Iii-nitride Semiconductor Materials

Ill-Nitride Semiconductor Materials Zhe Chuan Feng editor Imperial College Press Ill-Nitride Semiconductor Materia...

Author: Zhe Chuan Feng (Editor)

307 downloads 3050 Views 22MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

Ill-Nitride

Semiconductor Materials

Zhe Chuan Feng editor

Imperial College Press

Ill-Nitride

Semiconductor Materials

This page is intentionally left blank

Ill-Nitride

Semiconductor Materials

editor

Zhe Chuan Feng National Taiwan University, Taiwan

4fl^

Imperial College Presj

Published by Imperial College Press 57 Shelton Street Covent Garden London WC2H 9HE Distributed by World Scientific Publishing Co. Pte. Ltd. 5 Toh Tuck Link, Singapore 596224 USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601 UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE

British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library.

III-NITRIDE SEMICONDUCTOR MATERIALS Copyright © 2006 by Imperial College Press All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means, electronic or mechanical, including photocopying, recording or any information storage and retrieval system now known or to be invented, without written permission from the Publisher.

For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy is not required from the publisher.

ISBN

1-86094-636-4

Printed in Singapore by World Scientific Printers (S) Pte Ltd

CONTENTS Preface

xi

Chapter 1

Hydride vapor phase epitaxy of group III nitride materials V. Dmitriev, A. Usikov 1-40 1 1. Introduction 2. Experiment 4 3. Material Properties 5 3.1. Undoped GaN layers 6 3.2. Si-doped GaN layers 10 3.3. Mg-doped GaN layers 11 3.4. Zn-doped GaN layers 13 3.5. A1N layers 14 3.6. AlGaN layers 16 3.7. InN and InGaN layers 19 4. New directions in HVPE development 21 4.1. Large area and multi wafer HVPE growth 21 4.2. Multi-layer structures 23 4.3. P-n junctions 24 4.4. Structures with two dimensional carrier gas 26 4.5. Nano structures and porous materials 26 5. Applications of HVPE grown group III nitride materials 27 5.1. Substrate applications 27 5.1.1. Template substrates 27 5.1.2. Free-standing substrates 28 29 5.1.3. Bulk substrates 31 5.2. Device Applications 6. Conclusions 33

Chapter 2

Planar MOVPE technology for epitaxy of Ill-nitride materials M. Dauelsberg, B. Schineller, J. Kaeppeler 1. History of Reactor Development for Ill-Nitrides 2. Types of Planar Reactors 3. Reactor Modeling 3.1. Growth Kinetics of Group-Ill Nitride MOVPE 3.2. Modeling of MOVPE processes 3.3. Horizontal Tube Reactors: Flow Dynamics and Reactor Technology

41-71 41 43 45 46 48 50

VI

Contents

3.4. Planetary Reactors: Transport Phenomena & Parameter Dependencies 4. In-situ Technology in Nitride MOCVD Systems 5. The Mass Production of GaN and Related Materials 5.1. Optoelectronic Device Structures 5.2. Growth in the 8x4 inch Configuration 5.3. Special Aspects of Growth on 4 inch 5.4. The Growth of InGaN Structures 5.5. The Growth of AlGaN Structures 5.6. High Growth Rates in Mass Production Reactors

53 57 62 62 63 63 67 68 69

Chapter 3

Close-Coupled Showerhead MOCVD technology for the epitaxy of GaN and related materials E. J. Thrush, A. R. Boyd 73-116 1. Introduction 74 2. Reactor Development 75 2.1. Close-Coupled Showerhead Technology 76 2.2. Modeling 80 2.3. Scale-up of CCS Reactors 85 3. In-situ Optical Monitoring 86 3.1. Optical Pyrometry 86 3.2. Laser Interferometry (Reflectometry) 88 3.3. Emissivity Corrected Pyrometry 91 3.4. Spectral Reflectometry 93 4. Reagent Concentration Monitoring 94 5. Growth Results 97 5.1. GaN Template Growth in Close-Coupled Showerhead Reactors 97 5.2. Quantum Wells in Close-Coupled Showerhead Reactors 97 5.3. HEMT Structures in Close-Coupled Showerhead Reactors 102 6. Mass Production of GaN Based Devices 707 6.1. Cost of Ownership Dependence on System Size 707 6.2. Process Cycles 709 6.3. Growth Data from 19x2" Production System 111 7. Acknowledgements 114

Chapter 4

Molecular beam epitaxy for III-N materials H. Tang, J. Webb 1. Introduction 2. MBE technology 2.1. Ammonia vs Plasma Assisted MBE

117-160 777 775 727

Contents

Semi-insulating GaN 3.1. Growth of semi-insulating GaN 3.2. Carbon doping 3.3. Other dopants for semi-insulating GaN Electrical Characteristics 4.1. Polarization Induced Two Dimensional Electron Gas 4.2. Reproducibility and Uniformity 4.3. Low Temperature Mobility 4.4. Magnetotransport Properties AlGaN/GaN HFET Device Characteristics 5.1. DC and Small-Signal RF Characteristics 5.1.1. Devices Grown on Sapphire and SiC by Ammonia MBE 5.1.2. Devices Grown on Resistive Silicon by Ammonia MBE 5.1.3. Devices Grown on Sapphire and SiC by PAMBE 5.2. Large-Signal RF Performance 5.3. Device Scalability and Large Periphery Performance Conclusion

Chapter 5

vn

123 123 124 129 130 130 132 135 138 144 145 145 147 149 150 154 156

Growth and properties of nonpolar GaN films and heterostructures J. Sun, 0. Brandt 161-202 Introduction 161 Substrate: y-LiAlO., 163 2.1. Properties of the substrate 163 2.2. Polarity of the substrate 166 2.3. In-plane orientation relationship 168 Optimization of Heteroepitaxy of GaN Films Ony-LiAIO2(100) 168 3.1. Impact of nucleation conditions on phase purity 169 3.2. Influence of nucleation temperature on surface quality 172 Ga Adsorption and Desorption Kinetics 174 4.1. Surface reconstruction 174 177 4.2. Ga adsorption/desorption kinetics M-plane (In,Ga)N/GaN Multiple Quantum Wells 185 5.1. In incorporation and surface segregation 186 5.2. Recombination mechanism 190 Conclusion and Outlook 196

Contents

Vlll

Chapter 6

Chapter 7

Indium-nitride growth by high-pressure CVD: real-time and ex-situ characterization N. Dietz 203-235 1. Introduction 204 2. HPCVD Reactor Characteristics 206 3. Reactor flow characterization 209 4. Precursor characterization: Ammonia and Trimethlyindium 211 4.1. Optical characterization of Trimethylindium, TMI, [In(CH3)3] 212 4.2. Optical characterization of ammonia (NH3) 214 5. Flow kinetics: Analysis utilizing pulsed gas injection 216 5.1. Flow characterization during pulsed precursor injection 218 6. Precursor decomposition dynamics at higher pressures 220 7. Growth of InN: Real-time optical monitoring 222 8. Ex-situ characterization of InN layers 227 9. Summary and Outlook 232

A new look on InN L.-W. Tu, C.-L. Hsiao, M.-H. Tsai 1. Introduction 2. Growth and Structures 3. Compositions and Electrical Properties 4. Optical Characterizations 5. Four Possibilities 6. Conclusions and Outlook

Chapter 8

237-256 237 239 245 247 249 250

Growth and optical/electrical properties of AlxGa!_xN alloys in the full composition range F. Yun 257-304 1. Introduction 257 2. Material Growth of AlGaN 260 2.1. Growth Techniques and Full Range Control of Al 260 2.2. Conduction type control of AlGaN 271 2.3. Dislocation and Morphology Analysis 272 2.4. Chemical Ordering in AlGaN 280 3. Electrical Properties 282 3.1. Electrical Transport 282 3.2. Deep Level Defects 286 4. Optical Properties 289 5. Band Gap Bowing of AlGaN 295 6. Summary 299

Contents

Chapter 9

IX

Optical investigation of InGaN/GaN quantum well structures grown by MOCVD Tao Wang 305-343 1. Introduction 305 2. Strain-relaxation in InGaN/GaN MQW 307 3. Quantum-confined Stark effect and exciton-localization Effect 313 4. Optical Investigation of InGaN/GaN MQWS under high excitation 320 5. Study of stimulated emission from InGaN/GaN multiple quantum well structures 326 6. Femtosecond studies of electron capture times in InGaN/GaN MQWs 331 7. Summary 337

Chapter 10 Clustering nanostructures and optical characteristics in InGaN/GaN quantum-well structures with silicon doping Y.-C. Cheng, C.-Y. Chen, C. C. Yang 345-368 1. Introduction 346 2. Sample Preparation and Measurement Conditions 348 3. Photoluminescence and Photoluminescence Excitation 349 3.1. Amplified Spontaneous Emission 353 3.2. Cathodoluminescence 356 3.3. Strain State Analysis 360 4. Discussions 361 5. Conclusions 366 Chapter 11 Ill-nitrides micro- and nano-structures H. M. Ng, A. Chowdhury 369-398 1. Introduction and Overview 369 2. Nanostructures by Epitaxy 370 3. Nanostructures by Selective Etching and Regrowth 375 3.1. Photoelectrochemical Etching 376 3.2. Polarity Selective Chemical Etching (PSCE) 377 3.3. ID and 2D Periodic Structures Formed by PSCE 383 4. Applications of GaN Micro- and Nanostructures 387 4.1. Second Harmonic Generation 387 4.2. Electron Field Emission 389 5. Summary and Future Outlook 393

X

Contents

Chapter 12 New developments in dilute nitride semiconductor research W. Shan, W. Walukiewicz, K. M. Yu, J.W. AgerIII, E.E. Haller 1. Introduction 2. Material Properties 3. Device Applications 3.1. Long-wavelength Laser Diodes for Telecommunications 3.2. Solar Cells for Photovoltaic Solar Power Conversion 4. Origin of Band-gap Reduction in Dilute Nitrides 4.1. Large B and-gap B owing and Early Impurity Models 4.2. Band Anticrossing 4.3. E. and E+ transitions 4.4. Enhancement in Maximum Free Electron Concentration 5. Concluding Remarks: From Dilute III-N-V Nitrides to Dilute II-O-VI Oxides

399-428 399 400 404 404 406 408 408 409 413 416 422

PREFACE

Group Ill-Nitrides semiconductor materials, including GaN, InN, A1N, InGaN, AlGaN and AlInGaN, i.e. (Al, In, Ga)N, are excellent semiconductors, covering the spectral range from UV to visible and to infrared, with unique properties very suitable for modern electronic and optoelectronic applications. Remarkable breakthroughs have been achieved in recent years for research and development (R&D) in these materials and devices, such as high-power and high-brightness blue-green-white light emitting diodes (LEDs) and blue laser diodes (LDs) as well as other optoelectronics and electronics devices and applications. III-Nitrides-based industry is forming up and new economic developments from these materials are promising. It is expected that IIINitrides-based LEDs might replace the traditional light bulbs to realize a revolution in lightings and change the entire human life in this century, similar to Edison's invention of the electric light bulb more than one hundred years ago. The developments on these materials and devices are moving so quickly. Many data or knowledge, even published quite recently, have been modified and need to upgrade. This new book will cover the rapidly new developments and achievements in the field, in particular those made after entering the 21st century. It is not pure science, but engineering and materials. The book is organized for a wide range of audiences and covers each of the basic and critical aspects of Ill-Nitrides science and technology. Each chapter, written by experts in the field, reviews the important topics and achievements in recent years, discusses progresses made by different groups, and suggests further works needed. This book provides useful information to material growth, characterization, application and development on the Ill-Nitrides semiconductor materials. This book consists of 12 well-written review chapters and the topics of the book include: 1) Hydride vapor phase epitaxy of group III nitride materials, 2) Planar MOVPE technology for epitaxy of Ill-nitride XI

Xll

Preface

materials, 3) Close-Coupled Showerhead MOCVD technology for the epitaxy of GaN and related materials, 4) Molecular beam epitaxy for IIIN materials, 5) Growth and properties of nonpolar GaN films and heterostructures, 6) Indium-nitride growth by high-pressure CVD: realtime and ex-situ characterization, 7) A new look on InN, 8) Growth and optical/electrical properties of AlxGai_xN alloys in the full composition range, 9) Optical investigation of InGaN/GaN quantum well structures grown by MOCVD, 10) Clustering nanostructures and optical characteristics in InGaN/GaN quantum-well structures with silicon doping, 11) Ill-nitrides micro- and nano-structures, and 12) New developments in dilute nitride semiconductor research. As seen, the current book presents the key properties of Ill-Nitrides materials, describes key technologies and demonstrates the remaining challenging issues in material growth and device fabrication for future R&D in the 21st century. This book is useful to material growers and evaluators, device design and processing engineers as well as potential users of Ill-Nitrides technologies, including new comers, postgraduate students, engineers and scientists in the Ill-Nitrides field.

Zhe Chuan Feng (Professor in Electro-Optical & Electrical Engineering, National Taiwan University)

CHAPTER 1 HYDRIDE VAPOR PHASE EPITAXY OF GROUP III NITRIDE MATERIALS

Vladimir Dmitriev and Alexander Usikov Technologies and Devices International, Inc., 12214 Plum Orchard Dr., Silver Spring, MD 20904, USA e-mail: [email protected] In this chapter we describe recent experimental results on hydride vapor phase epitaxy (HVPE) of group III nitride materials including epitaxial layers and multi-layer device structures. Properties of GaN, InN, AIN, and AlGaN layers grown by HVPE are presented. For GaN layers, ntype and p-type doping during HVPE growth is reported. Thick crack free AIN layers grown by stress control HVPE are described. New directions in HVPE technology including large area growth, multiwafer growth, sub-micron multi-layer growth, and fabrication of nanosize structures including GaN and InN nanowires are briefly discussed. Properties of HVPE grown AlGaN/GaN hetero structures with twodimensional electron and two dimensional hole gases are presented. Applications of HVPE grown group III nitride materials including substrate applications (template substrates, free standing substrates, and bulk substrates) and device structures for both optoelectronic application (blue and ultraviolet light emitting diodes grown by HVPE) and electronic applications (high electron mobility transistors grown by HVPE) are discussed as well.

1. Introduction The HVPE technology has been demonstrated to deposit single crystal layers of both GaN 1 and AIN 2 more than 30 years ago. Fig. 1 illustrates the basic idea of the method. For GaN growth, source materials are

1

2

Dmitriev, Usikov

gallium chloride and ammonia gases. Gallium chloride is formed inside growth apparatus by a reaction of gaseous HC1 and Ga metal. Formed gallium chloride is transported into a growth zone of the apparatus where it reacts with ammonia forming GaN. Substrates are located in the growth zone and, if growth conditions including substrate parameters are suitable, single crystal GaN is formed on the substrate. The method provides deposition rates of several microns per minute making it possible to grow hundred microns thick layers.

Growth zone Fig. 1. Schematic illustration for GaN HVPE growth: 1 - main reactor tube, 2 - heating elements, 3 - Ga source gas channel, 4 - boat with Ga melt, 5 - Ga melt, 6 - substrate, 7 substrate holder, and 8 - ammonia source tube.

Due to well recognized potential of GaN materials for light emitters, numerous attempts to create GaN-based blue light emitting diodes by HVPE, including the demonstration of the first GaN violet light emitter,3 have been taken in the 70th. AlGaN alloy growth by HVPE has also been performed. However, despite substantial progress in material quality and process understanding, background n-type carrier concentration in grown

Hydride Vapor Phase Epitaxy: III- Nitrides

3

materials remained high and p-type GaN materials have not been fabricated by that time. Rapid progress in another epitaxial technology, metal organic chemical vapor deposition (MOCVD) for p-type GaN and AlGaN materials in the early 90th5'6 and its ability to form Ga(Al)N/InGaN quantum well structures made this method a technology of choice for the fabrication of GaN-based devices including green, blue, ultra violet (UV), and eventually white light emitting diodes (LEDs).7 The MOCVD was also the first epitaxial method to make GaN-based laser diodes (LDs) and high performance electronic devices. However, many material issues in the field remain unsolved. Due to absence of native GaN and A1N substrates, GaN-based LEDs and HEMTs are currently manufacturing by MOCVD and molecular beam epitaxy (MBE) on foreign substrates, sapphire or silicon carbide. A poor lattice match and difference of thermal expansion coefficient for these substrates usually lead to the formation of threading defects in the epitaxial layers, cracking of layers during the post growth cooling, and residual strains in the epitaxial layers. These defects contribute to high background doping and degrade both electrical and optical properties of the layers and devices. !

10"

V 10'"

1

a sc

K

10°

N

-P

10° 10' ' 10°

**

10z

]

O x

T

10"'

10°

10'

10z

103 10"

Distance from the substrate (urn)

10"" 10" Distance (m)

10"4

Fig. 2. Defect density in GaN layers vs. layer thickness. Left picture: GaN grown by HVPE on sapphire12. Right picture: GaN layers grown by HVPE on 6H-SiC substrate13.

In the 90th, focus of HVPE development for GaN materials shifted to substrate applications8'9. Slow progress and substantial technical obstacles in other crystal growth techniques to form bulk GaN or A1N crystals, put HVPE method on a leading position to produce substrate

4

Dmitriev, Usikov

materials for group III nitride semiconductor devices. High deposition rate intrinsic to GaN HVPE method and ability to form thick single crystal GaN layers on foreign substrates have led to intense developments in the field and to the demonstration of the first large area (3 cm diameter) GaN free standing wafers.10 The later developments of HVPE technology for free standing GaN wafers resulted in GaN materials with a record room temperature electron mobility of 1320 cm2/V sec and donor concentration of 7.8xl015cm"3.H It was demonstrated by several research teams that dislocation density in GaN HVPE layers decreases dramatically with layer thickness increase (Fig. 2).12 For thick GaN layers grown on SiC substrates, dislocation density measured by transmission electron microscopy (TEM) was about lxlO 7 and 4xl05cm"2 at 10 and 100 microns distance from SiC/GaN interface, respectively.13 Growth of thick GaN layers and subsequent fabrication of low defect free standing GaN wafers led to the fabrication of advanced GaN-based power devices' high-frequency transistors, and blue lasers. Currently, HVPE technology is established for commercial production of various types of GaN and A1N substrate materials (Section 5). For more than 30 years of GaN HVPE development various designs of growth apparatuses and process arrangements were described in numerous scientific publications.14'15'16 In this paper we report on our resent results on the fabrication and characterization of GaN, A1N, InN, AlGaN, and InGaN materials and heterostructures by HVPE. We also describe several novel directions in HVPE technology for group III nitrides such as multi-layer sub micron device structures, nano-structures, and large area HVPE growth. 2. Experiment Group III nitride materials described in this paper were grown on proprietary homebuilt HVPE growth equipment. The growth processes were carried out in atmospheric pressure on a multi-wafer HVPE machines having a hot wall quarts tube reactors with a resistively heated multi zone furnaces (Fig. 3). Ammonia (NH3) and hydrogen chloride (HC1) were used as active gases and argon served as a carrier gas. Ga, Al,

Hydride Vapor Phase Epitaxy: III- Nitrides

5

and In metals were used as group III source materials. The metals were located in boats placed in separate channels in the source zone of the growth machine. Substrates were placed on a holder, which got in and out from the growth zone of the rector by a quarts moving rod. Usually we use sapphire or silicon carbide substrates. Substrate capacity of the growth machines was up to seven 2-inch wafers. Size of the substrates ranged from 2 to 6-inch. Typical growth temperature varied from 900 to 1100°C, except for InN grown at lower temperatures (Section 3.7). For n-type doping, silane (SiH4) gas was used. To grow p-type layers, Mg and/or Zn metals were placed in the source zone of the reactor. Growth rate was controlled from 0.05 to 1 microns per min by changing HC1 flow through the corresponding metal source channel. Thickness of grown layers was varied from a few nanometres to several millimetres and was controlled by the growth rate and deposition time. The machine design allows us to grow multi layer hetero structures with thick (>10 |im) and thin (<0.1 |0.m) layers in the same epitaxial process. Metal sources HCI •

J

HCI„ • f SiH 4 Doping^ L Ar NH,

MgorZn

^SKls

\^a^ I

Exhaust

~~ i V = < ^r

Main reactor quartz tube

f^f^l ^

^

5S

Substrates holder

Fig. 3. Schematic of HVPE growth reactor used for experimental work.

