference publications for the present study

183 184

184 186 186 191 196 198

198 200 200 207 207

211 215

216 217

223

225

228

Chapter 1 General Introduction

1.1 Wide bandgap semiconductors for light emitting devices

Unique material properties of wide bandgap semiconductors make them extremely promising for high-power, high-temperature, high-frequency applications in electronic and optoelectronic devices. Particularly, due to their large bandgap energy these semiconductors offer the remarkable ability to realize short wavelength light emitting devices. Generally speaking, a wide band gap semiconductor is a semiconductor with an energy bandgap wider than about 2 eV, which results in a short wavelength light emission.

Figure 1 shows the variation of bandgap energy as a function of the lattice constant for most common semiconductors with the corresponding wavelengths, and highlights some materials investigated in the present study [1]. In general, lattice constants are smaller and bandgaps larger for light elements because of their high bond strength.

The first visible spectrum of light emitting diodes (LEDs) has been reported by Holonyak and Bevacqua in 1962 [2]. They have reported on the emission of coherent visible light from gallium-arsenic-phosphide (GaAsP) junctions. In 1963 and 1964, the first reports of gallium-phosphide (GaP) p-n junction LED have been reported by Allen et al. [3] and Grimmeiss et al. [4], respectively. Green LEDs were formed with efficiencies as high as 0.6% by doping GaP with nitrogen (N) isoelectronic impurities [5].

Visible light laser diodes (LDs) were first developed using InGaAlP quaternary alloy system [6-8]. The InGaAlP alloy has direct bandgap energies ranging from 1.9 to 2.3 eV, which covers the red to green portion of the visible spectrum, when grown lattice-matched on gallium arsenide (GaAs) substrates. This alloy system is still the dominant material system for high-brightness light emitting devices in the red, orange and yellow wavelength range today due to the high external quantum efficiency as well as that of GaN based materials in the blue – blue violet wavelength region, as shown in Fig. 1.2 [9]. Materials based on zinc selenide (ZnSe) grown on a GaAs substrate have been studied intensively for use in blue-green light emitting devices. However the lifetime of these laser diodes is only about 100 hours, a limitation that has prevented

Figure 1.1 Variation of bandgap energy as a function of the lattice

constant for most common semiconductors with corresponding

wavelengths.

Figure 1.2 External quantum efficiency as a function of peak

wavelength of LEDs. (Lumilleds Lighting LLC, ICNS-5, 2003)

Indium-gallium-nitride (InGaN) based nitride materials have a direct bandgap that is suitable for blue light emitting devices. Although these materials had a serious

problem due to the lack of p-type doping, the breakthrough came in 1989, when H.

Amano et al. obtained thin films of p-type GaN for the first time [12]. Furthermore, Nakamura et al. obtained low resistive p-type GaN by post-growth annealing in 1992 [13]. After that, blue LDs and LEDs have been rapidly developed with these materials [14, 15], although green or UV region LEDs still show considerably low external quantum efficiency, as shown in Figs.1.2 and 1.3.

With the growing demand for III-nitride-based light-emitting diodes (LEDs), zinc oxide (ZnO) has been attracting renewed attention due to its superior properties for fabricating ultraviolet laser diodes and LEDs. The impact of their performance is expected to span from low-threshold operation of excitonic lasing to low-cost and clean productivity of UV LEDs, taking advantage of the abundant supply and commercial availability of single-crystalline substrates [16].

Diamond has also attracted attention as a material for deep UV light emitting devices because of the high density of excitons even at high temperature. 10 years ago, successful doping of n-type [17] has led to achieve in the diamond p-n junction UV light emitting diodes [18, 19]. As shown in Fig. 1.3, it is still difficult to obtain high external quantum efficiency in UV region LEDs. Diamond is one of the promising materials for deep UV region LED.

1.2 Importance of doping technologies for light emitting devices

In general, high brightness characteristics are obtained by introducing a double heterostructure (DH). Figure 1.4 shows a schematic primitive DH LED structure.

DH LEDs offer inherent advantages, including the following: enhanced injection efficiency of electrons or holes, confinement of injected carriers in the active layer, and formation of transparent window and substrate layers for improved current spreading and light extraction. Doping of cladding layers has a strong influence on the efficiency of DH LEDs. The resistivity of cladding layers is one factor in determining the doping concentration in the cladding layers. The resistivity should be low to avoid resistive heating and to enhance the current spreading in the cladding layers. Another factor is the residual doping concentration in the active (light emitting) layer. The active layer has a residual doping concentration, even if not intentionally doped. The residual impurities are possible to give deterioration effects on light emitting as non-radiative center. And another thing, the doping concentration in the cladding layers must be

Figure 1.3 Reported values of external quantum efficiency as a function of peak wavelength of LEDs.

External Quantum Efficiency (%)

Peak Wavelength of LED (nm)

200 300 400

10

10

10

10

10

10

Diamond

InGaAlN & Diamond

Figure 1.4 Schematic structure of a primitive double heterostructure

LED.

Figure 1.5 Schematic band diagram on an InGaAlP double 

heterostructure.

higher than the doping concentration of the active layer to define the location of the p-n junction. As shown in Fig. 1.5, low doping concentrations in the cladding layers also facilitate electron or hole escape from the active layer, thereby lowering the internal quantum efficiency in LEDs and the increase in the threshold current in LDs [20].

Therefore, the high concentration doping in the cladding region and low concentration of residual doping in the active region are essential to realize high performance of light emitting devices. However, it has been expected to be difficult to obtain high concentration doping in p-type or n-type cladding region in wide bandgap semiconductors due to the large bandgap energy and high strength in chemical bonding between atoms [21]. Actually, it has been difficult to obtain low resistive p-type layers for III-V materials and II-VI materials, and low resistive n-type layers for diamond as described in the next section.

1.3 Difficulties in doping control for wide bandgap semiconductors

It has long been recognized that doping of wide bandgap semiconductors often poses problems: typically, one (electrical conduction) type of doping is hard to achieve. High p-type or n-type doping is limited mainly by two key factors: low (dopant) impurity concentration, and low ionization ratio (low electrical activation) of the incorporated dopant impurity. The former factor includes low impurity incorporation efficiency and low solubility of the impurity. Lowering of impurity incorporation efficiency is often observed during high temperature growth due to the high pressure of the dopant material. In case of wide bandgap semiconductors, the small lattice parameters have a tendency to result interference with incorporation of large dopant impurities.

The latter factor includes: large ionization energy of the dopant and compensation by defects (such as residual impurities, native defects and alternative atomic configurations). Figure 1.6 shows the reported ionization energy of acceptor or donor in several semiconductors as a function of bandgap energy. As shown in this figure, wide bandgap semiconductors have a large ionization energy value of the dopant. The large ionization energy should give low carrier concentration in ordinary temperature.

Most common compensation defects are formed by hydrogen atoms (hydrogen passivation). Hydrogen is unintentionally introduced during several growth techniques, such as metalorganic chemical vapor deposition (MOCVD), gas source molecular beam epitaxy (GSMBE), and plasma- enhanced CVD. The introduced hydrogen makes complexes with the dopants and passivates them. Figure 1.7 shows the microscopic model to explain hydrogen passivation of the four types of donors and acceptors in GaAs

[22]. Oxygen contamination is also severe problem for Al containing semiconductors. It has been reported that the presence of oxygen can also result in dopant compensation or the creation of efficient non-radiative recombination centers in AlGaAs [23].

In the present study, doping characteristics have been systematically investigated for InGaAlP alloys, GaN based materials and diamond to obtain high performance in visible-UV region light emitting devices. These materials have a common issue that either conduction type is difficult to achieve. InGaAlP alloy system and GaN based materials have difficulty in p-type doping, and diamond has difficulty in n-type doping.

We have revealed the influence of residual impurities on doping properties and device characteristics. High device performances are realized by using the developed doping technologies.

Ionization Energy of Donor or Acceptor (eV)

Bandgap Energy (eV)

Diamond

GaN ZnSe

InGaAlP Si GaAs

Donor Acceptor

1 2 3 4 5 6

0.0

0.5

(B)

(B) (P)

(Zn) (N)

(Li)

(Mg)

(Si) (Si) (Si) (Zn)

Figure 1.6 Reported ionization energies of acceptors and donors in

several semiconductors.

Figure 1.7 Microscopic models of acceptor and donor passivation by

hydrogen in GaAs. (J. Vetterhoffer et al., Phys. Rev.B 53, 12835 (1996).)

1.4 Topical overview of the present study

The following chapters in this dissertation are organized as follows: Chapter 2 provides the necessary background information for the study of InGaAlP alloys, and presents doping characteristics and electrical characteristics of zinc (Zn)-doped p-type InGaAlP and silicon (Si)-doped InGaAlP. Achievements in high concentration Zn doping of the p-type cladding layer and low concentration residual oxygen doping of the active layer by use of off-angle substrates are shown. Improved device performances with better doping techniques are shown. Chapter 3 provides the necessary background information for the study of GaN based materials, and the doping characteristics of Mg-doped p-type layers and Si-doped n-type layers. The Mg acceptor level is shown for AlGaN as a function of Al concentration. Characteristics of blue-violet laser diodes are shown based on these materials are demonstrated. Chapter 4 presents doping and electrical characteristics of boron (B)-doped p-type diamond and phosphorus (P)-doped n-type diamond. Furthermore, electrical characteristics of metal/diamond interfaces are shown. The dissertation concludes with Chapter 5, which summarizes the presented results.

References

[1] G. Hatakoshi, “Short-wavelength semiconductor lasers for optical memory” Rev.

Laser Eng., 25, 557 (1997).

[2] N. Holonyak Jr. and S. F. Bevacqua, “Coherent (visible) light emission from Ga(As1-xPx) junctions” Appl. Phys. Lett. 1, 82 (1962).

[3] J. W. Allen, M. E. Moncaster and J. Starkiewicz, “Electroluminescent devices using carrier injection in gallium phosphide” Solid State Electronics 6, 95(1963).