3. Material Properties

3.1. Undoped GaN Layers GaN layers having thickness from 2 to 100 jam and 0.1 to 2 jam were grown on 2-inch sapphire and silicon carbide (SiC) substrates,

6

Dmitriev, Usikov

respectively. The growth was performed on (0001) c-plane sapphire or (0001)Si face of 6H- and 4H-SiC substrates. The upper thickness limit reflects cracking problem, which is more sever for SiC substrates.

0

2

4(im

Fig. 4. Optical photo of 20 |im thick GaN layer (left) and AFM image for 100 (xm GaN layer (right) grown on sapphire by HVPE (courtesy of Prof. S. Nikishin).

Surface of grown GaN layers was sooth and mirror like. Typical surface roughness, RMS, even for thick layers grown on sapphire was less than 0.5 nm (Fig. 4). For the smoothest layers, as-grown surface had RMS values below 0.2 nm. Characterization performed using reflectance high energy electron diffraction (RHEED) indicated that as-grown GaN surface is single crystal (Fig. 5).

Fig. 5. RHEED image for as grown GaN surface measured after air exposure.

Crystal structure of grown layers was investigated using x-ray diffraction methods. Typically crystal quality of the layers grown on sapphire improves with layer thickness. For layers thicker than 20 |!m,

Hydride Vapor Phase Epitaxy: III- Nitrides

7

the full width at a half maximum (FWHM) values of x-ray diffraction coscan rocking curves measured for GaN (00.2) and (10.2) reflections were less than 200 arcsec and 300 arcsec, respectively (Fig. 6). Dislocation density measured by transmission electron microscopy (TEM) for GaN grown on sapphire was about (l-3)xl0 8 cm"2 for 10 u.m thick layer.

0

5

10

15

20

25

30

GaN layer thickness (&im)

Fig. 6. The FWHM values of x-ray co-scan rocking curves measured for GaN layers with various thicknesses. The layers were grown by HVPE on sapphire substrates: diamonds correspond to the (00.2) reflections and circles correspond to the (10.2) reflections.

°

1E + 1 5 - I 0

1

1

2

4

Depth

6

(|im)

Fig. 7. SIMS depth profile for H, O, C, and Si impurities in undoped GaN layer grown on sapphire by HVPE.

For GaN layers grown on SiC, x-ray rocking curve FWHM values ranged from about 70 to 200 arc sec and from 200 to 800 arc sec for the (00.2) and the (10.2) reflections, respectively. Etch pit density for about 2 (xm thick GaN layer grown on SiC substrate was in a mid 107 cm"2 range. Crystal structure of thin GaN layers grown on SiC was described in the Ref.17 P-n heterojunctions diodes formed by HVPE grown n-GaN

Dmitriev, Usikov

8

and p-SiC substrates were studied. Properties of n-n and p-n GaN/SiC diodes have also been reported.19 Impurity contamination for undoped GaN layers was studied by means of SIMS. Results of SIMS measurements showed that background impurity concentrations are typically less than 5xl0 16 cm"3, SIMS impurity depth profiles for undoped GaN layer grown on sapphire are given in Fig. 7 show silicon, oxygen, hydrogen, and carbon atomic concentrations of about 5xl0 15 , 2xl0 16 , lxlO17, and 8xl0 15 cm"3, respectively. Concentrations of background metallic impurities were also below 1015 cm"3 level as determined by SIMS and glow discharge mass analysis (Table 1). A concentration of electrically active noncompensated donors, ND-NA, measured for undoped GaN layers typically was below 5xl0 16 cm"3. Data collected for over a year time period for several HVPE growth machine are summarized in Fig. 8. Table 1. Background impurity concentrations measured in undoped GaN layers by glow discharge mass analysis. Element

Concentration Concentration

Element

Concentration Concentration [ppm wt]

[at/cm3]

Zn

<0.01

< 5.7E+14

<7.6E+14

Ge

<0.1

<5.1E+15

0.08

1.1E+16

Se

<0.01

<4.7E+14

P

<0.01

< 1.2E+15

Y

< 0.005

<2.1E+14

Ca

<0.05

<4.6E+15

Mo

< 0.005

< 1.9E+14

Ti

< 0.005

< 3.9E+14

Cd

<0.01

<3.3E+14

V

< 0.005

<3.6E+14

In

<0.01

< 3.2E+14

Cr

< 0.005

< 3.6E+14

Sb

< 0.005

< 1.5E+14

Mn

<0.01

< 6.7E+14

Te

< 0.005

<1.5E+14

Fe

< 0.005

<3.3E+14

W

< 0.005

< 1.0E+14

Co

< 0.005

<3.1E+14

Re

< 0.005

<9.9E+13

Ni

< 0.005

<3.2E+14

Bi

< 0.005

<8.9E+13

[ppm wt]

[at/cm ]

B

< 0.005

< 1.7E+15

Mg

< 0.005

Si

Most values of the concentration ND-NA ranges from 2x10 to 2x10 cm"3 irrespective on the layer thickness. The data points being close to

Hydride Vapor Phase Epitaxy: III- Nitrides

9

1017 cm"3 range were caused probably by some accidental contamination. The ND-NA values in the 1014 cm"3 range may be originated by uncontrollable compensation. In general, for GaN layers thicker than 3 |J.m no dependence of the concentration ND-N A on the layer thickness was detected. On the other hand, material crystal quality improvement with layer thickness is obvious. We may suggest that background electrical active shallow centers in grown undoped GaN are not related to crystal structure defects. Electron mobility and concentration determined by Hall measurements at room temperature for 26 u.m thick undoped GaN layer grown on sapphire was about 760 cm2/V sec and 4xl0 16 cm"3 (Fig. 9), respectively. 1.E+18

-

1.E+17

•; 1.E+16

1.E+15

l.E+14 -, 1.0

10.0

100.0

Thickness (urn)

Fig. 8. Concentrations ND-NA for undoped GaN layers grown on sapphire substrates.

— » • -s

2000

^X

>

XX 1

~-

1000

t 1

O- — B 0 — •

/ / 0

uncorrected corrected for interface layer

""•

1 A

0

100

200

300

T(K)

Fig. 9. Electron mobility measured by Hall method for 26 |im thick GaN layer grown on sapphire. The data are corrected for highly conducting GaN/sapphire interface layer (courtesy of Prof. D. Look).

Optical properties of undoped GaN layers grown on sapphire were evaluated by photoluminescence (PL). For 26 ^m thick GaN layer the

Dmitriev, Usikov

10

near band PL peak had the FWHM of about 1.7 meV at 6 K (Fig. 10). For about 1 p.m thick GaN layers grown on SiC, the near band edge PL peak was located at 3.384eV (300 K) and 3.439 eV (77 K) and had the FWHM of 23 meV and 58 meV at 77 K and 300 K, respectively. 10 7

10 6

3

10 5

°

103

10*

10' 1.8

2.0

2.2

2.4

2.6 2.8 Energy <eV)

3.0

3.2

3.4

Fig. 10. Photoluminescence spectra for GaN layer grown on sapphire substrates. Two lines represent measurements made at two different locations on the sample. The FWHM for the XAD°peak is about 1.73 meV (courtesy of Dr. J. Freitas, Jr.).

3.2. Si-doped GaN Layers Silicon is well known donor in GaN and is widely used to control n-type conductivity of GaN and AlGaN materials. By varying silane gas flow we controlled concentration ND-NA in HVPE grown GaN layers from lxlO 17 to 8xl0 18 cm 3 . Si depth profile measurements were performed by SIMS for HVPE grown epitaxial structure having five GaN layers (Fig. 11). The first and fifth (top) layers are undoped. Three intermediate layers were grown with various Si doping levels. It is clearly seen that Si doped layers have sharp interfaces. Due to small activation energy of Si donor, at room temperature concentration N D - N A is almost equal to electron concentration. Electron Hall mobilities of about 312 and 214 cm2/V sec were measured at 300 K

Hydride Vapor Phase Epitaxy: III- Nitrides

11

for the materials having electron concentration of about lxlO 18 and 4xl0 18 cm"3, respectively. Crystal quality of highly Si doped (>1018 cm"3) GaN layers are slightly deteriorated in comparison to undoped GaN layers. Another issue related to high Si doping is crack formation. Typically thickness for crack free highly Si doped GaN layers is kept bellow 10 \\m. 1E+19

j 1E+08 •a

-r 1E+06 O

2

3

4

Depth (urn)

Fig. 11. Si depth distribution measured by SIMS in an HVPE-grown GaN epi structure comprising Si doped layers. The sample surface is on the left. The Al profile is shown to indicate the GaN-sapphire interface. Initial undoped GaN layer is about 2 um thick.

3.3. Mg-doped GaN Layers n E

1x101!

1x1018 •o CD

A

1x10" a z < z

- ^M

1x10"

1x1016 1x10" 1x1018 1x1019 NA-ND, as-grown wafer (cm"3)

Fig. 12. Concentration NA-ND after annealing vs. concentration NA-ND for as-grown samples: Mg-doped (triangles) and Zn-doped (circles) GaN layers.

Mg doping20 resulted in the demonstration of the first HVPE grown ptype GaN layers and pn junctions. Thickness of p-GaN layers ranged from 3 to 5 jim. As grown Mg doped GaN layers had p-type conductivity

12

Dmitriev, Usikov

with concentration NA-ND up to 2xl0 18 cm"3. After sample anneal at 750°C in argon ambience for 40 min an increase of the concentration NAND in 1.5-3 times was detected (Fig. 12). One possible explanation of high p-type doping measured for as-grown GaN layers is low hydrogen concentration in HVPE grown GaN layers. Hydrogen is known to form H:Mg complex in GaN and post-growth thermal treatment (annealing) is required to withdraw H from the complex and to obtain p-type conductivity for MOCVD grown materials. Note that in MOCVD technology hydrogen is used as a carrier gas and present in large concentration in the growth zone. Besides hydrogen, threading defects can bond Mg atoms that also prevent p-type material fabrication. Developed HVPE process is free of the hydrogen gas (except of small amount of hydrogen produced by ammonia decomposition). Argon is used as a carrier gas in TDI HVPE processes. Thus, good crystalline quality and lower hydrogen concentration explain high p-type conductivity of as grown Mg-doped GaN layers fabricated by HVPE. T 1E+06 •o

c o v1E+04 S

tr1E+02 * 3 O

O 1E+14

1E+00 Depth (|im)

Fig. 13. Mg depth distribution measured by SIMS in an HVPE-grown Mg-doped GaN layer. As grown sample had p-type conductivity. The Al profile is shown to indicate the GaN-sapphire interface. Undoped part of the structure is about 3.5 tim thick.

Good control for Mg doping was proved by SIMS measurements (Fig. 13). The studied structure consists of undoped 3.5 \xm thick GaN layer deposited on sapphire substrate and three subsequently grown Mgdoped GaN layers with increasing doping concentration toward the structure surface. All four layers were grown in the same HVPE run.

Hydride Vapor Phase Epitaxy: III- Nitrides

13

The SIMS profiling demonstrates ability of HVPE technique to fabricate abrupt Mg doping profiles during GaN growth. Recent Mg doping experiments21 resulted in p-type GaN layers showing hole mobility of 80 cm2/V sec at hole concentration of lxlO18 cm"3 at room temperature. These results present record high carrier mobility for p-type GaN and must be further investigated. The FWHM values for XRD rocking curves measured for Mg doped samples were slightly larger than the ones for undoped and Si doped samples ranging from 350 to 410 arc sec and from 600 to 800 arc sec for symmetric (00.2) and asymmetric (10.2) reflections, respectively. Highly Mg doped GaN layers have been applied to form blue and UV LEDs by HVPE technology (Section 5). 3.4. Zn-doped GaN Layers

a

2

1000/T (K"1)

3

Fig. 14. Temperature dependence of electrical resistivity measured in several locations for semi insulating GaN compensated with Zn. The layer was grown on conductive SiC substrate.

GaN doping with Zn during HVPE growth has been widely used for formation of light emitting and photo detector device structures.3 In these structures Zn-doped layers were semi insulating. We have fabricated

14

Dmitriev, Usikov

semi insulating Zn-doped GaN layers grown by HVPE on SiC substrate (Fig. 14) and sapphire substrates. Grown materials demonstrated room temperature electrical resistivity above 1010 Ohm cm.22 Detailed study of Zn doping showed that by controlling Zn vapor pressure in the growth zone it is possible to fabricate three basic strictures: (1) n-type GaN layers compensated with Zn, (2) semiinsulating Zn-doped layers, and, surprisingly, (3) p-type GaN layers doped with Zn. The first p-type GaN doped with Zn has been reported by Polyakov et al.23. For p-type GaN layers doped with Zn a maximum concentration NA-ND of about 6xl0 17 cm"3 was measured (Fig. 12). We believe that p-type doping of GaN with Zn is due to low background impurity concentration and low defect density in HVPE grown layers. More information of two-dimensional hole gas detected in p-type Zn doped GaN/AlGaN structures is given in Section 4. 3.5. AIN Layers HVPE growth of AIN layers is studied substantially less than that for GaN layers. In 1972, Yim and co-workers grew AIN on sapphire substrates in the temperature range 1000-1100°C by HVPE method where A1C13 was generated by passing HC1 over molten aluminum.2 However, results on HVPE growth of AIN are limited.24,25'26 Recently, we demonstrated HVPE growth technology for thick AIN layers. The first free-standing AIN wafers were fabricated by growing up to 1 mm thick AIN layers on silicon substrates and subsequent removal of the substrate by chemical etching.27 Properties of free-standing AIN wafers fabricated by HVPE are given in Section 5. Lately, HVPE technology has been modified to deposit high quality thin AIN and AlGaN epitaxial layers on SiC28' 29 and sapphire substrates. 30 AIN layers were deposited on (OOOl)Si on-axis surface of SiC substrates. Thickness of AIN layer was varied from 0.1 to 0.7 microns. Surface roughness, RMS measured by AFM ranged from 1 to 5 nm. RHEED measurements showed that grown AIN layers have single crystal surface meaning that air exposure does not result in a formation of thick amorphous oxide layer. X-ray diffraction study indicated high

Hydride Vapor Phase Epitaxy: III- Nitrides

15

crystal quality of the grown layers having a minimum FWHM value of co-scan (0002) rocking curve of about 80 arc sec. Specific electrical resistivity of AIN exceeded 1011 Ohm cm at room temperature (Fig. 15). 6-19-1

*

1E+12 r 1E+11

/

0.77 eV

1E+10 r 1E+9 r 1E+8 r

/

1E+7 r

/ . . . . f....

1.0

1.5

i

. . . .

2.0

i . . . .

i . . . .

25 3.0 1000/T(K"')

i . . .

.

3.5

Fig. 15. Temperature dependence of electrical resistivity for semi insulating undoped AIN layer grown on SiC substrate.

/\y^s^~^'

x=i.o 80-

1 °'78 /fcri^/~ / 1 il°J>s ' / 67

60«•a N

3

ji o t: Q

40-

/

// A

/

// /

1

11/

20(1200

250

/

0.46/ j

Z 1 " 3 ? /°- 2 7 /0.11 300

350

f x=0.0 400

Wavelength (nm)

Fig. 16. Optical transmission spectrum for AIN/sapphire, GaN/sapphire and AlxGa^N (0.1 l<x<0.78)/sapphire epitaxial wafers grown by HVPE.

Cathodoluminescence measurements performed at room temperature on thin AIN layers grown on SiC substrates revealed near band edge luminescence peaking at 5.9 eV. Device structures based on HVPE grown thin AIN and AlGaN layers are described in Section 5. Thin (<1 |im) AIN layers with mirror like surface were grown on sapphire substrates. However AIN surface was rougher that that for GaN layers. Surface roughness, RMS for AIN on sapphire ranged from 3 to 15

16

Dmitriev, Usikov

nm. Grown AIN/sapphire samples exhibited good optical transparency in UV spectrum region (Fig. 16). Relatively narrow x-ray (00.2) AIN (0-scan rocking curves with the FWHM values of about 80 arc sec were measured for several samples, but for such samples the asymmetric (10.2) x-ray rocking curves were much wider with the FWHM values exceeding 1500 arc sec. The narrowest (10.2) x-ray co-scan rocking curve measure so far for thin AIN layer grown on sapphire had the FWHM of 310 arc sec with the FWHM for the (002) peak of 177 arc sec. Currently, HVPE process for low defect thin AIN layers on sapphire is under development. Thick (> 5 microns) crack-free AIN layers were grown on 2-inch and 3-inch SiC and sapphire substrates by stress control HVPE method31. Thickness of grown AIN layers ranged from 1 to 20 microns for material grown on sapphire and from 5 to 75 microns for material grown on SiC. Both 6H- and 4H-SiC wafers were used as substrates. The layers were deposited on (0001)Si face of SiC wafers and on (0001) c-plan of sapphire wafers. Surface of AIN layers had characteristic pyramidal relief with RMS roughness ranging from 20 to 70 nm. After polishing, AIN surface had roughness less than 0.5 nm. The minimum FWHM values for the (002) AIN co-scan x-ray rocking curves were about 250 arc sec and 500 arc sec for layers grown on SiC and sapphire substrates, respectively. Wet chemical etching of thick AIN layers grown on SiC resulted in etch pit density ranging from mid 107 cm"2 to low 108 cm"2. 2inch diameter AIN free standing wafers were fabricated by removing SiC substrate by reactive ion etching (Section 5). 3.6. AlGaN Layers Initial results on HVPE grown AlGaN alloys were reported in the 70 th4 . In that study Ga and Al were transported to a reaction zone as subchlorides by passing HC1 diluted with nitrogen gas over a mixed melt of Ga and Al in a graphite boat. Sapphire substrates had the (21-31) orientation resulted in (11-20) AlGaN films. Composition of the grown layers was varied from 0 to 100 mol. % AIN by controlling Ga to A] ratio in the source material. Cathodoluminescence and optical absorption of grown materials were studied for entire composition range. Electrical

Hydride Vapor Phase Epitaxy: III- Nitrides

17

properties of HVPE grown AlGaN alloy layers having AIN content up to 45 mol. % were studied33 showing background electron carrier concentration in the 1019 cm"3 range, which decreased as AIN content increases. Recently, up to 75 |am thick AlGaN layers were grown on sapphire by HVPE using mixed Ga+Al sources.34 To deposit AlGaN layers by HVPE we used separate Ga and Al metal sources. The whole composition range of AlGaN alloys was covered by controlling a ratio of aluminum and gallium chlorides partial pressures in the growth zone (Fig. 17). Composition of AlGaN layers was determined using x-ray diffraction data assuming that Vegard's law is valid. The experimental uncertainty in that determination of the AlxGai_xN composition was estimated to be x -0.02. Composition for grown alloy epitaxial layers was reasonably uniform across the wafer (Fig. 18). 1

00

1— I

|

-i

1

r— |

-i

1

80 60

11 .

40 20

•

^«»-i-.~r 0

0.2

0.4

•—i—,—i—,—i 0.6

0.8

1

[AI]/[Ga]+[AI] ratio in vapor phase

Fig. 17. Vapor phase-solid phase composition diagram for HVPE grown AlGaN alloys.