[4] H. G. Grimmeiss and H. J. Scholz, “Efficiency of recombination radiation in GaP”

Phys. Lett. 8, 233 (1964).

[5] R. A. Logan, H. G. White and W. Wiegmann, “Efficient green electroluminescence in nitrogen-doped GaP p-n junctions” Appl. Phys. Lett. 13, 139 (1968).

[6] M. Ishikawa, Y. Ohba, H. Sugawara, M. Yamamoto and T. Nakanishi,

“Room-temperature cw operation of InGaP/InGaAlP visible light laser diodes on GaAs substrates grown by metalorganic chemical vapor deposition” Appl. Phys. Lett. 48, 207 (1986).

[7] K. Kobayashi, S. Kawata, A. Gomyo, I. Hino and T. Suzuki, “Room-temperature CW operation of AlGalnP double-heterostructure visible lasers” Electron. Lett. 21, 931 (1985).

[8] M. Ikeda, Y. Mori, H. Sato, K. Kaneko and N. Watanabe, “Room-temperature continuous-wave operation of an AlGaInP double heterostructure laser grown by atmospheric pressure metalorganic chemical vapor deposition” Appl. Phys. Lett. 47, 1027 (1985).

[9] Lumilleds Lighting LLC, ICNS-5, 2003.

[10] K. Katayama, H Matsubara, F Nakanishi, T Nakamura “ZnSe-based white LEDs”

J. Cryst. Growth 214/215, 1064 (2000).

[11] R. M. Potter, J. M. Brank and A. Addamiano “Silicon carbide light-emitting diodes”

J. Appl. Phys. 40, 2253 (1969).

[12] H. Amano, M. Kito, K. Hiramatsu and I. Akasaki, P-type conduction in Mg-doped GaN treated with low electron beam irradiation (LEEBI)”, Jpn, J. Appl. Phys. 28, L2112 (1989).

[15] S. Nakamura, M. Senoh, S. Nagahama, N. Iwasa, T. Yamada, T. Matsushita, Y.

Sugimoto and H. Kiyoku, “ Room-temperature continuous-wave operation of InGaN multi-quantum-well structure laser diodes” Appl. Phys. Lett. 69, 4056 (1996).

[16] A. Tsukazaki, A.Ohtomo, T. Onuma, M. Ohtani, T. Makino, M. Sumiya, K. Ohtani, SF Chichibu, S. Fuke, Y. Segawa, H. Ohno, H. Koinuma, and M. Kawasaki, “Repeated temperature modulation epitaxy for p-type doping and light-emitting diode based on ZnO” Nature Mater. 4, 42 (2005).

[17] S. Koizumi, M. Kamo, Y. Sato, H. Ozaki and T. Inuzuka, ”Growth and characterization of phosphorous doped {111} homoepitaxial diamond thin films” Appl.

Phys. Lett. 71, 1065 (1997).

[18] S. Koizumi, K. Watanabe, M. Hasegawa and H. Kanda, "Ultra Violet Emission from a Diamond pn Junction" Science 292, 1785 (2001).

[19] T. Makino, N. Tokuda, H. Kato, M. Ogura, H. Watanabe, S-G. Ri, S. Yamasaki and H. Okushi, “High-efficiency excitonic emission with deep-ultraviolet light from (001)-oriented diamond p–i–n junction” Jpn. J. Appl. Phys. 45, L1042 (2006).

[20] R. F. Kazarinov and M. R. Pinto, “Carrier transport in laser heterostructures” IEEE J. Quantum Electronics 30, 49 (1994).

[21] W.Walukiewicz, “Intrinsic limitations to the doping of wide-gap semiconductors”

Physica B 302, 123 (1999).

[22] J. Vetterhoffer and J. Weber, “Hydrogen passivation of shallow donors S, Se, and Te in GaAs” Phys. Rev. B 53, 12835 (1996).

[23] H. Terao and H. Sunakawa, “Effects of oxygen and water vapour introduction during MOCVD growth of GaAlAs” J. Cryst. Growth 68, 157 (1984).

Chapter 2 Doping technologies for InGaAlP light emitting devices

2.1 Introduction to InGaAlP alloys

The InGaAlP quaternary alloy system is an important material for visible light emitting devices. The InGaAlP bandgap energy, the corresponding wavelength, and the nature of the bandgap (direct or indirect), are shown in Fig. 2.1[1]. The vertical dashed line in this figure indicates the composition at which InGaAlP is lattice matched to GaAs. As shown in this figure, the In0.5(Ga1-xAlx)0.5P alloy has a direct bandgap from 1.9 to 2.26 eV (x=0 to x~0.5), which covers the red to green portion of the visible spectrum, when grown lattice matched on gallium arsenide (GaAs) substrate. The direct-indirect transition of the bandgap occurs at the energy of 2.23-2.33 eV corresponding to 556-532 nm. Unlike GaAsP, which is grown on GaAs, InGaAlP does not suffer from lattice mismatch. Unlike nitrogen (N)-doped GaP, InGaAlP does not rely on deep-level-mediated transitions that tend to saturate. These properties make it an attractive material for making high-efficiency double hetero structure (DH) devices with a wide color range. The development of molecular beam epitaxy (MBE) and metalorganic chemical vapor deposition (MOCVD) has enabled the production of high quality, thin-film crystals for InGaAlP alloy system [2-4].

This alloy system characteristically exhibits CuPt structure atomic ordering of the group III atoms along the [111] direction (Fig. 2.2), which is in common with a number of other ternary and quaternary III-V alloys. Atomic ordering has a pronounce effect on the bandgap energy of the semiconductor layer with a difference of around 70 meV between fully disordered (largest bandgap) and fully ordered material [5]. It has been shown that ordering of InGaAlP is a strong function of growth conditions [5-7].

Doping of p-type or n-type cladding layers has a strong influence on the performance of double heterostructure LDs and LEDs. The resistivity of the cladding region is one factor in determining the doping concentration in the cladding layers. The resistivity should be low to avoid heating of the cladding layers. One of the other factors is carrier leakage current from the active layer to the cladding layers. In an

Figure 2.1. The bandgap energy of the InGaAlP alloy and its

corresponding wavelength versus lattice constant [1].

The barrier height, δp in Fig. 2.4, can be determined by the bandgap difference ∆Eg

between the active layer and the cladding layer and the quasi-Fermi level Φp for the p-type cladding layer, and is given as follows:

),

(

p g

p

E φ φ

δ = ∆ − Φ − +

act.

g clad g

g E E

E = −

∆ (2.2)

The parameters, φn and φp in Eq. (2.1), are quasi-Fermi levels for electrons and holes in the active layer, respectively, and are related to the injected carrier densities n and p by the following equations.

_1/2

,

⎥⎦ ⎤

⎢⎣ ⎡

= N F kT n

φ

(2.3)

_1/2 ⎥,

⎦

⎢ ⎤

⎣

= ⎡ F kT N

p _V

φ

(2.4)

Where NC and NV are the effective densities of states for the conduction band and valence band, respectively, and F1/2[x] represents the Fermi-Dirac integral. Τhe value of Φp in Eq. (2.1) is determined by the acceptor concentration p in the p-cladding layer, and can be decreased by increasing p. This means that carrier overflow can be reduced by using a highly doped p-cladding layer.

As the wavelength decreases, so the band offset must decrease due to the associated increase of the active region bandgap. The band offsets can be increased by increasing the Al concentration in the cladding layer, however, this has proven difficult. The reason for this is mainly due to the low incorporation and the low electrical activity of Zn in higher Al concentration InGaAlP layers (see Fig. 2.5 [13]).

Another factor limiting the shortest useable wavelength is the efficiency of the active region. The overall optical efficiency of the active region decreases with decreasing wavelength. This is mainly due to the fact that to construct an active region to produce a short wavelength the easiest technique is to increase the bandgap of the region by increasing the Al content of the layer [14]. However, increasing the Al content tends to result lower optical efficiency, as shown in Fig. 2.6 [15].

(x=0-1.0). The n-type doping characteristics were systematically investigated, and two types of Si donor- related states (shallow and deep) were shown to exist in this alloy system. In Section 2.5, the influences of residual impurities on n-type and p-type doping properties in InGaAlP are shown. It is shown that Oxygen and hydrogen significantly affects the ionization of the dopant impurity. In Section 2.6, the effects of substrate misorientation on n-type and p-type doping characteristics are shown. The Zn incorporation efficiency is shown to be improved by using off-angle substrate. In Section 2.7, effects of substrate misorientation on reduction in residual impurity concentration and the related deep level concentrations are shown. To use off-axis substrates makes great advantage for InGaAlP light emitting devices. In Section 2.8, improved device performances of an LED and an LD with InGaAlP alloys are shown.

Section 2.9 is the summary of this chapter.

Cu Pt

In Ga

P (a)

(b)

Figure 2.3. Energy band offsets for the conduction and valence

bands for unordered In

(Ga

Al

)

P relative to In

Ga

P as a

function of Al mole fraction x. The maximum conduction band

offset (∆E

) is approximately 0.2 eV and occurs in the range of x ~

0.5 to 0.7. The maximum valence band offset (∆E

) occurs for

In

Al

P and is approximately 0.24 eV.

Figure 2.4. Schematic band diagram for InGaAlP double  

heterostructure.

Figure 2.5. Hole concentration and Zn concentration versus Al mole

fraction.

Figure 2.6. Photoluminescence (PL) intensity and bandgap

energy of InGaAlP alloys as a function of aluminum mole fraction

x.

2.2 MOCVD growth of InGaAlP alloys

In the present study, the InGaAlP layers were grown on GaAs substrates by low-pressure (LP)-MOCVD. Fig. 2.7 shows a schematic of a gas delivery system for the InGaAlP MOCVD reactor. For the alkyl delivery, individual pressure –controlled alkyl bubblers in temperature- controlled baths provide reliable and constant delivery. It is always critical in MOCVD that the gas plumbing provides rapid, efficient, and controllable switching of gases into the reactor chamber without perturbing the reactor pressure or the delivery pressure from the alkyl bubbler. For this reason it is typical to implement a run-vent arrangement of injection a set-up gas flow into a reactor.