Best 2-inch wafers had standard deviation for AIN composition of about 1 mole % of AIN. Typical thickness of crack-free AlGaN layers grown on sapphire substrates ranged from 0.3 to 2 urn. The FWHM of XRD rocking curves changes with AlGaN composition attaining the lowest values of 350 arcsec and 1400-1800 arcsec for the (00.2) symmetric and the (10.2) asymmetric reflections, respectively for high range of AlxGai_xN composition (0.65<x<0.85). Minimum estimated screw and edge dislocation densities in AlGaN layers were <6xl0 8 cm"2 and <2xl0 9 cm"2, respectively. For undoped AlxGa]_xN layers grown on sapphire, background concentration ND-NA depended on alloy

18

Dmitriev, Usikov

composition x. For x < 0.3, AlGaN layers show n-type conductivity and concentration ND-NA is in 1017 - 1018 cm"3 ranges. Hall measurements showed electron mobility of about 166 cm2/V sec and electron concentration of 9xl0 17 cm"3 (300 K) for alloy having 18 mol. % of A1N. For x values near 0.5 or larger, AlxGai_xN materials were almost insulating and correct C-V measurements were not possible.

Fig. 18. Photoluminescent map for 2-inch AlGaN-on-sapphire epitaxial wafer. Standard peak lambda deviation is 0.65%. Average AlGaN composition is about 16 A1N mole % (courtesy of Accent Optical Technologies).

Normalized optical transmission spectra of crack-free HVPE-grown AlxGaj.xN layers are shown in Fig. 16. The interference fringes appeared on the transmission curves at the wavelengths longer than cut-off wavelength indicate smooth surfaces of these AlGaN layers. At 265 nm optical transmission of 71% and 73% was measured for AlxGai_xN layers having composition of x~0.73 and x~0.78, respectively. Energy gap, Eg, obtained from transmission measurement found to correlate with the PL peak position that can be ascribed to near-band-edge recombination in the layer grown. The transmission and PL data are collected together in Fig. 19 to plot optical band-gap energy of AlxGaLXN alloys as a function of composition x (A1N mole fraction). The Eg values for undoped GaN layer defined from absorption and PL measurements, corresponded well to each other within +10 meV accuracy range, were equal to 3.41 eV. For A1N band-gap energy was found to be 6.13 eV. Dotted line is linear interpolation between EgGaN

Hydride Vapor Phase Epitaxy: III- Nitrides

19

(3.4 eV) and EgAiN (6.1 eV). Solid line was calculated using Eg=xEgAiN+(I-x)EgCaN-bx(l-x) equation with bowing parameter b=l.l eV. Refractive index of AlGaN layers grown by HVPE was studied in the Ref.35 Undoped AlGaN layers grown by HVPE on SiC substrates were described in ref.36'37 Doping of AlGaN alloys grown by HVPE is not investigated in details yet. The fact that AlGaN/AlGaN p-n heterostructures are working as 300 nm UV light emitters38 indicates that p-type doping is possible at least up to 27 mol. % of A1N. We may conclude that A1N and AlGaN epitaxial layers can be grown by HVPE with a reasonable composition control and crystal quality. However, HVPE growth of AlGaN and A1N layers is investigated much less than HVPE growth of GaN. 7.0 6.0

I 5.0 LU

4.0 3.0 0.0

0.2

0.4

0.6

0.8

1.0

AIN mole fraction

Fig. 19. Dependence of AlxGa].xN optical band gap on AIN mole fraction (x) derived from room temperature photoluminescence (triangles) and optical transmission data (circles).

3.7. InN and InGaN Layers Results on HVPE growth of InN and InGaN materials are limited. The first deposition process was performed more than 25 years ago when InN epitaxial layers were grown using reaction between ammonia and InCl3 39. InCl3 was synthesized in a separate experiments from In and CI2. The layers were deposited on sapphire and quartz substrates. Deposition rate ranged from 1 to 8 um per hour depending on growth conditions. Source temperature and substrate temperature were controlled from 200 to 600°C and from 450 to 700°C, respectively. The maximum growth rate was observed at 600°C. Grown materials had n-type conductivity with

Dmitriev, Usikov

20

electron concentration from 2xl0 20 cm"3 to 8xl0 21 cm"3 and mobility 35 50 cm2IN s. Later several research teams reported HVPE growth of InNcontaining materials40,41. Thermodynamics of HVPE growth process for InGaN layers was analysed in the Ref.42. InN growth using InCl and InCl3 sources was reported43,44 InGaN alloy layers were grown employing InCl3, GaCl3 and NH3 sources.45 When InCl-NH3 material system was used, very low growth rate (<0.05 p.m/hr) was observed. In contrast, use of InCl3-NH3 material source system resulted in InN growth rate up to 0.3 ixm/hr. The growth rate obtained at 750°C was about 0.2 ixm/hr. The minimum FWHM value of x-ray rocking curve of about 24.7 arc min was measured for InN grown at 700°C. Growth of InGaN thin films by HVPE was reported in ref.46. In that work, separate sources of metal Ga and In were place in a single HC1 gas channel to form gallium and indium chlorides. The composition of alloy InGaN layers was controlled by varying the ratio of the effective metal surface areas in the range from 0.3 to 35 (fn/Ga). '

'

' ,'

InN

1

I'

i •*

/

: •

HVPE InN

A /

t

>

/

wm

•

n

A%n

.^4*

-JK*n

»

~

Fig. 20. The energy of the optical absorption edge as a function of the free electron concentration in InN material grown by different techniques (after J. Wu, W. Walukiewicz 48 ). Position of TDI InN HVPE grown material is shown.

Recently, our team has reported novel results on HVPE growth of InN materials and heterostructures.47 The growth was performed on 2-

Hydride Vapor Phase Epitaxy: III- Nitrides

21

inch sapphire substrates and also on GaN/sapphire, AlN/sapphire, and AlGaN/sapphire templates (Fig. 20). Growth temperature was varied from 500 to 700°C. For continuous InN layers, thickness ranged from 0.3 to 2 microns. X-ray diffraction rocking curves measured usingcoscanning geometry for the (00.2) InN reflection had the FWHM of about 460 arc sec for continues InN layers. It was observed that depending on growth conditions, InN materials may be grown as continues layers or nano-size structures. The nanostructures had shape of nano-rods or nano-wires with a typical diameter of about 60 - 500 nm and length ranging from a few hundred nanometers to a few microns (Fig. 21). Investigation of InN HVPE process and materials is under way.

Fig. 21. SEM photographs of GaN nano wire (left) and nano-size InN structures (right) grown on sapphire (courtesy of Dr. A. Davydov and Prof. S. Nikishin). Image width is about 2 u.m for top picture and about 16 u.m for bottom picture.

4. New Directions in HVPE Development

4.1. Large Area and Multi Wafer HVPE Growth Novel HVPE growth equipment developed at TDI, Inc. allowed us to grow GaN, A1N, and AlGaN layers and multi-layer structures on large area (>2-inch) substrates, for the first time. GaN layers grown on 4-inch sapphire substrates showed material parameters similar to that for 2-inch diameter samples. Interestingly, undoped GaN layers grown on 4-inch

22

Dmitriev, Usikov

sapphire showed some reduction of ND-NA concentration being in high 1014 cm"3 to low 1015 cm"3 range. Typical standard thickness deviation for GaN layers grown on 4-inch substrates was about 5 %. Uniform n-type doping was achieved for Si-doped GaN layers grown on 4-inch sapphire with electrical resistivity uniformity better than 5 %. Recently, A1N and GaN layers were successfully grown on 6-inch diameter Si and sapphire substrates, respectively (Fig. 22). These results demonstrate ability of HVPE process and equipment to handle large size epitaxial structures.

Fig. 22. Photoluminescence map for 6-inch GaN-on-sapphire epitaxial wafer (courtesy of Accent Optical Technologies). Standard deviation of peak PL wavelength for the whole wafer is 0.2 %.

Fig. 23. Thickness of seven AlGaN layers grown on 2-inch sapphire substrates in the same HVPE process. The layer thickness was measured in the wafer centre.

Hydride Vapor Phase Epitaxy: III- Nitrides

23

Another direction of the development of HVPE production technology is design of multi wafer HVPE growth machines and processes. Such machines existing for HVPE growth of other compound semiconductor have not been developed for GaN materials until recently. TDI, Inc. is developing and putting in operation HVPE machines with large wafer capacity to reduce manufacturing cost. Novel HVPE equipment is capable to carry out simultaneous growth seven 2-inch (7x2") wafers (Fig. 23), or 3x3" wafers, or 2x4" wafers and further scaling of the equipment and process is under way. 4.2. Multi-lmci Strut lurc\

Fig. 24. SEM image of cleaved edge of multi layer AlGaN/GaN epitaxial structure grown on sapphire substrate. Dark regions and light regions corresponds GaN and AlGaN layers, respectively.

Multi layer HVPE growth of GaN-based structures has been used to fabricate M-I-S light emitting structures in the 1970th.49 Recently, novel avenue for nitride HVPE has been opened by the demonstration of controllable growth of nm-thick layers and fabrication multi-layer structures. Growth rate of GaN and AlGaN layers was controlled below 2 ixm/hr making it possible to fabricate multi layer structures with submicron layers. Equipment and process were modified to grow

24

Dmitriev, Usikov

GaN/AlGaN heterostructures with sharp interfaces. One example of such multi-layer AlGaN/GaN is demonstrated in Fig. 24. Thickness of different AlGaN and GaN layers in this structure was controlled from about 20 to 260 nm and 50 - 250 nm, respectively, by varying growth time for each layer. 1E+006

Al 0 2 2 GaN/Al 0 1 6 GaN/GaN/AI„ 1 6 GaN/AI 0 double confined heterostructure

1E+005

—

2 2 GaN/SiC

lllllllll|IIIIIIMI|IIIIIMII|IMIIIIII|MIIIIIII|lllllllll|IIIM(lll

0

0.1

0.2

0.3

0.4

O.S

0.6

Depth (microns)

Fig. 25. SIMS element depth profiles for AlGaN/GaN/AlGaN double heterostructure grown by HVPE on SiC substrate.29

SMS depth analysis of another HVPE grown AlGaN/GaN/AlGaN double heterostructure is given in Fig. 25. GaN well was of about 20 nm thick. Stimulated emission under optical pumping conditions was observed from this structure at room temperature at a wavelength of 370 nm with a threshold of about 540 kW/cm2.29 4.3. P-n Junctions Two-layer GaN pn-structures were fabricated by in-situ p-type doping during HVPE growth.20 First, undoped GaN layer was grown directly on 6H-SiC substrate. Than Mg was introduced as an acceptor dopant to grow p-type GaN layer. The Mg atomic concentration ranged for 20 different samples from 5x10 to 5x10 cm". Mesa diodes with a

Hydride Vapor Phase Epitaxy: III- Nitrides

25

vertical current flow geometry were formed by reactive ion etching. Current-voltage and capacitance-voltage characteristics of the p-n diodes indicating good rectifying properties were measured. 300 'o 200 CO

80 90100

300

Temperature, K i

18

' g 8x10 o

.2 7x1018 CO

A

•

a

1 * fat>£«-*«

98

Aft<

Ml

4r

1 e•

> ffi * 1 A

9

§ 6x10 18 C

oo 5x1018

* i

6

,

1I

i

i

8 10 1 1000/T, K

1

12

Fig. 26. Temperature dependencies of hole mobility (top) and hole concentration (bottom) measured for p-type GaN/AlGaN hetero structure with 2DHG.21 Triangles and dots represent data measured at different locations on the sample.

The electroluminescence from the p-n diodes peaking at 420 nm was detected. These results have demonstrated an ability of HVPE

26

Dmitriev, Usikov

technology to form GaN-based structures for optoelectronic devices (Section 5). 4.4. Structures with Two Dimensional Carrier Gas Another example of multi layer structures grown by HVPE is AlGaN/GaN heterostructures with two dimensional electron gas (2DEG).50 These structures were grown on sapphire substrates. Thickness of GaN and AlGaN layers were 3 |J.m and 25 nm, respectively. A1N concentration in the alloy layers was about 25 mol. %. The structures exhibited room temperature electron mobility of about 800 cm2/V sec at sheet charge density of about lxlO 13 cm"2. At low temperature electron mobility up to 4000 cm2/V sec was measured. For p-type GaN/AlGaN structures formation of two dimensional hole gas (2DHG) was also detected (Fig. 26).21 Hole mobility of 120 cm2/V sec at hole concentration of 6x101S cm"3 was measured at room temperature. Investigation and further development of epitaxial structures containing 2DHG may substantially benefit fabrication of GaN-based bipolar devices. 4.5. Nano Structures and Porous Materials In this chapter we are focusing on continues epitaxial layers and structures grown by HVPE. However, we would like to mention a fabrication of nanowires and nano rods by HVPE. Recently, GaN51 and InN47 nano structures (Fig. 21) have been demonstrated. Porous nanosize GaN materials were grown from single crystal continues GaN layers by wet anodization.52'53 GaN growth on porous substrates resulted in stress and defect density reduction in grown epitaxial layers.54 The nano size materials have an important meaning for future device applications. Novel directions in group III nitride HVPE developments also include epitaxial overgrowth experiments to reduce defect density and GaN growth on non polar planes of sapphire substrates. However, these very interesting areas are out of scope of this chapter.

Hydride Vapor Phase Epitaxy: III- Nitrides

27

5. Applications of HVPE Grown Group III Nitride Materials

5.1. Substrate Applications

5.1.1. Template substrates The idea of template substrate comprising thin (from 0.5 to 100 microns) nitride layer on a foreign material, for example 10 (xm thick GaN on 2inch sapphire, has been discussed in numerous publications.55' 56' 57 Template substrates provide native GaN (AIN or AlGaN) surface for subsequent epitaxial growth. Buffer layer is not needed to form high quality nitride epitaxial layer/structure on template substrates. Various types of template substrates have been discuss in the literature include (1) low defect from 5 to 20 |im thick n-type GaN-on-sapphire templates for high-power, high-brightness blue, green, and white LEDs, (2) p-type GaN-on-sapphire templates, (3) UV transparent AlGaN-on-sapphire templates for UV LEDs, (4) semi insulating i-GaN-on-sapphire templates for electronic devices, and (5) AIN-on-sapphire and AIN-on-SiC templates for high power high electron mobility transistors (HEMTs). HVPE is a method of choice to fabricate template substrates due to high growth rate, low defect density in grown materials, and relatively low cost of the technology. Growth of group III nitride materials on template substrates was investigated using HVPE,58 MOCVD59 and MBE60'61 techniques. Defect density reduction and material quality improvements in homoepitaxial GaN layers grown on GaN template substrates have been demonstrated. 30 Template substrates have been successfully used for group III nitride optoelectronic and electronic32'62'63 devices including blue LEDs on GaN/SiC templates64 and GaN/sapphire templates.57 Ultra violet (UV) AlGaN-based LEDs have been fabricated on UV transparent AlGaN/sapphire templates.66 Laser diodes and high quality LEDs were made on HVPE grown templates with low defect density reached due to lateral overgrowth approach.67 HVPE grown semi insulating GaN/sapphire and AlN/SiC templates have been used as substrates for

28

Dmitriev, Usikov

high frequency GaN-based HEMT devices.68 Fabrication of template substrates by HVPE is technically and economically attractive due to high deposition rate and relatively low defect density in thick HVPE grown epitaxial layers. Definite advantage of template substrates is a possibility to use large size initial substrate such as 4-inch sapphire or 6inch Si. The main technical challenges are relatively high defect density, cracking and substantial bowing of template substrates containing thick nitride layers. Some promising results on the bow control of AIN layers grown by HVPE on foreign substrates have been recently achieved by stress control HVPE method.31 5.1.2. Free-standing substrates 30 mm diameter GaN free standing wafers have been reported in 1998.10 Thick (up to 200 ixm) GaN layers were grown by HVPE on 6H-SiC substrates. The SiC substrate was removed by reactive ion etching resulting in freestanding GaN wafer. These wafers exhibited very low defect density measured by TEM. Dislocation density was lxlO 7 cm"2 and 4xl0 5 cm" for material located at 10 microns away and 100 microns away from GaN/SiC interface, respectively (Fig. 2).13 Free standing GaN wafers have also been fabricated using HVPE growth of thick GaN layers on sapphire,70' 71 LiGa0 2 substrates,72 and GaAs substrates.73 Typical dislocation density in GaN free standing wafers is in 106 - 107 cm"2 range. These GaN materials exhibited record high electron mobilities and very good optical properties. Properties of freestanding GaN wafers and their applications are described in numerous publications.65'69-74"78 AIN27 and AlGaN79 freestanding wafers have also been reported. AIN free standing wafers were fabricated using Si and SiC initial substrates. Thick AIN layers were grown by HVPE and initial substrates were removed after HVPE growth. Mechanical80, electrical81, and structural13 properties of free standing AIN crystals have been reported. AIN materials fabricated by HVPE using silicon initial substrates demonstrated record high thermal conductivity of 3.3 WK^cm"182 and high electrical resistivity (>10 n Ohm cm, 300 K).81 However, due to

Hydride Vapor Phase Epitaxy: III- Nitrides

29

significant lattice and thermal mismatch to initial Si substrate, dislocation density in grown AIN layers exceeded 108 cm"2 and corresponding x-ray rocking curves were relatively wide, of about 40 arc min for the rocking curves measured using co-scanning geometry for the (0002) AIN reflex.13 Recently, 2-inch diameter freestanding AIN wafers have been fabricated using SiC initial substrates31, which were removed by reactive ion etching after HVPE deposition of thick AIN layers. Free standing AlGaN wafer79were fabricated by growing thick (up to 0.6 mm) undoped AlGaN layer on SiC substrate and subsequent removal of the SiC substrate by reactive ion etching. AIN concentration in fabricated AlGaN wafers was estimated of about 35 mol. %. The wafers demonstrated n-type conductivity with electron concentration in the 1017 era"3 range. Currently, free-standing technology is the only commercial method to fabricate 2-inch GaN substrates with defect densities below 107 cm"2. Laser diodes83' 84and LEDs have been fabricated on GaN free standing wafers.85 High voltage M-I-M device structures were fabricated using free standing AIN wafers.81 However, due to the fact that every wafer is fabricated from a thick epitaxial layer grown on a foreign substrate and, as a result, fabricated material is highly stressed and contains defects originated by heteroepitaxial growth, this approach has significant technical and economical limitations. 5.1.3. Bulk substrates Recently HVPE-based bulk growth technologies were proposed to fabricate bulk GaN and AIN crystals.86 The method involves growth of GaN boule on GaN seed, boule slicing, and wafer polishing (Fig. 27). High quality GaN homoepitaxial layers (Fig. 28) with defect density in the 106 cm"2 range and blue LEDs were fabricated on bulk GaN wafers. Although the bulk approach has very good potential for cost effective manufacturing of low defect large size GaN substrates, current size of demonstrated GaN boule is small and substantial development for this technology is needed. Bulk GaN technology based on HVPE approach is under development at several organizations.87'8S Bulk AIN substrates were fabricated from AIN boules grown by a vapor transport technique based on HVPE technology.89 AIN wafers up

30

Dmitriev, Usikov

to 1.75-inch diameter were sliced from single crystal AIN boules. Specific resistivity of the AIN bulk material was evaluated in the temperature range from 250 to 650 K. Material electrical resistivity exceeds 10g Ohm cm at room temperature and 106 Ohm cm at 500 K.

Fig. 27. Photo of 35 mm diameter GaN boule under diamond wire slicing. The boule was grown by modified HVPE technology at TDI, Inc.

Fig. 28. PL spectra for 2 u.m thick GaN homoepitaxial layer grown by HVPE on bulk GaN substrate (courtesy of Dr. J. Freitas, Jr.). The FWHM of the PL peak is below 0.5 meV.