Following the delivery of metalorganic and hydride sources through the plumbing manifold into the reactor chamber, chemical reactions occur within the thermal boundary near the GaAs substrate and on the surface for epitaxial growth of InGaAlP.

Because InGaAlP is a quaternary system, careful control is required to ensure lattice-matching of InGaAlP to the GaAs substrate. This is crucial for obtaining high-quality materials. Second, Al is very reactive and binds easily to oxygen. It was found that one of the most effective ways to suppress oxygen incorporation is to increase the growth temperature [16]. However, high growth temperature is not ideal for In-containing alloys, which are typically grown at much lower temperatures.

Re-evaporation of In can be also serious problem at higher temperature. Thus, the optimal growth temperature “window” for InGaAlP must be narrow.

The growth conditions in the present study were determined considering aforementioned matters. Typical growth conditions of InGaAlP used for this study are summarized in Table 2.1.

Material sources: Group-III Trimethylgallium (TMG) Trimethylaluminium (TMA) Trimethylindium (TMI)

Group-V Phosphine (PH

)

Arsine (AsH

)

Dopant Dimethylzinc (DMZ)

Silane (SiH

)

Substrate temperature: 730 ℃ Total pressure: 25 Torr

V/III ratio: 450

Growth rate: 3µm/h

Table 2.1 Typical growth conditions of InGaAlP alloys.

2.3 Background of p-type and n-type doping of InGaAlP alloys

Zn and Mg are widely used as p-type dopants. Generally, Mg acceptors are expected to have smaller ionization energies than Zn. It is possible to achieve a higher hole concentration for InGaAlP with Mg, especially for high Al content InGaAlP [15].

In addition, several studies have shown that Mg has a lower diffusion coefficient than Zn in III-V materials [18-24]. However, it was difficult to control the Mg doping profile due to the severe memory effect of Mg, which means remaining Mg on the surface of the reactor or the source gas pipe after Mg source gas turning off causes a residual impurity in undoped or n-type layer on the Mg-doped layer. The “delay” and the “tail” can be seen in the Mg doping profile [18, 19]. On the other hand, it has been shown that the Zn incorporation efficiency (the ratio of Zn concentration in the solid phase to Zn concentration in the vapor phase) and Zn electrical activity (the ratio of the net acceptor concentration to the Zn concentration in the solid phase) in Zn-doped InGaAlP decreases with decreasing Al mole fraction x, as shown in Fig. 2.5 [13]. The maximum acceptor concentration in Zn-doped InAlP is limited to 2 X 10¹⁷ cm^-3 [25].

For n-type doping, of early InGaAlP samples grown by MOCVD, Se was used as n-type dopant. However, Se has become less popular because it generates a severe memory effect [26], and the incorporation is decreasing with increasing Al content [14].

The advantage of using Si is that the Si dopant source, SiH4, generates essentially no memory effect and Si exhibits relatively low diffusivity in the solid. Te is also used as an n-type dopant in this alloy system [27].

2.4 Si doping characteristics in InGaAlP alloys

2.4.1 DX centers in III-V semiconductor alloys

N-Type dopants, such as Si, Se, Te, in III-V semiconductor alloys display a somewhat complex behavior. The presence of such donor related deep states has previously been detected in a number of alloy semiconductors such as GaAsP [28, 29], InGaAlP [30, 31] and AlGaAs [32-35]. In particular, for n-type AlGaAs, there has been a considerable amount of experimental and theoretical effort directed towards a deeper

In Se-doped In0.5(Ga1-xAlx)0.5P (x > 0.4) grown by MOCVD [30] and in Si-doped InGaAlP (x = 0.48) grown by MBE [31], donor-related and DX-like deep states have been found. However, few investigations of electrical properties and/or of the conduction band structure in relation with these donor states have been reported in the literature.

In the present study, we have systematically investigated for the first time the donor states and electrical properties of Si-doped InGaAlP (x = 0-1.0) grown by MOCVD.

The Si doping characteristics were investigated, and two types of donor-related states (shallow and deep) are shown to exist in this alloy system.

2.4.2 Experiments

The Si-doped In0.5(Ga1-xAlx)0.5P layers were grown by LP-MOCVD. The growth temperature was 730℃, the total pressure was 25 Torr, and the growth rate was 3µm/h.

Trimethylindium (TMI), trimethylgallium (TMG), trimethylaluminium (TMA), phosphine (PH3) and arsine (AsH3) were used as the source materials. The dopant sources were silane (SiH4) and dimethylzinc (DMZ) as n-type and p-type dopants, respectively. A 0.5 µm Zn-doped p-type GaAs layer, followed by a 0.5 µm Zn-doped p-type InGaAlP layer (to eliminate the influence of parallel conduction) and a 2 µm Si-doped InGaAlP were successively grown on the (100)-oriented semi-insulating GaAs substrate. The Al mole fraction, x, was varied in the range 0 < x < 1.0. Schottky diodes were fabricated by evaporating and alloying gold (Au)/ gold-germanium (AuGe) for ohmic contacts and evaporating Au to construct a 200 µm diameter Schottky barrier contacts. The cross sectional view of the schematic sample structure is shown in Fig.