True bulk technical approach involving boule growth on a native seed has substantial potential for material quality improvement by selecting the best quality seed for each subsequent growth run. However,

Hydride Vapor Phase Epitaxy: III- Nitrides

31

many technical issues, including relatively low growth rate and longterm process and equipment instability, have to be resolved before this technique will be commercially valuable. 5.2. Device Applications Multi layer structures grown by HVPE were used for the fabrication of several types of GaN-based devices including electronic and optoelectronic devices, HEMTs90 and LEDs.91' 92, 38' 93 For all these devices, the whole device epitaxial structure was grown by HVPE technology. Transistors were fabricated based on AlGaN/GaN90, 94 epitaxial structures grown by HVPE. The HEMT structure consisted of a 2 u,m thick undoped GaN layer grown on sapphire, followed by a 30 nm thick Alo.22Gao.7gN layer. Fabricated devices with 1 um gate length showed transconuctances in excess of 110 mS/mm and drain-source current above 0.6 A/mm. A current gain cut-off frequency, ft, of 11.5 GHz and a maximum frequency of oscillation, f^,, of 20.5 GHz were achieved for 1x100 |j.m2 device at a gate voltage of 0 V and drain voltage of 2 V (Fig. 29).95 30 U,

11.5 GHz 20.65 GHz

Frequency (GHz) Fig. 29. RF frequency response of a 1x100 u.m2 gate dimension HVPE AlGaN/GaN HEMT measured at Vg = 0 V and Vds = 3 V96.

Blue,92 ultra violet (Fig. 30),38'93 and white LEDs were fabricated based on AlGaN/GaN and AlGaN/AlGaN p-n heterostructures grown by

32

Dmitriev, Usikov

HVPE. Peak wavelength for HVPE blue LEDs ranged from 415 to 420 nm. Packaged LEDs had output power and external efficiency up to 1.4 mW and 2.5 %, respectively. These numbers represent highest external efficiency reported for In-free LED structures emitting in the blue-violet spectral region. Recently, HVPE grown blue LEDs have been used for sign fabrication .96 Energy (ev)

5x10"5

-.

4x1

°5

i

3x10 5

I 2x10-5 0)

_c

1x10"5 0 300

350

400

450

500

Wavelength (nm) Fig. 30. EL spectra at various currents measured for AlGaN/AlGaN double heterostructure UV LED grown on sapphire by HVPE at TDI, Inc. Peak emission is at 331 nm.

AlGaN/AlGaN double heterostructures for UV LED were grown by HVPE on sapphire substrates. Peak wavelength for HVPE grown UV LEDs was controlled from 300 to 350 nm by varying AIN content in the light-emitting region. For packaged 340 nm LED (Fig. 28) 2 mW output power was measured at 110 mA pulsed current.38 6. Conclusions The technology of group III nitride epitaxy and bulk growth has shown grate progress in the last decade. It appears that HVPE is the current choice of technology for GaN and AIN substrate materials including template, free standing, and bulk substrates. Resent demonstration of 6inch GaN and AIN HVPE technology and multi wafer growth HVPE equipment are important steps toward large area low cost substrate

Hydride Vapor Phase Epitaxy: III- Nitrides

33

materials. However, defect densities are still higher than 105 cm"2 and growth rate itself is below 1 mm/hr required for true bulk crystal growth. New achievements in the HVPE field include fabrication of HVPE grown multi layer device structures with sharp interfaces, LEDs and HEMTs. The technology progressing very rapidly and we expect dramatic progress to be achieved in a few years in both scientific and industrial aspects of group III nitride HVPE technology. Acknowledgments Authors would like that thank O. Kovalenkov, V. Soukhoveev, Y. Shapovalova, V. Ivantsov, A. Syrkin, V. Maslennikov, Yu. Melnik, A. Nikolaev, A. Cherenkov, and V. Syzov for their priceless input into technological developments described in this paper. Experimental results described in this paper have been achieved in collaboration with a large number of academic, government, and industrial organizations. Authors are very grateful to all colleagues who supported this work. Materials characterizations performed by M. Reshchikov, S. Nikishin, J. Freitas, C. Collins, M. Wraback, M. Kneissl, N. Johnson, N. Shmidt, V. Davidov, A. Davydov, N. Kuznetsov, T. Dang, D. Thomson, and D. Look made these developments possible. We thank Lehighton Electronics, Inc Candela, and Accent Optical Technologies for their help in material investigation. Encouraging and governing discussions with C. Wood, M. Yoder, and J. Carrano and their support were very important for this work. Financial support for HVPE development at TDI was provided by the Department of Defense, Department of Energy, and Department of Commerce. References 1. H.P. Maruska and J.J. Tietjen, The preparation and properties of vapor-deposited single-crystalline GaN, Appl. Phys. Lett. 15, 10, 327-329 (1969). 2. W.M. Yim, E.J. Stofko, P.J. Zanzucchi, J.I. Pankove, M. Ettenberg, and S.L. Gilbert, Epitaxially grown A1N and its optical band, J. Appl. Phys. 44, 1, 292-296 (1973). 3. J.I. Pankove, E.A. Miller, D. Richman, and J.E. Berkeyheiser, Electroluminescence in GaN, Journal of Luminescence, 4, 63-66 (1971). 4. J. Hagen, R.D. Metcalfe, D. Wickenden, and W. Clark, Growth and properties of Ga x M.xN compounds, J. Phys. C: Solid State Phys. 11, L143-L146 (1978).

34

Dmitriev, Usikov

5. H. Amano, M. Kito, K. Hiramatsu and I. Akasaki, P-Type Conduction in MgDoped GaN Treated with Low-Energy Electron Beam Irradiation (LEEBI), Jap. J. Appl. Phys. 28, N12, pp. L2112-L2114 (1989). 6. J S. Nakamura, T. Mukai and M. Senoh, High-Power GaN P-N Junction BlueLight-Emitting Diodes, Jap. J. Appl. Phys. 30, N12A, pp. L1998-L2001 (1991). 7. S. Nakamura and G. Fasol, The Blue Laser Diode - GaN-Based Light Emitters and Lasers, Heidelberg: Springer-Verlag (1997). 8. K. Naniwae, S. Itoh, H. Amano, K. Itoh, K. Hiramatsu, and I. Akasaki, Growth of single crystal GaN substrates using hydride vapor phase epitaxy, J. Cryst. Growth 99,381-384(1990). 9. S.T. Kim, Y.J. lee, D.C. Moon, C.H. Hong, T.K. Yoo, Preparation and properties of free-standing HVPE grown GaN substrates, J. Cryst. Growth 194, 37-42 (1998). 10. Yu. Melnik, A. Nikolaev, I. Nikitina, K. Vassilevski, V. Dmitriev, Properties of free-standing GaN bulk crystals grown by HVPE, Mat. Res. Soc. Symp. Proc. Vol. 482, 269-274 (1998). 11. D.C. Look, J.R. Sizelove, J. Jasinski, Z. Liliental-Weber, K. Saarinen, S.S. Park, and J.H. Han, Electrical, Optical, Structural, and Analytical Properties of Very Pure GaN, Mat. Res. Soc. Symp. Proc. Vol. 743, 575-590 (2003). 12. A. Usui, Bulk GaN crystal with low defect density grown by hydride vapor phase epitaxy, Mat. Res. Soc. Symp. Proc. Vol. 482, 233-244 (1998). 13. M. Albrecht, LP. Nikitina, A.E. Nikolaev, Yu.V. Melnik, V.A. Dmitriev, and H.P. Strunk, Dislocation Reduction in A1N and GaN Bulk Crystals Grown by HVPE, Phys. Stat. Sol. (a), 176, 453-458 (1999). 14. L. Liu, J.H. Edgar, Substrates for gallium nitride epitaxy, Material Science and Engineering R 37, 61-127 (2002). 15. O. Ambacher, Growth and applications of Group Ill-nitrides, J. Phys. D: Appl. Phys. 31, 2653-2710 (1998). 16. S.Yu. Karpov, A.S. Segal, D.V. Zimina, S.A. Smirnov, A.P. Sid'ko, A.V. Kondratyev, Y.N. Makarov, D. Martin, V. Wagner, and M. Ilegems, Transport and Chemical Mechanisms in GaN Hydride Vapor Phase Epitaxy, Mat. Res. Soc. Symp. Proc. Vol. 743 (2003) 17. A. Kazimirov, N. Faleev, H. Temkin, M.J. Bedzyk, V. Dmitriev, Yu. Melnik, Highresolution x-ray study of thin GaN films on SiC, J. Appl. Phys. 89, 11, 6092-6097 (2001). 18. A.E. Nikolaev, S.V. Rendakova, LP. Nikitina, K.V. Vassilevski, and V.A. Dmitriev, GaN Grown by Hydride Vapor Phase Epitaxy on p-Type 6H-SiC layers, J. Electron. Materials 27, 288-291 (1998). 19. E. Danielsson, C-M. Zetterling, M. Ostling, A. Nikolaev, LP. Nikitina, and V. Dmitriev, Fabrication and Characterization of Heterojunction Diodes with HVPEGrown GaN on 4H-SiC, IEEE Transactions on Electron Devices, 48, 444-449 (2001).

Hydride Vapor Phase Epitaxy: III- Nitrides

35

20. A.E. Nikolaev, Yu.V. Melnik, N.I. Kuznetsov, A.M. Strelchuk, A.P. Kovarsky, K.V. Vassilevski, and V.A. Dmitriev, GaN pn-structures grown by hydride vapor phase epitaxy, Mat. Res. Soc. Symp. Proc. Vol. 482,251-256 (1998). 21. A. Usikov, O. Kovalenkov, V. Ivantsov, V. Soukhoveev, V. Dmitriev, N. Shmidt, D. Poloskin, V. Petrov, V. Ratnikov, P-type GaN epitaxial layers and AlGaN/GaN heterostructures with high hole concentration and mobility grown by HVPE, Mater. Res. Soc. Symp. Proc. Vol. 831,453-457 (2005). 22. N.I. Kuznetsov, A.E. Nikolaev, A.S. Zubrilov, Yu.V. Melnik, V.A. Dmitriev, Insulating GaN:Zn layers grown by hydride vapor phase epitaxy on SiC substrates, Appl. Phys. Lett. 75, 3138-3140 (1999). 23. A.Y. Polyakov et al., Studies of deep centers in high-resistivity p-GaN films doped with Zn and grown on SiC by hydride vapor phase epitaxy, Solid-State Electronics 45, 249-253 (2001). 24. I. Akasaki and M. Hashimoto, Infrared lattice vibration of vapour-grown AIN, Solid State Communications, 5, 851-853 (1967). 25. M.P. Callaghan, E. Patterson, B.P. Richards and C.A. Wallace, The growth, crystallographic and electrical assessment of epitaxial layers of aluminum nitride on corundum substrates, J. Crystal Growth 22, 85-98 (1974). 26. T. Yamane, H. Murakami, Y. Kangawa, Y. Kumagai, and A. Koukitu, Growth of thick AIN layer on sapphire (0001) substrate using hydride vapor phase epitaxy, Phys. Stat. Sol. (c) 2, 2062-2065 (2005). 27. A. Nikolaev, I. Nikitina, A. Zubrilov, M. Mynbaeva, Yu. Melnik, V. Dmitriev, AIN wafers fabricated by hydride vapor phase epitaxy, Mat. Res. Soc. Symp. 595, W6.5.1-W6.5.5(2000). 28. Yu. Melnik, D. Tsvetkov, A. Pechnikov, I. Nikitina, N. Kuznetsov, V. Dmitriev, Characterization of AIN/SiC epitaxial wafers fabricated by hydride vapour phase epitaxy, Phys. Stat. Sol. (a) 188, Nl, 463-466 (2001). 29. D. Tsvetkov, Yu. Melnik, A. Davydov, A. Shapiro, O. Kovalenkov, J.B. Lam, J.J. Song, V. Dmitriev, Growth of submicron AlGaN/GaN/AlGaN heterostructures by hydride vapor phase epitaxy (HVPE), Phys. Stat. Sol. (a) 188, Nl, 429-432 (2001). 30. A.S. Usikov, Dae-Woo Kim, A.I. Pechnikov, Yu.V. Ruban, M.A. Mastro, Yu. Melnik, V.A. Soukhoveev, O.V. Kovalenkov, G.H. Gainer, S. Mahajan, V.A. Dmitriev, Material quality improvements for homoepitaxial GaN and AIN layers grown on sapphire-based templates, Phys. Stat. Sol. C, N7, 2580-2584 (2003). 31. O. Kovalenkov, V. Soukhoveev, V. Ivantsov, A. Usikov, V. Dmitriev, Thick AIN layers grown by HVPE, J. Cryst. Growth 281, 87-92 (2005). 32. J.R. LaRoche, B. Luo, F. Ren, K.H. Baik, D. Stodilka, B. Gila, C.R. Abernathy, S.J. Pearton, A. Usikov, D. Tsvetkov, V. Soukhoveev, G. Gainer, A. Pechnikov, V. Dmitriev, G.-T. Chen, C.-C. Pan, J.-I. Chyi, GaN/AlGaN HEMTs grown by hydride vapor phase epitaxy on AIN/SiC substrates, Solid State Electronics 48, 193-196 (2004).

36

Dmitriev, Usikov

33. B. Baranov, L. Daweritz, V.B. Gutan, G. Jungk, H. Neumann, and H. Raidt, Growth and properties of AlxGal-xN epitaxial layers, Phys. Stat. Sol. (a) 49, 629-636 (1978). 34. K.H. Kim, J.Y. Yi, H.J. Lee, M. Yang, H.S. Ahn, C.R. Cho, S.W. Kim, S.C. Lee, Y. Honda, M. Yamaguchi, and N. Sawaki, Growth of thick AlGaN by metalorganichydride vapor phase epitaxy, Phys. Stat. Sol. (c) 1, 2474-2477 (2004). 35. N.A. Sanford, L.H. Robins, A.V. Davydov, A. Shapiro, D.V. Tsvetkov,, V.A. Dmitriev, S. Keller, U.K. Mishra, and S.P. DenBaars, Reflective index study of AlxGa!.xN films grown on sapphire substrates, J. Appl. Phys. 94, 2980-2991 (2003). 36. G.A. Onushkin, A.E. Nikolaev, A.V. Fomin, O.Yu. Ledyaev, A.E. Cherenkov, E.V. Kalinina, LP. Nikitina, O.V. Kovalenkov, and V.A. Dmitriev, Photoelectrical Properties of AlGaN Epitaxial Layers Grown by HVPE, Phys. Stat. Sol. (c) 1, 465468 (2002). 37. Yu.V. Melnik, A.E. Nikolaev, S.I. Stepanov, A.S. Zubrilov, LP. Nikitina, K.V. Vassilevski, D.V. Tsvetkov, A.I. Babanin, Yu.G. Musikhin, V.V. Tretyakov and V.A. Dmitriev, AIN/GaN and AlGaN/GaN heterostructures grown by HVPE on SiC substrates, Mat. Res. Soc. Symp. Proc. Vol. 482, 245-249 (1998). 38. G.A. Smith, T.N. Dang, T.R. Nelson, J.L. Brown, D. Tsvetkov, A. Usikov, V. Dmitriev, 341 nm emission from hydride vapor-phase epitaxy ultraviolet lightemitting diodes, J. Appl. Phys. 95, 8247 (2004). 39. L.A. Marasina, I.G. Pichugin, M. Tlaczala, Preparation of InN epitaxial layers in InCl3 - NH3 System, Kristall und Technik, 12, 6, 541 - 545 (1977). 40. H. Sunakawa, A.A. Yamaguchi, A. Kimura and A. Usui, Growth of InN by chloride-transport vapor phase epitaxy, Jpn. J. Appl. Phys. 35, LI395 - LI397 (1996). 41. Y. Sato and S. Sato, J. Cryst. Growth 144, 15 (1994). 42. Y. Kumagai, K. Takemoto, T. Hasegawa, A. Koukitu, H. Seki, Thermodynamics on tri-halide vapor-phase epitaxy of GaN and InxGai.xN using GaCl3 and InCl3, J. Cryst. Growth 231, 57-67 (2001). 43. N. Takahashi, J. Ogasawara, A. Koukitu, Vapor phase epitaxy of InN using InCl and InCl3 sources, J. Cryst. Growth 172, 298-302 (1997). 44. N. Takahashi, R. Matsumoto, A. Koukitu and H. Seki, Growth of InN at high temperature by halide vapor phase epitaxy, Jpn. J. Appl. Phys. 36, L743 - L745 (1997). 45. N. Takahashi, R. Matsumoto, A. Koukitu and H. Seki, Vapor Phase Epitaxy of InxGai.xN using InCl3, GaCl3 and NH3 sources, Jpn. J. Appl. Phys. 36,L601 - L603 (1997). 46. Y. Sato and S. Sato, Hydride Vapor Phase Epitaxy of InxGa,_xN Thin Films, Jpn. J. Appl. Phys. 36, 4295-4296 (1997). 47. A. Syrkin, A. Usikov, V. Soukhoveev, O. Kovalenkov, V. Ivantsov, V. Dmitriev, C. Collins, E. Readinger, N. Shmidt, V. Davydov, S. Nikishin, V. Kuryatkov, D. Song, D. Rosenbladt, M. Holtz, InN-based heterostructures grown by modified HVPE,

Hydride Vapor Phase Epitaxy: III- Nitrides

48. 49.

50.

51.

52. 53.

54.

55.

56.

57.

58.

59.

60.

37

Abstracts of the 6th International Conference on Nitride Semiconductors, August 28-September 2, 2005, Bremen, Germany. J. Wu, W. Walukiewicz, Superlattices and Microstructures 34, 63-75(2003). H.P. Maruska, W.C. Rhines, and D.A. Stevenson, Preparation of Mg-doped GaN diodes exhibiting violet electroluminescence, Mat. Res. Bull. Vol. 7, 777-782 (1972). M.A. Mastro, D.V. Tsvetkov, A.I. Pechnikov, .A. Soukhoveev, G.H. Gainer, A. Usikov, V. Dmitriev, B. Luo, F. Ren, K.H. Baik, S.J. Pearton, AlGaN/GaN heterostructures grown by HVPE: growth and characterization, Mat. Res. Soc. Symp. Proc. Vol. 764,39-44 (2003). G. Seryogin, I. Shalish, W.M. Chan, and V. Narayanamurti, Catalytic Growth of GaN Nanowires by Hydride Vapor Phase Epitaxy, Abstracts of the 2004 MRS Fall Meeting, Boston, MA, November 29-December 3, p. 148 (2004). M. Mynbaeva, D. Tsvetkov, Inst. Phys. Conf. Ser. 155, 365-368 (1997). M. Mynbaeva, A. Titkov, A. Kryganovskii, V. Ratnikov, and K. Mynbaev, Structural characterization and strain relaxation in porous GaN layers, Appl. Phys. Lett. 76,1113-1115(2000). M. Mynbaeva, A. Titkov, A. Kryzhanovski, I. Kotousova, A.S. Zubrilov, V.V. Ratnikov, V.Yu. Davydov, N.I. Kuznetsov, K. Mynbaev, D.V. Tsvetkov, S. Stepanov, A. Cherenkov and V.A. Dmitriev, Strain relaxation in GaN layers grown on porous GaN sublayers, MRS Internet Journal Nitride Semiconductor Research, 4, 14 (1999). R.L. Aggarwal, P.A. Maki, R.J. Molnar, Z.-L. Liau, and I. Melngailis, Optically pumped GaN/A10.1Ga0.9N double-heterostructure ultraviolet laser, J. Appl. Phys. 79,2148-2150(1996). Yu. Melnik, A. Nikolaev, S. Stepanov, I. Kalinina, K. Vassilevski, A. Ankudinov, Yu. Musikhin and V. Dmitriev, HVPE GaN and AlGaN "Substrates" for Homoepitaxy, Materials Science Forum Vols. 264-268,1121-1124 (1998). K.S. Boutros, J.S. Flynn, V. Phanse, R.P. Vaudo, G.M. Smith, and J.M. Redwing, InGaN double-heterostructures and DH-LEDs on HVPE GaN-on-sapphire substrates, Mat. Res. Soc. Symp. Proc. Vol. 482, 1047-1052 (1998). V. Dmitriev, A. Nikolaev, A. Cherenkov, D. Tsvetkov, S. Stepanov, N. Kuznetsov, I. Nikitina, A. Kovarsky, M. Yagovkina, V. Davidov, Properties of GaN homoepitaxial layers grown on GaN epitaxial wafers, Mat. Res. Soc. Symp. Proc. Vol. 512,451-456(1998). K. Naniwae, S. Itoh, H. Amano, K. Itoh, K. Hiramatsu and I. Akasaki, Growth of single crystal GaN substrates using hydride vapor phase epitaxy, J. Cryst. Growth 99,381-384(1990). R.J. Molner, W. Gotz, L.T. Romano, N.M. Johnson, Growth of gallium nitride by hydride vapor-phase epitaxy, J. Cryst. Growth. 178, 147-156 (1997).