2.8.

Secondary ion mass spectroscopy (SIMS) was used to determine the Si concentration in InGaAlP layers. Calibration was carried out using ion-implanted samples. The net donor concentration, ND-NA, was evaluated by capacitance-voltage (C-V) measurements at room temperature. The carrier concentrations, n, and electron thermal activation energies were determined employing Van der Pauw-Hall effect measurements over the temperature range 77-350K. X-ray diffraction measurement confirmed the lattice mismatch of InGaAlP layers to GaAs substrates within + 0.1%.

In and Al mole fraction were confirmed by x-ray microanalysis (XMA).

To characterize deep levels, deep level transient spectroscopy (DLTS), C-V measurements and photocapacitance (PHCAP) measurements were performed. DLTS was performed over the temperature range 77 to 400 K.

Figure 2.8. Cross sectional view of the schematic sample structures.

2.4.3 Doping characteristics

Figure 2.9 shows the net donor concentration, ND-NA, and the Si concentration, NSi, in In0.5(Ga0.3Al0.7)0.5 at room temperature as a function of the molar ratio of SiH4 to the group-III sources ([SiH4]/[III]). The net donor concentration and the Si concentration increases in proportion to the [SiH4]/[III] ratio. The Si electrical activity (the ratio of donor concentration to Si concentration) was found to be about 0.8, being independent of [SiH4]/[III]. The same dependences, of ND-NA, and NSi on [SiH4]/[III], were obtained for other alloy compositions.

2.4.4 Deep levels

Figure 2.10 shows the typical DLTS spectra obtained for several Al contents of Si-doped In0.5(Ga1-xAlx)0.5P alloys (x=0, 0.4, 0.7 and 1.0). The net donor concentration for these samples was between 5 and 7 X 10¹⁷ cm^-3. There were no detectable deep levels in In0.5Ga0.5P (x=0). The broadened spectra shown in Fig. 2.10, as has been widely reported for other compound semiconductors [28-35], also are thought to be the result of an alloy effect [29, 40]. Ahrenius plots show that the activation energy of electron thermal emission, EDLTS, is consistently 0.42 eV, independent of x.

The correlation between observed deep level concentration and donor concentration, for x=0.7, is shown in Fig. 2.11. The deep level concentration, NT, can be determined by employing a combination of C-V measurements at room temperature and at 77K [30].

When the C-V measurement is carried out at room temperature, the total donor concentration including the shallow donor concentration (ND-NA) and the deep level concentration (NT). On the other hand, the ND-NA can be directly obtained from the C-V measurement at 77K due to the small time constant even at such low temperature. The deep level concentration was found to increase linearly with increasing net donor concentration. This indicates that the deep levels are either donor-related levels or generated by doping.

Figure 2.12 shows the normalized concentration of donor-related deep levels, NT/(ND-NA), as a function of x. The concentration increases rapidly with increasing x up to a maximum at around x=0.5, it then decreases over the range x=0.5 to x=1.0. These results show that the deep trap level concentration is comparative or higher than the shallow donor concentration for the Al mole fraction x > 0.4.

The optical emission energy, required to remove electrons from deep donor states, Eopt, was determined by PHCAP measurements [41]. A constant capacitance method was used. For x = 0.7, Eopt of the main electron trap level was found to be 1 eV. The concentration of the trap levels, determined by the capacitance change due to optical

Figure 2.10. DLTS spectra of In

(Ga

Al

)

(x = 0, 0.4, 0.7, 1.0).

Figure 2.11. Deep level concentration versus net donor

concentration for [SiH

]/[III] in In

(Ga

Al

)

P.

Figure 2.12. Normalized deep-level concentration versus Al mole

fraction x of In

(Ga

Al

)

P.

emptying of traps at an excitation energy of 1eV, was found to be nearly equal to that found for the deep levels detected by DLTS. Therefore, this 1 eV optical stimulation energy can be identified as corresponding to the deep levels detected by DLTS.

According to previous work on the DX center in AlGaAs by Lang et al. [42], this large value of optical emission energy indicates that electron captured at this trapping level is accompanied by a large lattice relaxation.

2.4.5 Shallow and deep donor states

Figure 2.13 shows the dependence of net donor concentration, the carrier concentration, at room temperature, and thermal activation energy of the equilibrium carrier concentration on the Al mole fraction. The energy values were determined by Hall effect measurements (150 to 300 K). The net donor concentration was found to be almost independent of x. However, the carrier concentration gradually increased with increasing x. The thermal activation energy was found to be rather small for x < 0.3 and to increase rapidly with increasing x for 0.3 < x < 0.5. It reached a maximum at x = 0.5, and then gradually decreased as the Al mole fraction approached the ternary end point. It is obvious that the minimum carrier concentration at x = 0.5 can be attributed to the maximum value of EC-ED at the point [43].

The donor levels from conduction band edge, EC-ED, shown in Fig. 2.13 and the Γ, X and L minima of the conduction band [9] are shown in Fig. 2.14, as a function of x.