38

Dmitriev, Usikov

61. P.A. Maki, R.J. Molner, R.L. Aggarwal, Z-L. Liau, and I. Melngailis, Optically Pumped GaN-AlGaN Double-Heterostructure Lasers Grown by ECR-GSMBE and HVPE, Mat. Res. Soc. Symp. Proc. Vol. 395, 919-924 (1996). 62. J.W. Johnson, J. Han, A.G. Baca, R.D. Briggs, R.J. Shul, J.R. Wendt, C. Monier, F. Ren, B. Luo, S.N.G. Chu, D. Tsvetkov, V. Dmitriev, S.J. Pearton, Comparison of AlGaN/GaN high electron mobility transistors grown on AIN/SiC templates or sapphire, Solid-State Electronics 46 (2002) 513-523. 63. M.J. Manfra, L.N. Pfeiffer, K.W. West, H.L. Stomer, K.W. Baldwin, J.W.P. Hsu, D.V. Lang, R.J. Molnar, High-mobility AlGaN/GaN heterostructures grown by molecular-beam epitaxy on GaN templates prepared by hydride vapor phase epitaxy, Appl. Phys. Lett. 77, 2888-2890 (2000). 64. V. Schwegler, C. Kirchner, M. Seyboth, M. Kamp, K.J. Ebeling, Yu.V. Melnik, A.E. Nikolaev, D. Tsvetkov, and V.A. Dmitriev, GaN/SiC Quasi-Substrates for GaN-based LEDs, Phys. Stat. Sol. (a) 176, 99-102 (1999). 65. Y. Shibata, A. Motogaito, H. Miyake, K. Hiramatsu, Y. Ohuchi, H. Okagawa, K. Tadatomo, T. Nomura, Y. Hamamura, and K. Fukui, Fabrication and characterization of UV Schottky detectors by using a freestanding GaN substrates, Mat. Res. Soc. Symp. Proc. Vol. 831 (2005) 137-142. 66. M. Kneissl, Z.H. Zang, M. Teepe, C. Knollenberg, N.M. Johnson, S.B. Schujman, L.J. Schowalter, A. Usikov, and V. Dmitriev, reported at the 6th International Conference on Nitride Semiconductors, August 28-September 2, 2005 Bremen, Germany; to be published. 67. C. Sasaoka, H. Sunakawa, A. Kimura, M. Nido, A. Usui, A. Sakai, High-quality InGaN MQW on low-dislocation-density GaN substrate grown by hydride vaporphase epitaxy, J. Cryst. Growth 189/190, 61-66 (1998). 68. J.K. Gillespie, R.C. Fitch, N. Moser, T. Jenkins, J. Sewell, D. Via, A. Crespo, A.M. Dabiran, P.P. Chow, A. Osinsky, M.A. Mastro, D. Tsvetkov, V. Soukhoveev, A. Usikov, V. Dmitriev, B. Luo, S.J. Pearton, F, Ren, Uniformity of dc and rf performance of MBE-grown AlGaN/GaN HEMTS on HVPE-grown buffers, Solid State Electronics 47, 1859-1862 (2003). 69. C. Lee, H.Y. Lee, H. Shin, C. Kim, H. Ko, J. Han, H. Kim, and K. Lee, The influence of threading dislocation density on Hall effects in free-standing GaN substrates, Phys. Stat. Sol. c, 2, (2005) 2137-2140. 70. M.K. Kelly, R.P. Vaudo, V.M. Phanse, L. Gorgens, O. Ambacher and M. Stutzmann, Large Free-Standing GaN Substrates by Hydride Vapor Phase Epitaxy and Laser-Induced Liftoff Jpn. J. Appl. Phys. 38, pp. L217-219 (1999). 71. F. Yun, M.A. Reshchikov, K. Jones, P. Visconti, H. Morkoc, S.S. Park, K.Y. Lee, Electrical, structural, and optical characterization of free-standing GaN template grown by hydride vapor phase epitaxy, Solid-Sate Electronics 44, 2225-2232 (2000). 72. O. Kryliouk, M. Reed, M. Mastro, T. Anderson, and B. Chai, GaN substrates: Growth and Characterization, Phys. Stat. Sol. (a) 176, 407-410 (1999).

Hydride Vapor Phase Epitaxy: III- Nitrides

39

73. K. Motoki, T. Okahisa, N. Matsumoto, M. Matsushima, H. Kimura, H. Kasai, K. Takemoto, K. Uematsu, T. Hirano, M. Nakayama, S. Nakahata, M. Ueno, D. Hara, Y. Kumagai, A. Koukitu and H. Seki, Preparation of Large Freestanding GaN Substrates by Hydride Vapor Phase Epitaxy Using GaAs as a Starting Substrate, Jpn. J. Appl. Phys. 40, L140-L143 (2001). 74. C.R. Miskys, M.K. Kelly, O. Ambacher, G. Martinez-Criado, and M. Stutzmann, GaN homoepitaxy by metalorganic chemical-vapor deposition on free-standing GaN substrates, Appl. Phys. Lett. 77,1858-1860 (2000). 75. R.P. Vaudo, X. Xu, A. Salant, J. Malcarne and G.R. Brandes, Characteristics of semi-insulating, Fe-doped GaN substrates, Phys. Stat. Sol. (a) 200, No 1 , 18 21(2003). 76. B. Monemar, H. Larsson, C. Hemmingsson, I.G. Ivanov, D. Godova, Growth of thick GaN layers with hydride vapour phase epitaxy, J. Cryst. Growth 281, 17-31 (2005). 77. R.P. Vaudo, X. Xu, A. Salant, J. Malcarne, and G.R. Branes, Characterization of semi-insulating, Fe-doped GaN substrates, Phys. Stat. Sol. a, 1 (2003) 18-21. 78. P.P. Paskov, T. Paskova, P.O. Holtz, B. Monemar, Polarization-dependent spectroscopy of the near-bandgap excitonic emission in free standing GaN, Mat. Res. Soc. Symp. Proc. Vol. 798 (2004) 601-606. 79. Yu.V. Melnik, V.A. Soukhoveev, K.V. Tsvetkov, and V.A. Dmitriev, First AlGaN free-standing wafers, Mat. Res. Soc. Symp. Proc. 764, 363-368 (2003). 80. I. Yonenaga, A. Nikolaev, Yu. Melnik, V. Dmitriev, High-temperature hardness of bulk A1N single-crystal A1N, Jpn. J. Appl. Phys. 40, L426-L427 (2001). 81. B. Luo, J.W. Johnson, O. Kryliouk, F. Ren, S.J. Pearton, S.N.G. Chu, A.E. Nikolaev, Yu. V. Melnik, V.A. Dmitriev, T.J. Anderson, High breakdown M-I-M structures on bulk A1N, Solid-State Electronics 46, 573-576 (2002). 82. D.I. Florescu, V.M. Asnin, F.H. Pollak, Compound Semicond. 7, 61 (1998). 83. S. Nakamura, M. Senoh, S. Nagahama, N. Iwasa, T. Yamada, T. Matsushita, H. Kiyoku. Y. Sugimoto, T. Kozaki, H. Umemoto, M. Sano, and K. Ohocho, Continues-wave operation of InGaN/GaN/AlGaN-based laser diodes grown on GaN substrates, Appl. Phys. Lett. 72, 2014-2016 (1998). 84. M. Kuramoto, C. Sasaoka, Y. Hisanaga, A. Kimura, A.A. Yamaguchi, H. Sunakawa, N. Kuroda, M. Nido, A. Usui, and M. Mizuta, Continues-Wave Operation of InGaN Multi-Quantum-Well Laser Diodes Grown on an n-GaN Substrate with a Backside n-Contact, Phys. Stat. Sol. (a) 176, 35 (1999). 85. T. Nishida and N. Kobayashi, Ten-Milliwatt Operation of an AlGaN-Based Light Emitting Diode Grown on GaN Substrate, Phys. Stat. Sol. (a) 188, 113-116, (2001). 86. Vladimir Dmitriev, Alexander Fomin, Yuri Melnik, Vitaliy Soukhoveev, Vladimir Ivantsov, Gordon Gainer, and Katie Tsvetkov, Properties of Bulk GaN Wafers and Homoepitaxial Layers, reported at the the International Workshop on Bulk Nitride Semiconductors, May 18th through 23rd, 2002, Amazonas, Brazil [unpublished].

40

Dmitriev, Usikov

87. X. Xu, R.P. Vaudo, J. Flynn, J. Dion, and G.R. Brandes, MOVPE homoepitaxial growth on vicinal GaN(OOOl) substrates, Phys. Stat. Sol. (a) 5, 727-731 (2005). 88. D. Hanser, L. Liu, E.A. Preble, D. Thomas, M. Williams, Growth and Fabrication of 2-inch Free-Standing GaN Substrates via the Boule Growth Method, Mat. Res. Soc. Symp. Proc. Vol. 798, 257 - 262 (2004). 89. Yu. Melnik et al., A1N substrates: fabrication via vapor phase growth and characterization, Phys. Stat. Sol. (a) 1, 22-25 (2003). 90. M.A. Mastro, D. Tsvetkov, V. Soukhoveev, A. Usikov, V. Dmitriev, B. Luo, F. Ren, K.H. Baik, S.J. Pearton, Hydride vapor phase epitaxy-grown AlGaN/GaN high electron mobility transistors, Solid State Electronics 47, 1075-1079 (2003). 91. J.B. Lam, G.H. Gainer, S. Bidnyk, A. Elgawadi, G.H. Park, J. Krasinski, J.J. Song, D.V. Tsvetkov, V.A. Dmitriev, Comparative study of HVPE- and MOCVD-grown nitride structures for UV lasing applications, Mat. Res. Soc. Symp. Proc. Vol. 639, G6.4.1-G6.4.6(2001). 92. A.S. Usikov, D.V. Tsvetkov, M.A. Mastro, A.I. Pechnikov, V.A. Soukhoveev, Y.V. Shapovalova, O.V. Kovalenkov, G.H. Gainer, S.Yu. Karpov, V.A. Dmitriev, B.O'Meara, S.A. Gurevich, E.M. Arakcheeva, A.L. Zakhgeim, H. Helava, Indiumfree violet LEDs grown by HVPE, Phys. Stat. Sol. N7, 2265-2269 (2003). 93. A.S. Usikov, Yu.V. Melnik, A.I. Pechnikov, V.A. Soukhoveev, O.V. Kovalenkov, E. Shapovalova, S.Yu. Karpov, V.A. Dmitriev, HVPE-grown AIN-GaN based structures for UV spectral region, in: M.S. Shur and A. Zukauskas (eds.), UV SolidState Light Emitters and Detectors, 15-29. 2004 Kluwer Academic Publishers. 94. B. Luo et al., HVPE-grown AlGaN/GaN HEMTs, Mat. Res. Soc. Symp. Proc. 764, 45-55 (2003). 95. M.A. Mastro, , D. Tsvetkov, V .Soukhoveev, A. Usikov, V. Dmitriev, B. Luo, F.Ren, K.H. Baik, S.J. Pearton., RF performance of HVPE grown AlGaN/GaN HEMTs, Solid-State Electronics 48, 179-182 (2004). 96. V. Dmitriev, A.Usikov, H.Helava, B. O'Meara: HVPE offers alternative route to AlGaN-based UV emitters, Compound Semiconductors, Dec. 2004.

CHAPTER 2 PLANAR MOVPE TECHNOLOGY FOR EPITAXY OF III-NITRIDE MATERIALS

Martin Dauelsberg, Bernd Schineller, Johannes Kaeppeler AIXTRONAG, Kackertstrasse 15-17, 52072 Aachen, Germany, info@aixtron. com This chapter gives an overview of horizontal and planetary reactor technology for MOVPE of Nitride layers. The specific growth kinetics of Nitrides and its role for process and reactor design is discussed, and the use of computational modeling in order to understand the fundamentals of the growth process and to give guidance for process and reactor design is indicated. Certain aspects of the flow dynamics of horizontal tube reactors are outlined, and the issues relevant to the transition to radial flow planetary reactors in order to enable volume production on a multi-wafer scale are discussed. Process parametric dependencies in large scale planetary reactors are exemplified, and the scalability of the planetary reactor is emphasized as one of its intrinsic advantages. In addition, in situ monitoring of the growth and the possibility of feedback control are gaining more and more importance in modern epi fabs. The determination of growth rate and layer quality as well as the accurate measurement of the surface temperature and its feedback into the reactor's control software will be discussed.

1.

History of Reactor Development for Ill-Nitrides

The MOVPE of Ill-Nitrides became important in the early 90s when I. Akasaki 1 and S. Nakamura 2 first demonstrated Nitride based light emitting diodes (LEDs). These results motivated a lot of research groups to embark on research into the MOVPE of Ill-Nitrides. These groups often started with modified reactors, which were originally used for the

41

42

Dauelsberg, Schineller, Kaeppeler

growth of GaAs or InP based materials. The precursors used for the group III component when growing Ga (Al, In) N are mainly the same as used for the growth of GaAs or InP based materials. These are TMGa, TEGa, TMA1 and TMIn. Regarding the group V precursor NH3 was normally used for growing Ill-Nitrides. A notable difference between this reagent and AsH3 and PH3 is the higher bond energies (-70.672.8kcal/mol for AsH3, -76.9-81.4kcal/mol for PH3 and -93105kcal/mol -depending on source data-, for NH3),3'4 which requires a higher growth temperature of about 1100°C for GaN. A more detailed description of the thermal chemistry is made in section 3. As the existing reactors were designed for maximum growth temperatures of about 850-900°C modifications were required to enable the reactors to run at temperatures up to, or above 1100°C. For the substrate holder (susceptor) to reach these temperatures most reactors are equipped with Radio Frequency (RF) heater or resistance heater. Most of the fundamental research was performed on these small homemade reactors. In the mid 90s a few groups were able to grow LED structures on these small systems and the first interest in larger scale reactors for the production of blue LEDs developed. At that time production processes for growing the GaAlInP based Epilayers used for LEDs operating in the yellow to red spectral region were established on multiwafer systems. These systems were not however able to grow GaN based material because the maximum operation temperature was limited to 900°C. Another limitation concerned the material for the wafer carrier (susceptor), which was mainly uncoated graphite. This could not be used in an NH3 containing ambient because NH3 or its decomposition products react with the pure graphite at these elevated temperatures. The most common solution to this problem is to coat the graphite with SiC, which does not react, with NH3 at process temperatures and therefore protects the graphite against corrosion. Another requirement, which came up, was the necessity of In situ monitoring of the wafer surface during growth. As there are still no commercial substrates available which match with the lattice constant of GaN, Sapphire and SiC are the most common selected substrates. For

Planar MOVPE - GaN

43

compensating the crystal lattice mismatch of the substrate compared to the GaN, a low temperature nucleation layer is required5. To improve the growth control, especially between the low temperature 3D Nucleation growth and the 2D GaN Growth, In situ reflectance measurement becomes a very useful instrument. Therefore optical access to the wafer surface is essential for MOVPE reactors. The current market requirements for production MOVPE machines are significantly changed from those of the late 90s. Today the market expects a system with high throughput and low cost of ownership combined with high performance regarding the uniformity of layer properties such as thickness, wavelength, electro-luminescence intensity and doping level. This market also requires an up-scaling of the reactor to permit growth on larger wafer areas whilst at the same time improving the performance of the reactor.

1994

1997

2000

2003

year Fig. 1. Development of planar reactor size in substrates per run versus years.

2.

Types of Planar Reactors

Two types of planar reactor principles will be introduces here. The horizontal reactor in its simplest shape is a tube with a heated wafer

44

Dauelsberg, Schineller, Kaeppeler

carrier and the process gases are introduced in one end and will be feed to the pump system through the other end.

Fig. 2.1. Schematic top view of a horizontal MOCVD Reactor.

The gas concentration of the process gases and its decomposed species will be depleting from the gas inlet side into the gas outlet direction. This given depletion will cause a decreasing growth rate on the substrate along the gas flow direction. The most efficient compensation method is to rotate the substrate, which will average the growth conditions into a radial symmetry. If the depletion curve is be tuned to a linear shape, the rotation will generate a flat uniform layer thickness on the wafer. The rotation will also help to homogeneous the temperature of the substrate. Today horizontal reactors are mainly used for R&D of IIINitride layers and for small scale production like for Laser Diodes. The planetary reactor can be regarded as an aggregate of such reactor sections to form a complete circular shape. P.M. Frijlink developed the concept of the Planetary Reactor® in the late 80s.6 The substrates are placed on a platen rotates around the gas inlet at the center of the reactor. The gas inlet is separated into two levels. Through the top inlet group III process gases are injected whilst the lower inlet is used to inject Group V precursor and its carrier gas into the reactor. A schematic top view of the reactor is shown in figure 2.2. In the original reactor built by P. M. Frijlink IR lamps heated the susceptor. As described in the above GaN growth requires susceptor temperatures of about 1100°C. Due to this requirement for higher process temperatures, planetary reactors for III-Nitride growth are

Planar MOVPE - GaN

45

equipped with an inductive RF-heater. Such equipped planetary reactors are able to reach growth temperature up to 1600°C.

Exhaust gas collector

Substrate Gas inlet 77TT

Fig. 2.2. Schematic top view of a MOCVD Planetary Reactor®.

Computational modeling of Nitride MOVPE processes and reactors are essential in the development of MOCVD equipment. The use of this technique improved the quality and reduced the time of the reactor development and therefore will be explained here. Both reactor types will be explained in later sections. 3. Reactor Modeling In the history of MOVPE (Metal Organic Vapor Phase Epitaxy) growth of GaN and related materials, reactor design has always been a crucial aspect of growth technology. Common MOVPE reactor types for growth of group-Ill Nitrides as well as for conventional III-V compound semiconductors include horizontal type reactors, either configured as linear tube reactors or as radial flow Planetary reactors. These can be seen as distinct from rotating disk vertical reactors or close space showerhead reactors (see Chapter 3). This section will elucidate various

46

Dauelsberg, Schineller, Kaeppeler

aspects of horizontal reactor technology, including flow dynamics and transport phenomena, reactor modeling, the discussion of parametric dependencies and process tuning strategies, as well as reactor scaling to accommodate a larger number of substrates during a single growth run. Reactor design and functionality are intimately linked to the specific requirements of the Nitride growth process. It has been recognized during the course of Nitride development that the typically wide process window during typical multi-layer growth sequences for optoelectronic or electronic devices calls for customized hardware design solutions and adaptations based on reactors for conventional III-V compound semiconductor MOVPE. In subsection 3.1 the current understanding of gas phase kinetics encountered with the growth of GaN and related materials is outlined. Computational modeling of flow dynamics and reaction kinetics provides for fundamental insight into the growth mechanism and has proved to be a valuable and indispensable tool for both reactor vendors as well as device manufacturers to improve reactor design and process. Some background on the scope and methods of computational modeling is given in subsection 3.2. In subsection 3.3 the flow dynamics and species transport in horizontal tube reactors is discussed, and the motivation for the transition from the linear to the radial Planetary configuration is explained. Current state of the art Nitride production technology based on the Planetary reactor configuration is presented in subsection 3.4 and section 5 with focus on transport phenomena, parametric dependencies, and in-situ diagnostics.