This result indicates that two types of donor states exist. One is the shallow state observed for x < 0.3. This has a small donor binding energy, which is in agreement with the values calculated from the effective mass approximation [44]. The other state is the deeper state observed in the range x > 0.3. This has rather large binding energies than those estimated by the effective mass approximation. Considering the thermal emission energy (EDLTS = 0.42 eV), it is suggested that the electron capture barrier relevant to the occupation of the deeper state [45] is quite large. These results indicate that the behavior of the donor state is similar to that of AlGaAs, and especially that deeper state is similar to the DX levels commonly found in AlGaAs [32-35], GaAsP [28, 29] and Se-doped InGaAlP [30]. For x > 0.3, the shallow donor states could not be determined by the Hall effect measurements in this study, because the contact

Figure 2.13. Net donor concentration, electron concentration, and

electron thermal activation energy versus Al mole fraction x of in

In

(Ga

Al

)

P net donor concentration for [SiH

]/[III].

Figure 2.14. Donor levels and conduction band minima versus Al mole fraction x of In

(Ga

Al

)

P.

E

E

E

Energy (eV)

Al Mole Fraction x

In

(Ga

Al

)

P

Shallow

Deep

0.0 0.5 1.0

1.8

2.0

2.2

2.4

2.6

2.8

the deepest donor state occurring at around x = 0.5 can be explained by the dependence of the conduction band structure on alloy composition. The curious dependence of EC-ED on alloy composition can be explained by association of the carriers with the higher conduction band minima (X or L), as in the case of AlGaAs [37]. According to this model in the case of AlGaAs, EC-ED displays a maximum at the direct-indirect crossover point (x ~ 0.45) in AlxGa1-xAs [46]. In the case of InGaAlP, it has been reported that the direct-indirect crossover point, x = 0.53 [11] coincides with the alloy composition ratio x ~ 0.5 which gives the maximum value of EC-ED. The results of this study supported the previous report. Thus, the deeper states found in this study in InGaAlP are considered to be DX centers. As in the case of AlGaAs [46, 47], the local environment around a donor atom strongly should affect the nature and/or the activation energy of the deep states.

2.4.6 Electrical properties of Si-doped InGaAlP

The electrical resisitivity and the electron Hall mobility are shown in Figs. 2.15 and 2.16, respectively as a function of Al mole fraction. The resistivity rapidly increased with increasing below x ~ 0.5 and then saturated at around 0.3 Ωcm. The electron mobility was rapidly decreased over x = 0.2, and saturated at around 50 cm²/V/sec.

Because the electron concentration was almost independent of the Al mole fraction, the rapid increase in the resistivity is thought to be caused by the rapid decrease in the electron mobility. The rapid decrease in the electron mobility was considered to be due to the change of the conduction band minimum from direct to indirect at around x = 0.5 [8]. The fact that the electron effective mass in X-band is considerably larger than that in Γ-band should give a small value of electron mobility. These results also support that the direct-indirect (Γ - X) cross point is at around x ~ 0.5.

2.4.7 Summary of Section 2.4

Si doping characteristics were systematically investigated in Si-doped In0.5(Ga1-xAlx)0.5P alloys (x = 0-1.0). The net donor concentration and the Si concentration increased in proportion to the [SiH4]/[III] ratio. The Si electrical activity (the ratio of donor concentration to Si concentration) was found to be about 0.8, being independent of [SiH4]/[III]. DLTS measurements have revealed deep levels, with a constant thermal emission energy value (EDLTS = 0.42), for x > 0.3. There are indications that the electron capture to this level is accompanied by a somewhat large lattice relaxation, from the result of the large 1.0 eV optical stimulation energy, determined by PHCAP. The electron thermal activation energy, EC-ED, was rather

small in the Al mole fraction range x < 0.3. With increasing x, this energy value rapidly increased, reaching a maximum (~ 80 meV) at x = 0.5. The relation between the donor states and the conduction band structure is consistent with the same model as that used for DX centers in AlGaAs alloys.

Figure 2.15. Electrical resistivity versus Al mole fraction x of

In

(Ga

Al

)

P.

Figure 2.16 Hall mobility versus Al mole fraction x of In

(Ga

Al

)

P.

2.5 Residual impurities in InGaAlP alloys

2.5.1 Residual impurities in III-V semiconductors

Several authors have investigated the residual impurities in AlGaAs grown by MOCVD. Oxygen contamination is serious problem in the growth of Al-containing layers, since Al reacts vigorously with oxygen to form strong bonds. Degradation of the optical and electrical properties of AlGaAs due to the presence of O2 and H2O in the growth ambient has been observed [48-53]. The primary sources of O2 and H2O are considered to be gases such as H2 and AsH3, and the various dopant gases [52].

The phenomenon of hydrogen passivation of Zn acceptors has been reported in GaAs [54, 55], in InP [56-59] and in InGaAs [60]. Hydrogen is incorporated into the layer during growth and/or during cooling after growth. The source of atomic hydrogen is thought to be AsH3 and PH3. The incorporation of hydrogen into the crystal would make Zn acceptors electrically inactive.