3.1. Growth Kinetics of Group-Ill Nitride MOVPE In contrast to the MOVPE of conventional III-V compounds, the growth kinetics of Nitrides exhibit some particular features. Due to the disparate decomposition temperature of ammonia (by far the most common groupV precursor) and the typically used group-Ill metalorganics, like Trimethyl-Aluminum (TMA1), Trimethyl-Gallium (TMGa), TrimethylIndium (TMIn), and in order to prevent the growing crystalline layer from dissociating, very high V-III ratios and high flow rates of ammonia

Planar MOVPE - GaN

47

are necessary with inevitable implications to gas phase transport and flow dynamics. Other properties of Nitride MOVPE growth include the sensitivity of the growth process with respect to the kind of carrier gas, H2 or N2, and the complex adduct and gas phase nucleation chemistry that crucially determines group-Ill incorporation efficiency. The thermal decomposition of most metalorganic precursors typically involves the sequential release of the methyl ligands by cleavage of the metal-carbon bonds. Rate constants for each of the steps were determined by decomposition experiments under conditions close to MOVPE as reviewed.10 More recently ab initio computational chemistry was employed to determine bond strengths and reaction rate constants.3 Both approaches yield rather good agreement. The onset of thermal decomposition of TMGa for instance, occurs at a temperature of 350°C - 400°C by cracking of the first Ga-CH3 bond with an activation energy of ~ 59 kcal/mole. The thermal cracking of NH3 has been the subject of experimental studies in combustion chemistry resulting in the detailed kinetic mechanisms and the quantitative measurements of reaction rates.4 Gas phase decomposition proceeds as a second order reaction with a rather high activation energy of -95 kcal/mole, with initial cracking occurring at temperatures well above 800°C-900°C promoting ammonia dissociation of only a few percent at typical growth temperatures of GaN epitaxial layers (1000°C - 1100°C). Apparently, during MOVPE the majority of NH3 is decomposed catalytically on the growing surface through dissociate adsorption in conjunction with adsorbed group-Ill compounds. The use of alternative nitrogen sources, such as hydrazine derivatives like UDMH (unsymmetric dimethyl-hydrazine), has been investigated, too.12 Experimental decomposition studies indicate 50% cracking at 420°C, i.e. at a decomposition temperature well matched to that of the metalorganic group-Ill precursors. For the deposition of GaN, the formation of adducts between ammonia and the group-Ill precursors through a Lewis acid-base interaction plays a crucial role. A number of experimental and modeling

48

Dauelsberg, Schineller, Kaeppeler

studies have focused on the gas phase pre-reaction of NH3 and TMGa or TMA1.1314'15 According to Thon and Kuech13 the formation of the initial adduct between NH3 and TMGa takes place at temperatures below 200°C. Subsequent methane elimination leads to (CH3)2Ga-NH2 which can react to form dimeric and trimeric compounds with Ga and N atoms forming a ring. At temperatures above 700°C the abundance of dimeric and trimeric species tends to decline again owing to their thermal decomposition. The strength of the Lewis donor-acceptor bond itself is rather weak, unless subsequent alkane elimination occurs leading to the formation of more or less stable compounds with direct bonds between III- and Vatoms. The relative magnitude of adduct strength and the barrier to alkane elimination is crucial for the MOVPE process and depends on the participating III-V elements. Apparently, aluminum and nitrogen form the most stable compounds among the relevant group-Ill elements, which explains the high risk of particle formation during growth of A1N. The practical relevance of these processes to MOVPE lies in the fact that gas phase adduct formation results in the creation of gas phase species with low volatility, high molecular weight and poor diffusion properties, that tend to nucleate and fall out in the form of small nanoscale particles.16 These are subject to strong thermophoretic forces, usually opposed to the direction from bulk gas flow to the substrate, and carried away by the gas flow. In summary, these processes present mechanisms to deplete the gas phase from precursor species. The control of these parasitic processes and suppression of nucleation is key to the control of growth efficiency during Nitride MOVPE. 3.2. Modeling of MOVPE Processes The modeling of Nitride MOVPE processes has been used to gain insight into the fundamentals of the growth process itself.7'14 On the other hand, as MOVPE has evolved into a mass production technique, modeling has been of increasing value in supporting and guiding process development and equipment design.9 Semiconductor equipment vendors routinely use

Planar MOVPE - GaN

49

computational modeling in order to improve reactor design and processes, to shorten their development cycles and to cut down development and testing costs. The mathematical model used for computational predictions of the MOVPE growth process17 is based on the continuum mechanical description of flow, heat transfer and multi-component mass transport of chemical species, supplemented by physical models for thermal radiation transfer, gas phase and surface chemical reactions and electromagnetic inductive heating. The continuum description is adequate, since MOVPE reactors are operated at sufficiently high pressures (10 - 1000 mbar) and large characteristic dimensions, such that the Knudsen number, i.e. the ratio of the mean free path length over the characteristic dimension of the flow domain, is below Kn « 0.01, the condition for continuum flow. Computations involve the solution of coupled partial differential equations, e.g. the Navier-Stokes equation for flow, supplemented by appropriate boundary conditions.18 Transport phenomena are described either on a full three-dimensional computational domain or in a twodimensional representation by exploiting certain symmetries inherent to the problem. MOVPE growth exhibits a high sensitivity not only to local growth temperature and thermal uniformity on the substrate, but also to the thermal environment within the entire reaction chamber. For the purpose of predictive equipment scale modeling, the accurate calculation of temperatures is indispensable. In particular, the modeling of thermal radiation in the MOVPE reactor is important as it happens to be the predominant mode of heat transfer. The more accurate radiation models are based on ray tracing or view factors, and typically provide reasonably high spatial resolution, but imply high computational effort on the other hand. Another class of approaches, so called diffusive approximations are based on the solution of a radiation transfer equation, e.g. the popular Discrete Ordinates Method.21 In practice the choice of the radiation model usually involves a compromise between the degree of prediction accuracy and computational effort.

50

Dauelsberg, Schineller, Kaeppeler

Computational Fluid Dynamics (CFD) is the field of science dedicated to the numerical solution techniques for the system of coupled partial differential equations describing fluid flow and related transport phenomena. The most commonly used numerical techniques for the regime of incompressible, viscous flows, including multi-component gas mixtures are based on the Finite-Volume method. It consists of the discretization of the governing equations on structured or unstructured computational grids and the iterative solution of the resulting system of algebraic equations involving some pressure correction scheme. For more details on numerical techniques refer to 19 and 20. 3.3. Horizontal Tube Reactors: Flow Dynamics and Reactor Technology Historically, many of the first MOVPE growth attempts were made on horizontal tube reactors, typically consisting of a longitudinal quartz tube as flow duct and a heated graphite susceptor placed on the bottom, with the reactive gases passing over the substrate in a direction parallel to the growing surface. Modern versions of horizontal tube reactors, still popular within the research community, feature rectangular flow ducts (so called liners), substrate rotation, and access to in situ monitoring from the top, for instance. They are designed to accommodate single or multiple 2 inch substrates. With the high reactivity of group-Ill and group-V species in mind, the gas flow of both precursors often is separated at the inlet. The preferred means of heating is RF (radio frequency) induction heating in order to reach typical growth temperatures of up to 1100°C at a reasonably fast rate. In its simplest variant a graphite block is placed in the center of a cylindrical coil, whereas a flat coil is used whenever access to in situ diagnostics is desired. Figure 3.1 shows the setup of the commercial horizontal tube reactor AIX 200/4 RF-S with optical ports for in situ diagnostics on top of the liner, and a flat, pancake type RF induction coil underneath a graphite disk, that constitutes the susceptor and substrate holder.

Planar MOVPE - GaN

51

The control of growth efficiency and deposition uniformity in a horizontal tube reactor is governed by the distribution of mass transport of the growth limiting species to the growing surface and its interaction with gas phase and surface kinetics. The predominant mode of species delivery to the growth surface is flow itself (i.e. forced convective species transport) parallel to the liner walls and substrate surface. Superposed to flow is diffusion along the direction of steepest concentration gradient perpendicular to the substrate surface. In general the characteristics of convective and diffusive transport define the dependence of deposition uniformity on various reactor parameters, like gas flow rates. Overall growth rate and efficiency, however, will also depend on the consumption of growth nutrients by gas phase adduct formation (see subsection 3.1).

Fig. 3.1. Schematic of the horizontal tube reactor AIX 200/4 RF-S.

All horizontal reactors with main flow parallel to the growth plane, exhibit growth depletion to some degree. Depletion is related to the consumption of growth limiting species in the flow direction by deposition on the susceptor, on the substrate and on adjacent reactor walls, as well as to the loss of nutrients due to gas phase adduct

52

Dauelsberg, Schineller, Kaeppeler

formation and subsequent nucleation. The non-uniformity caused by depletion is compensated for by slow substrate rotation in order to average layer deposition rates on the substrate. The fine tuning of layer thickness uniformity is then conducted by optimizing gas flow rates and mixture composition. Computational modeling proves valuable in facilitating this tuning procedure while keeping the number of growth experiments low. On the other hand, the purely two-dimensional picture of balancing supply and depletion of growth limiting species parallel to the direction of the main gas flow does not fully account for all the root causes of layer thickness non-uniformity. Flow in a horizontal tube reactor develops into a parabolic pattern confined by the liner walls as illustrated in figure 3.2. In particular the confinement of flow by the liner side walls causes the total mass flow of the reactive gas mixture through the liner to distribute as a parabolic profile across the substrate. Since the deposition rate profile is primarily governed by the forced convective transport of growth limiting chemical species and their depletion, a parabolic flow pattern in the horizontal flow duct results in declining growth rates away from the central axis of the reactor. The loss of precursor material due to side wall coating adds to this effect as a typically minor contribution.

Fig. 3.2. Development of parabolic flow pattern in a horizontal flow duct.

Classical fluid dynamics teaches that the length L (hydrodynamic development length) for fluid flow entering into a rectangular duct with infinite depth in order to fully develop into parabolic flow (see figure 3.2 for a schematic) is governed by

Planar MOVPE - GaN

53

(1) L = 0 . 0 4 D R e , with

p - fluid density, D - duct width, U = fluid velocity, ju - dynamic viscosity. Based on these considerations, the lateral uniformity limitation due to side wall confinement in horizontal tube reactors can be overcome by increasing the hydrodynamic development length, such that a plug flow profile will extend from the entrance up to the substrate position. Substantial uniformity enhancement for MOVPE in a horizontal tube reactor on 1 x 2 inch could be obtained, for instance, by simply replacing H2 by N2 or a blend of H2 and N2 as carrier gas, in order to increase fluid density and promote the retention of plug flow characteristics to that region of the reactor tube containing the substrate.23 Another elegant approach to overcome side wall confinement encountered with horizontal tube reactors completely and enable more efficient susceptor surface coverage with substrates is the transition from linear tube reactors to radial planetary reactors, thereby avoiding all side wall effects. 3.4. Planetary Reactors: Transport Phenomena and Parameter Dependencies Planetary reactors can be considered the straightforward extension of the principle of horizontal tube reactors to multi-wafer processing. Instead of placing a single substrate on the bottom of a rectangular flow duct, multiple substrates are arranged on top of a cylindrical disk, with their centers lying on a notional circle around the reactor center. The gas inlet is located at the reactor center, with the reactant gases injected in a circularly symmetric manner. The advantage of this configuration is obvious. Liner side walls are omitted, thereby getting rid of one important root cause of deposition non-uniformity in horizontal tube reactors (see subsection 3.3) that typically requires that a certain margin is left between wafer edge and side wall. The Planetary reactor

54

Dauelsberg, Schineller, Kaeppeler

configuration allows multiple substrates to be located closer to each other. This way the usage of susceptor surface area and the consumption of gases becomes more efficient. Apart from these purely geometric considerations, Planetary reactors tend to utilize the precursor gases in a more efficient way due to their volume expansion along the direction of main gas flow and the corresponding slowdown of gas flow. Gas consumption indeed contributes to the total cost of ownership, which is becoming of increasing importance in a volume production environment. Planetary reactors are commercially available in various configurations, the most common among them being 6x2 inch, 11x2 inch and 24x2 inch for 2 inch wafers, 7x3 inch and 8x4 inch for larger area substrate sizes. In order to obtain reasonable heat-up rates and allow perfect thermal control of the process environment, they are RF (Radio Frequency) inductively heated by means of a planar, so called pancake coil. Provision is made for adjusting the distance of the coil to the graphite susceptor, and the reactor is equipped with a quartz plate process chamber ceiling that is thermally controlled within a certain temperature range. Figure 3.3 shows a photograph of the 24x2 inch reactor interior, the flagship product of AIXTRON AG for the volume production of GaN based layers.

Fig. 3.3. Commercial volume production Planetary Reactor® AIX 2600G3 HT for the MOVPE of GaN based heterostructures, AIXTRON AG.

Planar MOVPE - GaN

55

Like all generic horizontal flow reactors, Planetary reactors exhibit a characteristic depletion growth profile. The mechanisms that determine depletion of growth rates in radial direction in a Planetary Reactor are essentially the same as for horizontal tube reactors, as described previously in subsection 3.3. Growth limiting species are depleted in the gas phase along the direction of gas flow from the center to the exhaust. Depletion can be put down to the usage of these species by growth on the susceptor and the substrates themselves while traveling along the susceptor surface. The shape and the steepness of depletion, however, are determined by the local balance between species transport by forced convection parallel to the substrate surface and diffusion of species perpendicular to the substrate surface. Applying substrate rotation at a moderate rate is the most effective remedy for depletion induced nonuniformity. Rotation is facilitated by the so called gas foil rotation. The disks that carry individual wafers are elevated and kept rotating by a gas flow from underneath. Main rotation of the whole susceptor disk is superimposed on to individual substrate rotation in order to equilibrate the susceptor thermally and ensure that different regions of the wafers are exposed to the same average deposition conditions. Figure 3.4 shows the principle setup of a Planetary reactor and illustrates the averaging of the deposition profile along radial direction by rotation. The separation of group-V and group-Ill gas flows at the central injector serves several purposes. First, there is the avoidance of undesired premature gas phase reactions between group-Ill metalorganics and ammonia (s. subsection 3.1) that would inevitably occur right at the inlet in the case of using a premixed gas mixture. With the detailed chemistry of adduct species formation, decomposition and the nucleation of low volatile gas species during the Nitride growth process in mind, the gas flow pattern and the point of mixing are carefully determined by the appropriate design of the injector geometry.

56

Dauelsberg, Schineller, Kaeppeler

Cairier + MH3

Carrier + group EI elements + dopant

quartz ceiling

Growth rate on rotated wafer Rotating inductively heated susceptor

radius

Fig. 3.4. Schematic set-up of the Planetary Reactor , indicating gas flow, rotation and the distribution of growth rates.

Any disturbances of laminar flow at the inlet e.g. return flow must not be induced as this would kindle premature gas phase processes immediately. Likewise, the thermal management of the injection zone is carefully designed to avoid losses by coating and gas phase pre-reactions. Another advantage of supplying the group-Ill mixture separately from the group-V mixture is that it extends the point at which deposition on the susceptor starts to occur away from the inlet. Therefore, in a Planetary reactor as well as any other horizontal reactor ammonia is injected at the same height as the susceptor surface, whereas the gas mixture containing the growth limiting group-Ill precursors, coming from above, needs to reach the susceptor surface by diffusion through the layer of ammonia and carrier gas. By the independent supply and tuning the two gas flows at the injector, the point of onset of deposition in the radial direction can be adjusted. Likewise the precise control of growth rate and compositional uniformity under given process conditions (growth temperature, pressure, gas composition) can be achieved through

Planar MOVPE - GaN

57

the tuning of the flow rates. Figure 3.5 illustrates the dependence of the deposition rate profile on the susceptor and the rotating substrate holder on the flow rate of the group-Ill carrier gas. Experimental results in a production scale 24x2 inch Planetary Reactor® are compared with the predicted growth rate distribution of GaN by modeling of gas phase transport phenomena and the MOVPE growth process itself (as outlined in subsection 3.2).

distance from center (m) Fig. 3.5. Computed and measured growth rates of GaN in the Planetary Reactor AIX 2600G3 HT with 24 x 2 inch substrates.

4.

In-situ Technology in Nitride MOCVD Systems

In-situ technology enables the process engineer to directly observe the growth of the layer structures to determine key parameters such as growth rate, the timing of crucial growth steps and overall layer quality. The most commonly used in-situ technology is based on the measurement of Fabry-Perot interferences. A light beam at normal incidence to the layer surface gets partially reflected at the surface and the GaN/sapphire interface, since the difference of the indices of

Dauelsberg, Schineller, Kaeppeler

58

refraction differ greatly at these planes. Due to their respective phase shifts, the reflected light beams give rise to constructive and destructive interferences as the layer grows thicker. The frequency of the interferences allows for the determination of the growth rate if the refractive index of the layer material is known.

Sapphire

)

500

1000

1500

2000

Time [s]

Fig. 4.1. Microscopic disturbances of the

morphology of the surface.

In addition to the growth rate, the process engineer can draw conclusions on the quality of the layer surface. The in-situ kit collects the integrated signal from a macroscopic area which is considerably larger than the microscopic disturbances of the morphology of the surface (see figure 4.1 left). Thus, the amplitude of the interference signal decreases with increasing morphological disturbance, since signals with different phase shifts are combined. This quenching of the interference oscillations can be taken as a qualitative indicator for the surface morphology (see figure 4.1 right). The second most important in-situ technology is the pyrometric determination of the surface temperature. By using the black body equations the temperature of a surface can be measured if the emissivity factor is known. This emissivity factor, however, constantly changes as the layer grows. However, the emissivity factor is directly coupled to the reflectivity of the surface, making the combined measurement of reflectivity and temperature possible.

59

Planar MOVPE - GaN

*• via glass fiber into detector —

LED on: measure reflectance

prism

i

\

1=950 nm rt.=rau tini

L

20 Hz LED current

LED off: measure temperature

jwafer Fig. 4.2. Schematic of a Epitune® II sensor head.

Figure 4.2 shows a schematic of the Epitune® II sensor head. A pulsed LED (X=950 nm, f>20 Hz) is focused onto the wafer via a prism and a lens setup. The reflected light is conducted through a glass fiber to the detector system for evaluation. Here, reflectance and emissivity are calculated. During the dark-cycle of the LED the sensor head picks up the black-body radiation from the wafer surface and sends it through the same glass fiber for measurement. Since the rotation speeds of the wafer holder in modern epitaxy systems is low compared to the repetition cycle of the LED, reflectance, emissivity and pyrometric temperature can practically be measured on the same point on the wafer. Like this, in-situ scans of the growth rate and layer properties across the wafer can be done and the effect of changes in process conditions can be directly observed. Figure 4.3 shows the evolution of the growth rate across the diameter of the same 4 inch wafer during the same epitaxy run for different total flow rates which were manually changed by the process engineer during the run. To facilitate the measurement, 7 scan points across the 4 inch wafer diameter were monitored by the insitu system. As can be seen, the manual change of the total flow rate allowed the process engineer to change the growth profile from convex (low growth rate at the wafer center) to concave (high growth rate at the wafer center). Thus, the in-situ reflectance tool helps to optimize the value of the growth rate as well as its distribution across the wafer and without

60

Dauelsberg, Schineller, Kaeppeler

the necessity of interrupting the growth process and removing the wafer for ex-situ measurements.

B £ 1 o

-50

-40

-30

-20 -10 0 10 Radius [mml

20

30

40

50

Fig. 4.3. The evolution of the growth rate.

Qa&

; i w : $ ^ ':\*'•';'.* « ^ ' ',;f,-1'r

Fig. 4.4. Cross section of the mechanical setup of an Epitune® II sensor head.

Two of the prerequisites for in-situ technology in a mass production system are the seamless integration of the in-situ tool into the MOCVD system's hard- and software and the absence of interference of the measurement with the chemical and physical environment of the epitaxy

Planar MOVPE - GaN

61

process. Modern in-situ tools utilize minimal invasive access ports to the reaction chamber. Figure 4.4 shows the cross section of the mechanical setup of an Epitune® II sensor head on the reactor lid of an ADC 2400G3 HT Planetary Reactor® system. The sensor's lens focuses the optical rays through a small hole in the "ceiling" of the reaction chamber onto the wafer surface. A small purge flow through the hole prevents clogging and unwanted growth and ensures that the influence of the presence of the in-situ tool on the epitaxy process is negligible. The graphite susceptor with the individual wafer disks rotates at speeds around 10 rpm, ensuring that every wafer passes through the scanning beam 6 times per minute. An angle transmitter at the susceptor's rotation axis makes sure the position of the susceptor and the wafers under the beam are known at all times. 5iqtBrvtaBi bdri

B3PB17m3B&3i El PU

step 71

T1iw'£57

Himnieg

Fig. 4.5. An epitaxy system control software.

The epitaxy system's control software (see figure 4.5) records the traces, provides tools for expert evaluation and archives them together with the log files of the respective epitaxy run. Like this, irregularities in the results of processed devices can be traced back to the in-situ records of the layer structure on the particular wafers during growth.

Dauelsberg, Schineller, Kaeppeler

62

5. The Mass Production of GaN and Related Materials The requirements for the mass production of nitride based materials and devices are mainly driven by device prices and economic margins. Epitaxy as the key technology and starting point for any device production process must continually address these issues by reducing the cost of ownership (CoO) and improving the material quality to open up new markets and device performance ranges. All experimental results were achieved on AIXTRON multiwafer Planetary Reactor® large scale production MOCVD systems at AIXTRON's Material Research Laboratory. 5.1. Optoelectronic Device Structures Light emitting diodes (LED) have become the driving force for the optoelectronics industry. Optoelectronic applications based on nitride semiconductors are ranging from green through blue to ultra-violet (UV) LEDs, each requiring challenging conditions in the MOCVD process. The ternary InGaN material system used in the quantum wells of such devices is highly susceptible to slight temperature changed during growth. Its overall In composition and the distribution of the In-rich regions in the Ga-rich InGaN matrix are highly affected by the local growth temperature, resulting in high demands for the temperature stability and uniformity of the epitaxy tool. ry 1090 -

-20

0 20 radius [mm]

Fig. 5.1. Pyrometrically measured temperature profile.

60

Planar MOVPE - GaN

63

With this in mind, modern epitaxy tools must be checked against these requirements. Figure 5.1 shows the pyrometrically measured temperature profile along the diameter of a 4 inch wafer disk at a reactor control temperature of 1100°C, a pressure of 200 mbar and under realistic flow conditions of 10 slm NH3 and 10 slm H2. As can be seen, a temperature uniformity of less than +/-2°C could be achieved across the wafer disk. Since this disk can hold either one 4 inch or three 2 inch wafers, an excellent temperature uniformity for both wafer sizes can be provided to facilitate homogenous growth. 5.2. Growth in the 8x4 inch Configuration The need for increased production and reduced CoO has lead to the introduction of multiwafer reactors capable of processing more wafers at the same time. Modern epitaxy systems today provide the capacity of up to 24x2 inch wafers (e.g. AIXTRON's AIX 2600G3 HT). The well established 2 inch technology, however, shall not be covered in this chapter. Instead, we will focus on larger wafer sizes. Whilst 3 inch wafers have not reached high market penetration, 4 inch wafers pose a viable alternative. Easy transfer of existing processes from 24x2 inch to 8x4 inch reactors is possible, since both reactor types are essentially identical in terms of the geometry of the reactor chamber. Hence, similar process flows and temperatures can be used to grow on 4 inch. 5.3. Special Aspects of Growth on 4 inch The growth on 4 inch sized wafers poses special challenges to the growth of highly strained material, such as GaN. The difference in thermal expansion coefficients and lattice constants between the growing GaN layer and the sapphire substrate causes the wafers to bow during growth, resulting in loss of contact to the graphite substrate holder. This, in turn, leads to thermal gradients on the wafer surface, influencing the growth of temperature sensitive material such as InGaN. In addition, special care

64

Dauelsberg, Schineller,

Kaeppeler

must be paid to the manufacturing process of the sapphire raw material, the cutting of the sapphire core and the surface preparation procedure, to reduce the amount of built in material strain and stress, since these effects can cause the wafer to bow and warp in an unpredictable fashion. While those aspects, in general, also apply to the growth on 2 inch, their effects on the growth on smaller wafers is negligible. Given a substrate with perfectly uniform strain, the growth induced bow can be indirectly measured from the wavelength uniformity of an InGaN MQW structure. Prior to these experiments the gas phase and total flow conditions in the reactor must be tuned to provide perfect temperature, decomposition and precursor depletion conditions. Hence, process conditions must be chosen which would yield perfectly uniform wafers under non-bowing conditions. 30

25

T20

\\ —

\\

\\

s

\s

-•-

:x

115

S10

-

— I

F:;::: r — • • " — •

400

500

600 700 Wafer Thickness [|jm]

800

900

Fig. 5.2. "Bow indicator" as a function of substrate thickness.

Figure 5.2 shows the "bow indicator" as a function of substrate thickness for substrates with 500 urn, 640 um and 850 um. The "bow indicator" is a custom defined figure of merit which is based on the wavelength uniformity of the wafer and takes into account the different sensitivities of In-compositions towards the miscibility gap of InGaN. As can be seen, a clear dependence of the wafer bow during growth of the

Planar MOVPE - GaN

65

MQW structure on the thickness of the substrate can be observed. This, in turn, allows for the adaptation of the wafer pocket once the thickness of the substrate and the growing structure are known. With the possibility of substrate bowing and consequent loss of thermal contact to the disk in mind, an additional aspect to be paid attention to is the process temperature window of the sapphire surface for the nucleation and growth of the GaN buffer layers. This aspect of GaN growth has not been investigated to the full extent in the past. Since the properties affecting the process window are local surface properties, we investigated the influence of the off c-axis mis-cut as one of the most prominent wafer design parameters to be influenced during the fabrication of the substrates. C-plane substrates with the deliberate miscuts (on axis, 0.l°->m-plane, 0.2°->m-plane, 0.3°->a-plane and 0.4°->mplane) were overgrown with GaN under three different growth temperatures (1100°C, 1130°C and 1160°C) and characterized by optical microscopy and sheet resistance measurements using a Lehighton mapper.

Fig. 5.3. Optical micrographs for GaN layers grown.

Figure 5.3 shows optical micrographs for GaN layers grown at 1130°C on all four investigated mis-cuts (on axis morphology was similar to the 0.1° case). As can be seen, growth on orientations close to

Dauelsberg, Schineller, Kaeppeler

66

the c-axis is irregular with a rough microstructure. At mis-cuts of 0.2° and higher, smooth surfaces with terraces can be observed. The alignment of the terraces is marked with a white line for better visibility. Please note that the alignment of the terraces is different in case of the 0.3° material since the mis-cut of the underlying sapphire was towards the a-plane. In addition to the examination of the morphological structure by microscopy, the measurement of the sheet resistance gives an insight into the distribution and mobility of the carriers in the crystal. Figure 5.4 shows the average sheet resistance of the layers as a function of the growth temperature. As can be seen, wafers with mis-cuts of 0.3° and 0.4° have the lowest and most consistent sheet resistance over the investigated temperature range. Layers on mis-cuts of 0.2° show a clear increase of the sheet resistance with lower temperatures, indicating that the temperature window for growth on those substrates is smaller and closes at lower temperatures. Layers with on axis alignment and 0.1° mis-cuts are not shown on this plot since their sheet resistance was found to be a factor of 5-10 higher and strongly depends on the growth temperature. 1100 1050 ,1000 r

i 950 — 900 850 5> 800 % 750 \ .c 700

(/>

650 600 1 1090

1100

1110

1120 1130 1140 Growth Temperature [°C]

Fig. 5.4. Average sheet resistance of the layers.

1150

1160

1170

Planar MOVPE - GaN

67

From these results, one can conclude that mis-cuts of 0.3-0.4° are best suited for the growth of GaN, especially when temperature variations across the wafer surface are expected due to substrate deformation during growth. 5.4. The Growth oflnGaN

Structures

As discussed in the previous paragraphs, the growth of InGaN structures on 4 inch sapphire requires special attention to the quality of the sapphire and the design of the wafer holder in the reactor chamber. However, when deducing design parameters from the results on wafer bow elucidated above, good layer structures with excellent yield on wafer can be grown. . in ii i

nm 475.0 472.5 470.0 467.5 465.0 462.5 460,0 457.5 455.0 1452.5 1450.0

•*$.

.12

: 462.6

" 'ian : 462.7 " Dev: 0.281 % (1.300)

Fig. 5.5. Room-temperature photoluminescence mapping.

Figure 5.5 shows the room temperature photoluminescence mapping under low excitation conditions of a 5 period (In)GaN MQW structure grown on 4 inch using an adapted 4 inch wafer holder. The edge exclusion for this mapping was 0 mm. As can be seen, excellent wavelength uniformities with a standard deviation of 1.3 nm at 462.6 nm average wavelength can be achieved. The long wavelength spots at the anti-flat positions most probably stem from a disturbance of the lattice due to the tweezer-handling of the wafer during loading of the substrate. The slight non-rotational symmetry is attributed to internal strain fields

68

Dauelsberg, Schineller, Kaeppeler

in the sapphire substrate, which most probably stem from the polishing process. CoO is, among factors such as precursor utilization, mostly driven by the yield on the wafer. Taking a specification of +/-3 nm more than 96% or 7540 mm2 of the wafer surface can be used for device processing. 5.5. The Growth of AlGaN Structures The AlGaN ternary alloy is less susceptible to temperature variations than InGaN since it does not have a miscibility gap for the technically relevant composition ranges.24 However, AlGaN poses the challenge of possible pre-reactions between the TMA1 and NH3 precursors which can lead to formation of adducts in the gas phase. This leads to a reduction of the growth rate due to premature consumption of species and the formation of particulates. Careful reactor design and process optimization must, therefore, be used to avoid such pre-reactions. Figure 5.6a) shows the white light interference (WLI) thickness mapping of Alo.3Gao.7N grown on a 4 inch sapphire wafer grown at 30 mbar process pressure and a V/III ratio of 650-850. The layer consisted of an approximately 500 nm thick AlGaN layer grown onto an 850 nm strain relieving buffer structure composed of alternating GaN and A1N layers. Those interlayers are necessary to prevent cracking in thick AlGaN due to the strain that builds up between the epilayer and the substrate. As can be seen from the mapping, the resultant AlGaN was uniform in thickness and free from particles which would show up as erroneous points in the mapping. In addition, the growth rate measured by the insitu Epitune® II reflectometry kit during growth was found to be 1 um/h, showing that the reactor design in combination with the chosen process conditions actively prevents pre-reactions from happening. It is believed that the total pressure of 30 mbar effectively suppresses pre-reactions. Further experiments also show that the growth rate drops as the total pressure is increased.

Planar MOVPE - GaN

69

The distribution of the Al-concentration across the wafer was also found to be uniform. XRD measurements showed a mean Alconcentration of 32.9% and a standard deviation of 0.9%.

0.600 0 570 0 540 :0 510 0 480 |0.450 l O .420 •0.390 •0.360 •0.330 'o.300 • |e : 1.355 .1-dian : 1.348 •••• Dev: 2 . 1 2 4 % (0.029)

a)

e : 0.456 Jian : 0.466 Dev: 5.508 % (0.025)

b)

Fig. 5.6. a) White light interference (WLI) mapping, b) WLI thickness mapping thickness of an A1N layer.

Figure 5.6b) shows the WLI thickness mapping of an A1N layer grown directly onto a 25 nm thick low temperature GaN nucleation layer at a growth rate of 0.92 um/h. Temperature, total pressure and V/III ratio were similar to the growth conditions of the AlGaN layer. It can be seen that the total thickness was measured to be 400 nm. The layer was crackand particle-free. These experiments show that the growth of Al(Ga)N on 4 inch sapphire is feasible and commercially viable for mass production. 5.6. High Growth Rates in Mass Production Reactors One of the major factors influencing the CoO of epitaxy systems is the throughput, hence, the size of the usable device area that can be deposited in a given amount of time. Consequently, the achievable growth rate becomes an important issue, especially in structures that require thick buffer layers, such as LEDs or HEMTs. Depending on the layer design, more than 60% of the total growth time can be consumed

Dauelsberg, Schineller, Kaeppeler

70

by those layers. However, an increase of the growth rate is often limited by kinetics on the layer surface which could lead to incorrect incorporation of atoms, hence the formation of vacancies and interstitials. All those effects can also lead to rough surfaces.

0

2000

4000 6000 Q(TMGa) [nmol/min]

8000

Fig. 5.7. Dependence of the growth rate of GaN for different TMGa molar flows.

Figure 5.7 shows the dependence of the growth rate of GaN for different TMGa molar flows. Typical growth rates of GaN in MOCVD reactors are on the order of 2 um/h, yet this experiment on an AIX 2600G3 HT Planetary Reactor® in the 8x4 inch configuration shows that the growth rates can well exceed the 10 um/h range. This reduces the growth time for a 5 |im thick GaN buffer from 2.5 h to 30 min. References 1. I. Akasaki, H. Amano, K. Itoh, H. Sakai, T. Tanaka and K. Manaba, Inst. Phys. Conf.Ser. 129, 851 (1992). 2. S. Nakamura, T. Mukai and M. Senoh, Appl. Phys. Lett. 64, 1687 (1994). 3. H. Simka et al., Prog. Crystal Growth and Charact. 35, 117 (1997). 4. D. F. Davidsson, K. Kohse-Hoinghaus, A. Y. Chang, R. K. Hanson, Int. J. Chemical Kinetics 22, 513 (1990). 5. H. Amano, N. Sawakai, I. Akasaki, and T. Toyoda, Appl. Phys. Lett. 48, 353 (1986).

Planar MOVPE-GaN

71

6. P.M. Frijlink, A new versatile, large size MOVPE reactor. J. Crystal Growth, 93:207-215 (1988). 7. S. Yu. Karpov, R. A. Talalaev, Yu. N. Makarov, N. Grandjean, J. Massies, B. Damilano, Surface Science 450, 191 (2000). 8. T. G. Mihopoulos, V. Gupta, K. F. Jensen, J. Crystal Growth 195, 733 (1998). 9. S. W. Bland, J. Mat. Scu: Mat. in Electronics 13, 679 (2002). 10. G. B. Stringfellow, Organometallic Vapor-Phase Epitaxy, Academic Press, 1999. 11. H. Simka, B. G. Willis, I. Lengyel, K. F. Jensen, Prog. Crystal Growth and Charact. 35, 117(1997). 12. R. T: Lee and G. B. Stringfellow, J. Electronic Materials 28, 963 (1999). 13. A. Thon, T. F. Kuech, Appl. Phys. Lett. 69, 55 (1996). 14. T. Mihopoulos, Reaction and transport processes in OMCVD: Selective and groupIll Nitride growth, PhD thesis, MIT, 1999. 15. C. H. Chen, J. Electronic Materials 25, 1004 (1996). 16. J. R. Creighton, G. T. Wang, W. G. Breiland, M. E. Coltrin, J. Crystal Growth 261, 204 (2004). 17. C. R. Kleijn, Chemical vapor deposition processes, in M. Meyyappan, editor, Computational modeling in semiconductor processing, pp. 97-230, Artech House, 1995. 18. R. B. Bird, W. E. Stewart, E. N. Lightfoot, Transport phenomena, John Wiley & Sons, New York, 1960. 19. S. V. Patankar, Numerical Heat Transfer and Fluid Flow, Hemisphere, 1980. 20. J. H. Ferziger, M. Peric, Computational Methods for Fluid Dynamics, Springer, 1996. 21. M. F. Modest, Radiative Heat Transfer, McGraw-Hill, New York, 1993. 22. H. Schlichting, Boundary Layer Theory, McGraw Hill, New York, 1960. 23. M. Dauelsberg, H. Hardtdegen, L. Kadinski, A. Kaluza, P. Kaufmann, J. Crystal Growth, 223, 21 (2001).. 24. G. Steude, B. K. Meyer, A. Goldner, A. Hoffmann, F. Bertram, J. Christen, H. Amano, I. Akasaki, Applied Phys. Lett. 74 (17), 2456 (1999).

This page is intentionally left blank

CHAPTER 3 CLOSE-COUPLED SHOWERHEAD MOCVD TECHNOLOGY FOR THE EPITAXY OF GaN AND RELATED MATERIALS

Edward J. Thrush, Adam R. Boyd Thomas Swan Scientific Equipment Ltd., Anderson Road, Buckingway Business Park, Swavesey, Cambridge CB4 5FQ, UK. E-mail: ejthrush @ thomasswan. co. uk This chapter is concerned with the growth of GaN based structures in Close-Coupled Showerhead (CCS) MOCVD Reactors. The technical and financial issues that prompted the evolution from simple horizontal tube reactors to multi-wafer reactors are discussed and the merits of CCS technology are outlined. The principles determining the characteristics of CCS reactors are considered both analytically on idealised simplified systems and more rigorously using modern computational fluid dynamic techniques on real system architectures. A comparison between the predictions of such simulations and practical performance is outlined and in particular the suggestion that such reactors should have simple scaling rules seems to be borne out by experimental data. A section is included on the important topic of insitu monitoring. This includes a description of the optical techniques used to monitor growth rate, growth temperature and morphology and an ultrasonic technique for gas composition analysis. A selection of growth results are provided from CCS reactors of different wafer capacities and a section is included on production related issues and cost of ownership.

73

74

Thrush, Boyd

1. Introduction The group III nitrides represent an important class of materials with applications in high temperature electronics, optoelectronic devices operating in the visible and UV region of the spectrum, and cold cathodes. Owing to their combination of wide band gap, high thermal conductivity, high thermal stability and physical robustness they offer, in principle, an attractive route to such devices. There are however severe practical difficulties in their preparation which have prevented the realisation of their potential until relatively recently. Probably the most significant problem has been the absence of a reasonably priced, lattice matched, expansion matched and thermally stable substrate material which together with the difficulty of providing an adequate over pressure of active nitrogen during growth, has constituted a major hurdle to progress. The first epitaxial layers of GaN were grown in 1969 by Maruska and Tietjen1 using chloride vapour phase epitaxy. The material produced from this process, (which employed metallic gallium, HC1 and ammonia as the starting materials), was characterised by a very high n-type background doping level in the 1019 cm"3 range. Three years later Manasevit et al.2 using the then embryonic technique of Metal Organic Chemical Vapour Deposition (MOCVD) succeeded in preparing the nitrides of both gallium and aluminium. Work on nitrides continued at a modest level for the next two decades until the spectacular breakthrough of Nakamura et al.3 of Nichia Chemical Industries, whose results re-kindled a world-wide interest in the subject. Using a variant of the MOCVD process in which a radially injected stream of the reagents trimethyl gallium (TMG) and ammonia, was forced on to the horizontal surface of a rotating wafer by a downward "sub flow" of nitrogen and hydrogen, they were able to grow structures which allowed the fabrication of very bright violet and blue LEDs.