Residual impurities such as oxygen and hydrogen are considered to influence the electrical properties of InGaAlP, also.

In the present study, we examine hydrogen and oxygen in InGaAlP grown by LP-MOCVD and show its effects on Zn doping and Si doping properties.

2.5.2 Experiments

Layers of In0.5(Ga1-xAlx)0.5P are grown using LP-MOCVD. TMG, TMA, TMI, AsH3

and PH3 were used as source materials. DMZ and SiH4 were used as p-type and n-type dopants, respectively. The carrier gas was palladium-purified H2. Conditions were a growth rate of about 3 µm/h, a total pressure of 25 Torr and a substrate temperature (Ts) of 730 ^oC. V/III ratio was varied between 200 and 800. Figure 2.17 shows the sample structure studied here. A 0.5 µm thick undoped InGaAlP layer and a 1.0 µm thick Zn or Si doped InGaAlP layers were successively grown on a (100) GaAs substrate.

A cap layer of Zn doped p-GaAs or Si-doped n-GaAs was grown on the InGaAlP layer.

The cap layer was about 0.1 µm thick. The net donor concentration in the n-GaAs cap was 2 X 10¹⁸ cm^-3 and the net acceptor concentration in the p-GaAs was also 2 X 10¹⁸ cm^-3. The Al mole fraction x in In0.5(Ga1-xAlx)0.5P was varied between 0, 0.4, 0.7, 1.0.

C-V measurements were used to determine the net acceptor concentration, NA-ND, and the net donor concentration, ND-NA. The concentrations of H, O, Zn and Si (NH, NO, NZn and NSi) in the InGaAlP were examined by SIMS. These concentrations were calibrated using ion-implanted samples. Zn electrical activity was calculated from the

Figure 2.17. Cross-sectional view of the sample structure. The cap

layer is p-GaAs or n-GaAs.

net acceptor concentration divided by the Zn concentration (ηZn = (NA-ND)/ NZn). Si electrical activity, ηSi, was calculated as ηSi = (ND-NA)/ NSi.

2.5.3 Residual hydrogen doping in InGaAlP alloys

Figure 2.18 shows the net acceptor concentration in Zn-doped In0.5(Ga0.3Al0.7)0.5P with p-type GaAs or n-type GaAs cap layer as a function of DMZ introduction ([DMZ]/[III]), defined as the DMZ molar fraction divided by the total molar fraction of the group-III sources in the vapor phase. It can be seen that a higher net acceptor concentration is obtained with an n-type GaAs cap layer compared with a p-type GaAs cap layer. The net acceptor concentration in Zn-doped InGaAlP with an n-type GaAs cap layer increased with increasing DMZ introduction, while the net acceptor concentration with a p-type GaAs cap layer clearly saturated at about 1X10¹⁷ cm^-3. The Zn concentration in Zn-doped InGaAlP with an n-type GaAs cap layer was the same as that with a p-type GaAs cap layer. Thus, the difference in the net acceptor concentration is due to the difference in Zn electrical activity in InGaAlP. The Zn electrical activity is strongly affected by the conduction type of the cap layer grown on top of the Zn-doped layer. It is affected neither by the material nor by the thickness of the cap layer [61]. Larger Zn electrical activity can be achieved by growing an n-type cap layer.

Hydrogen passivation of Zn acceptors has been reported in other materials [54-60], and expected hydrogen to be involved in the change in Zn electrical activity in InGaAlP according to the conduction type of the cap layer. To clarify the relationship between hydrogen and Zn electrical activity, we examined the hydrogen concentration in samples (a) and (b), as marked in Fig. 2.18. The Zn concentration in sample (a) was 8 X 10¹⁷ cm^-3, just the same as in sample (b). However, the Zn electrical activity in sample (a) (ηZn = 0.13) was much lower than that in sample (b) (ηZn = 0.53), as shown in Fig. 2.18.

Figure 2.19 shows the hydrogen and Zn concentration profiles for sample (a). The p-type GaAs layer was chemically etched off before SIMS measurements. The hydrogen concentration in the Zn dope InGaAlP layer of about 1 X 10¹⁸ cm^-3 was higher than in the undoped layer. The hydrogen concentration in the undoped layer was below the SIMS detection limit (5 X 10¹⁷ cm^-3). Thus, the hydrogen was shown to be selectively incorporated in to Zn-doped InGaAlP. The hydrogen and Zn concentration profiles observed in sample (b) are shown in Fig. 2.20. The n-type GaAs cap layer was etched off before SIMS measurements. The hydrogen concentration in the Zn-doped InGaAlP layer was below the detection limit, as in the undoped layer. The Zn electrical activity in Zn-doped InGaAlP with a p-type GaAs cap layer was quite low and a hydrogen

Figure 2.18. Net acceptor concentration in Zn-doped

Figure 2.19. Zn and Hydrogen depth profiles in sample (a), which

has a p-type GaAs cap layer, measured by SIMS.

Figure 2.20. Zn and Hydrogen depth profiles in sample (b), which