Close-Coupled Showerhead MOCVD - GaN

75

In parallel with the explosion of activity in the field of MOCVD nitride growth which resulted from this success, aspirations have developed amongst the major players for multi-wafer equipment which would out perform the then current state of the art "research scale" reactors in terms of both uniformity and output. It is in response to this perceived demand that the Close-Coupled Showerhead reactor was developed to provide a readily scaleable growth platform. 2. Reactor Development As with the epitaxial growth of all III-Vs, the uniformity of thickness and composition of GaN based structures will be controlled by the flow patterns prevailing within the reactor and the spatial variations in the susceptor temperature. Although it is possible to tune "asymmetric" reactor configurations4 to achieve uniformity of a particular parameter, there is normally a compromise to be struck when attempting to optimise more than one characteristic at a time. For instance it is possible to compensate for the gas phase depletion of reactants in the direction of flow in a horizontal reactor, by progressively reducing the cross sectional area in the flow direction. Such action would increase the gas velocity and reduce the thickness of the "boundary" layer in the flow direction, thus providing a means of restoring the falling growth rate which would otherwise accompany the reduction in reagent concentration as the gas stream travels along the susceptor and the reagents are depleted by the deposition process. Whilst this may be a conceptually satisfactory gambit for the growth of binary compounds, where there is only one growth limiting reagent present in the gas stream, it is less satisfactory for more complex alloys or structures containing a number of different alloys (as device structures do), as the depletion characteristics will vary from one precursor to another. This will mean that only empirical trade-offs are possible. In addition reactors of this type are prone to thermal buoyancy driven recirculation cells and longitudinal rolls5 which can compromise both uniformity and interface quality, a tendency which becomes more serious as the size is increased.

76

Thrush, Boyd

In contrast to this situation, vertical axially symmetric, rotating disc "stagnation point" reactors can intrinsically achieve very uniform "boundary layers", deposition rates and higher alloy compositions.

Local horizontal flow vectors

Fig. 1. Streamlines, velocity profile and boundary layer for an idealised stagnation point rotating disk reactor (after Schlichting).

2.1. Close-Coupled Showerhead Technology The concept of the Close-Coupled Showerhead reactor is based upon a mathematical analysis of fluid flow undertaken more than 40 years ago. In a treatment by Schlichting6 which employed an analytical solution of the relevant Navier-Stokes equations to describe the flow streamlines, velocity distribution and boundary layer arising when a uniform flow of gas stalls against a perpendicular flat surface, it was shown that the boundary layer thickness is independent of the distance from the stagnation point (i.e. the point at which the flow separates to pass around the object in the path of the flowing gas). The boundary layer being a notional layer of gas, adjacent to the solid surface, through which the velocity increases from zero, at the surface, to the free stream velocity. A diagrammatic representation of this situation is given in Fig.l. As a conceptual aid it may also be worth considering some specific findings of this analysis, namely that the boundary layer thickness (5) is inversely

Close-Coupled Showerhead MOCVD - GaN

77

proportional to the square root of the vertical Reynolds number as shown in equation (1) and with the vertical Reynolds number being defined by the equation (2).

M u = flow velocity, D = chamber diameter, H = chamber height, r = gas density, n = dyn. viscosity

It is implied (as all of the terms in equation 2 other than " u" are fixed in the case we are considering) that as the gas flow direction veers away from the vertical towards the horizontal, and the magnitude of the velocity increases away from the stagnation point it does so by an increase in the horizontal component of the flow vector with the axial vertical component remaining constant as shown in Fig.2. Axial component

( f ) independent of radius

Fig. 2. Flow vectors in stagnation point vertical rotating disk reactor.

As the diffusion of growth precursors through the boundary layer, in the classical mass transfer limited regime commonly used for MOCVD, determines the growth rate and composition of higher alloys, its invariance with position on the susceptor implies that reactors of this geometry should inherently give good uniformity. It should be noted however that the analysis discussed previously applies to an idealised situation where the susceptor is immersed in an infinite volume of gas, at a fixed temperature, which flows toward the susceptor with no other

78

Thrush, Boyd

constraints. In practical reactors there are generally a number of deviations from this ideal situation, which complicate the simple concept. One of the most important of these is a perturbation of the boundary layer where the flow of gas bends over the edge of the susceptor. This local distortion of the boundary layer is less for reactors having a squat aspect ratio than for tall reactors where, in the limit, the flow streamlines adopt a "wine bottle" shaped profile as the gas is channelled over the susceptor edge, thus preventing the formation of a uniform boundary layer. A further perturbation is caused by a "wall drag" effect, (caused by the fact that the gas in contact with the reactor chamber wall must be stationary), that imparts a parabolic velocity profile to the gas stream flowing from the entry point to the deposition zone. The severity of this perturbation increases progressively as the distance from the injection point to wafer is increased, again being worse for tall reactors. Traditional designs which require the injected gas to spread into the reaction zone from a narrow inlet are also plagued by "jetting" effects which can lead to recirculation cells at the gas injection point. Additionally thermal buoyancy driven recirculation cells which arise from the change in density of the gas as it enters the hot reaction zone can also degrade both the uniformity and the interface quality of the structures grown. Finally for the MOCVD growth of III-V compounds from group III metal alkyls and group V hydrides the simple pyrolysis based chemistry is complicated by a parasitic low temperature reaction between these two types of reagent (the formation of addition compounds) which can undermine the control and chemical efficiency of the process and give rise to particulate contamination. A feature of the CloseCoupled Showerhead (CCS) reactor, shown diagrammatically in Fig.3, is that it provides the means of minimising these unwanted second order effects such that the idealised behaviour can be closely approximated to, in a simple manner. This is done by introducing the reagents across the whole of the reaction zone from a multiplicity of separate injector tubes, embodied in a water-cooled showerhead, located very close to the growing wafers. The graphite susceptor is in the form of a short inverted cylinder, open at the bottom, which rotates about a fixed multi-zone

Close-Coupled ShowerheadMOCVD - GaN

79

spiral heater. In the first CCS reactors, built to grow GaAs and InP based materials this heater was constructed from graphite. However for the growth of GaN based materials where a marked reaction between graphite and ammonia at high temperatures occurs, coating of the susceptor with protective SiC becomes mandatory and a different design of heater based upon tungsten elements supported by pyrolytic boron nitride insulators is used. The distribution of power between the different elements of the zoned heater can be adjusted within the control software to permit flat susceptor temperature profiles to be realised across the temperature range used for III-V epitaxy. The fact that the injection area is matched to that of the deposition zone mitigates against unwanted macroscopic jetting effects, whilst the close proximity of the injector to the growing wafers limits the intrusion of the susceptor edge distortion of the boundary layer as well as the scope for the development of the parabolic velocity profile referred to earlier. The proximity of the showerhead to the reaction zone also suppresses thermal buoyancy driven instabilities, which tend to be exacerbated in reactors having a large volume above the hot susceptor. For the growth of GaN based materials the reagent injector tubes of the showerhead are fed from two separate plenum chambers. In the latter arrangement the longer tubes from the upper plenum chamber and the shorter ones from the lower chamber are arranged on two interleaved matrices such that efficient mixing of the gases can be effected right at the edge of the reaction zone whilst still allowing complete mixing prior to the point of pyrolysis. The latter assertion is supported by both inhouse, and independent7 fluid dynamic modeling which has indicated that under typical operating conditions the gases from the two plenums are fully mixed, and the individual jets from the injection tubes are fully dissipated, by the time they arrive at the wafer surface. To appreciate the issues relating to showerhead design the generic chemistry of group III alkyls and group V hydrides need to be considered. It has been known for some time that low temperature reactions leading to the formation of low volatility addition compounds can occur8 which compromises the control of the reagents in the gas phase and can lead to particulate problems. In

80

Thrush, Boyd

the initial implementation of the Close-Coupled Showerhead, used for the growth of GaAs and InP based materials, a single plenum chamber was preferred as the pre-reaction between the group III alkyls and the group V hydrides (arsine and phosphine) was not severe and because many of the alloys and structures were sensitive to compositional uniformity on the group V sub-lattice (better spatial uniformity of gas distribution could be obtained by injection from the top plenum). However for GaN based materials where there is only one group V element, and the parasitic addition compound formation between the strongly basic NH3 precursor and the Lewis acid precursors used for the group III elements is more severe, the two plenum design is preferred. The reduced interaction afforded by the "late mixing" inherent to the two plenum showerhead yields a growth process with a high chemical efficiency, which together with the reactor's relative mechanical and operational simplicity contributes to a low cost of ownership.

Top plenum cham ber Gas inlet bottom plenum chamber Three zone heater Thermocouple

Fig. 3. Schematic diagram of 19x2" Close-Coupled Showerhead reactor.

2.2. Modeling Although the conclusions drawn from the analysis of ideal systems described in Section 2.1 encouraged the early investigation of the CloseCoupled Showerhead concept, it was obvious that this simplified approach would not be able to capture the detailed characteristics of such

Close-Coupled Showerhead MOCVD - GaN

81

reactors. To this end simulations were undertaken using a computational fluid dynamics package. The results of this exercise supported the suppostion that Close-Coupled Showerhead reactors should intrinsically yield uniform deposition characteristics. This package embodies the basic theory of gas flow dynamics and links it with algorithms to quantify heat flow, temperature distribution, and the progress of the growth chemistry at different locations within the reactor. It is worth noting that the growth chemistry is of particular importance in the simulation of the growth of group Ill-nitrides and is probably also the area of greatest uncertainty. This is due to the propensity of the electrophilic Lewis acid group III precursors to form addition compounds with the group V precursor, ammonia, owing to the latter's strongly basic character. Such reactions are reported to occur at low temperatures9 with the bonds formed by the sharing of the ammonia's lone pair electrons with the electrophilic group III alkyl molecule being relatively weak. This means that although the adducts which initially form can severely modify the chemical composition of the growth ambient (they have low vapour pressures compared to the precursors proper) they can nevertheless act as precursors themselves providing they can transport to regions of the reaction zone where the temperatures are adequate to promote their pyrolysis. However an alternative reaction pathway also exists where these weakly bound compounds are transformed into much more stable entities by alkane elimination (such as methane in the case of the (CH3)3Ga-NH3 adduct to leave dimethylgallium amide i.e. (CH3)2Ga-NH2 adduct. In this case the dative bond of the adduct is replaced by a stronger covalent bond leading to a relatively involatile compound which is even more difficult to pyrolyse. Such compounds can be further stabilized by the formation of cyclic oligomers where direct bonding between group III and group V atoms occurs. As the molecular weight and thermal stability of these various compounds increase the tendency for them to agglomerate in the gas phase and to form nano-particles also increases. Such nano-particles can deposit onto the growing wafer or be thermophoretically levitated11 and swept into the reactor exhaust. It is known that the propensity to follow a particular chemical pathway is

Thrush, Boyd

82

strongly dependent upon the material involved (i.e. the growth of aluminium containing materials is more likely to promote particle formation than is the case for the other group III compounds) and the growth conditions adopted (particularly the pressure). Owing to the potential complexity and high impact of this underlying chemistry it poses a significant challenge to those involved in the simulation of group III nitride growth and currently represents an area of intense activity.

6

0 mm from gas inlet 1 mm from gas inlet 5mm from gas inlet

5 !

4

ammonia plenum

o

MO plenum

> "TO

•S 2 1

i/M

CD

>

,

•

PJ. ifa-r^-,..-'-,,, •..T-V ; Lateral position on showerhead Fig. 4. Dissipation of individual jets away from showerhead surface.

In order to facilitate large scale modeling it is important to establish what approximations can be made without materially affecting the competence of the model. An example of such a determination was the detailed modeling to establish the nature of the flow profile in the vicinity of the showerhead outlet. The results of this investigation are shown diagrammatically in Fig.412. Here we see representations of the vertical velocity profile across a small section of the showerhead at distances of 5 mm (solid line), 1 mm (coarse dashed line) and 0 mm (fine dashed line) from the showerhead surface. From this analysis we have been able to establish that the velocity profiles from the individual jets are not discernible at distances beyond 5 mm from the showerhead surface for the outlets from the ammonia plenum and 2 mm from the showerhead

Close-Coupled Showerhead MOCVD - GaN

83

surface for the jets associated with the plenum supplying the metal organics. This means that the showerhead can be treated as a "porousplug" for the larger scale simulations needed to look at reactor scale uniformity and to carry out the parametric studies necessary for the optimisation of operating parameter space. Using such simulations it is convenient to spatially map the distribution of intermediate breakdown products of the growth precursors, under various operating conditions and to combine this with information depicting the temperature distribution within the reaction zone, as a means of predicting the likely compositional and thickness uniformity of the materials grown. It was found that the temperature and the concentration of the growth intermediary monomethyl gallium were very uniform within the reaction zone for the different sizes of reactor modeled. This result coincides with the observation that similar uniformities of layer characteristics are attainable across the range of multi-wafer CCS reactors. 1

-^3.0

1

1

1

1

1

1

<

1

'

—I

'

— i — < — i — < — i — • — i — • — i — • — i —

3.5 : p = 50Torr,T = 1030°C

; p = 100 Torr, T = 1030°C

^

model prediction

^-"-"''-^

^ 1^3.0

experimental

:

-3=2.5 0) TO

^~*^'

g2.5

»-0'

-c2.0

I

o 6 1-5 1.0

-

sz

v

"

?2.0 o

experimental

model prediction

< S 1.5 . -

•

i

25

i

i

i

i

•

i

i

i

30 35 40 45 Total flow rate (slm)

i

•

;

i

50

1.0

25

30 35 40 45 Total flow rate (slm)

50

Fig. 5. GaN growth rate Vs total flow rate in 19x2" reactor at (a) 100 Torr, (b) at 50 Torr.

In addition to this form of generic modeling, used to test and develop concepts of reactor design, the code can also be use to provide guidance in the navigation of the parameter space necessary to develop specific aspects of process performance. Examples of such applications are given in Figs.5 (a) and 5 (b) which indicate the variation in growth rate of GaN as a function of the flow rate of gas through a 19x2" CCS reactor at 100 Torr and 50 Torr respectively and Fig.6 which illustrates the linear dependency of growth rate on the molar input of the gallium

84

Thrush, Boyd

precursor (TMG) to the reactor. As can be seen the modeling in general captures the trends of these parametric dependencies quite well although some offset of the experimental values from those predicted implies that further refinement of the models will be worthwhile. 4.0 £-3.5 33.0 o 03 2.5

f 2.0 g C5 1.5 1.0 40

60

80

100

120

140

TMG bubbler flow rate (seem)

Fig. 6. Growth rate Vs TMG molar flow rate in 19x2" CCS reactor.

Modeling has also been used to quantify the "user friendliness" of the CCS reactor. One of the notable features of this design is the relative insensitivity of the reactor's performance to such parameters as total process flow (i.e. it is not critical to finely match this parameter to other growth conditions) and reactor geometry. An appreciation of this facet of the design can be made by the inspection of "performance diagrams" of the type described by Pawlowski et al.13 This analysis uses threedimensional plots to relate an important characteristic of the grown wafer (such as growth rate variation across it) with two other principal parameters used in the growth process (such as showerhead spacing and flow velocity of the gases flowing through the reactor). In such diagrams, which present the resultant "wafer" characteristic as a false colour map drawn on an area defined by the two growth parameters (on the "x" and "y" axes) it is desirable to have a large area of the diagram over which good performance can be obtained as this tends to facilitate the exploration of a wide region of parameter space and also makes the initial commissioning and set-up relatively straightforward. It has been found

Close-Coupled Showerhead MOCVD - GaN

85

that modern CCS reactors are particularly good in this respect. The ability to readily scale from the original multi-wafer 3x2" reactor to a 19x2" wafer variant, whilst maintaining a state-of-the-art performance, illustrates that the close spaced showerhead concept effectively addresses the various non-idealities discussed earlier, and the large "process window" intrinsic to this type of reactor allows robust growth process "recipes" to be formulated. 2.3. Scale-up of CCS Reactors As discussed previously the Close-Coupled Showerhead arrangement intrinsically yields a uniform boundary layer thickness irrespective of position on the susceptor radius, which in turn provides the foundation for uniform deposition. A corollary to this argument is that the design should produce these desirable characteristics on any length scale implying that there are no conceptual limitations to the realisation of good uniformity, even for very large reactors of this type. Moreover it can be argued that larger diameter reactors should have enhanced performance as any residual second order perturbations of the gas flow caused by the proximity of the reactor wall, or rotation induced recirculation cells at the susceptor edge, should have a proportionately smaller effect as the reactor size is increased. This reasoning has been borne out by detailed fluid dynamic modeling (as discussed in Section 2.2) which provided the reassurance necessary to invest the considerable engineering effort required to increase the size from the original single 2" wafer variant to the current large capacity multi-wafer reactors. Results from reactors ranging in capacity from 1x2" through 3x2", 6x2" to 19x2" have shown in fact that it is easier to optimise the larger reactors for highly uniform, chemically efficient, deposition than is the case for their smaller counterparts, although all of the multi-wafer reactors have produced state-of-the-art material specifications as discussed in Sections 5.1-5.3 and 6.3. In addition to the more favourable gas flow pattern discussed earlier another factor which makes the larger reactors easier to use is that thermal distinction between the separate zones of the multi-

86

Thrush, Boyd

zone heater is more easily achieved as the size is increased. This implies that the susceptor temperature uniformity can be more easily controlled which has a positive impact on the uniformity of parameters which are very temperature sensitive such as the composition of InGaN, and the growth and annealing of low temperature nucleation layers on which most GaN based structures are grown. 3. In-situ Optical Monitoring In order to navigate the complex parameter space employed for the growth of group III nitride based structures it is important to have some real-time insight into the progress of the growth process. Also the growth temperature is such a critical parameter that it is paramount to have an accurate measurement of it as close as possible to the growing wafer. To this end a number of in-situ optical monitoring tools based upon radiation pyrometry and reflectometry have been developed and adapted for use in the CCS reactor geometry. As a significant interruption of the injector tube matrix used in the showerhead design can have a deleterious effect upon process uniformity.13 It is important that these tools can work with minimal optical access.

3.1. Optical Pyrometry The most basic tool employed is an optical probe based pyrometer that is used to relate the control thermocouple reading to the temperature of the susceptor on which the wafers are carried. It also enables the spatial temperature profile of the susceptor to be set and its profile to be managed in the time domain. This arrangement is shown diagramatically in Fig.7.

Close-Coupled Showerhead MOCVD - GaN

87

Fig. 7. "Through showerhead" optical pyrometer set-up.

Radiation from the incandescent susceptor is collected by an array of quartz light guides (which protrude through penetrations to the showerhead matrix) and is relayed by flexible large core optical fibres to optical detectors which constitute the "front end" of the pyrometer. The photocurrents induced in these detectors (by the absorption of the fibre delivered radiation) is related to the temperature of the object "seen" by the optical probe by Planck's black body radiation equation modified by a factor related to the optical coupling between the hot susceptor and detector. The initial stage of the calibration therefore involves accounting for the optical coupling variations on individual channels by calibrating the probe and fibre assemblies to a blackbody calibration source. This relates the "perceived" temperature measured by the pyrometer to a series of "real" temperatures that are set on the black body calibration furnace. The probes are then installed in the reactor and measurements are made at a range of settings of the reactor temperature controller to allow the control thermocouple settings to be related to the susceptor temperature

88

Thrush, Boyd

read on the pyrometer and, by reference to the probe calibration data, to the "real" temperature. These measurements are made under the same sets of conditions as will be used during subsequent epitaxial deposition (e.g. different calibrations will apply when the reactor is being operated under a hydrogen ambient to when it is operating under nitrogen). During this exercise the opportunity is also taken to experimentally establish the heater zone settings (relative distribution of power between zones A, B and C of the heater) required to yield flat temperature profiles at a variety of key static temperatures as well as the dynamic evolution of the zone settings necessary to maintain a flat temperature profile when ramping from one temperature to another. This information is then used when constructing the software "recipes" used to run the reactor. It should be noted that simple optical pyrometry cannot be used accurately during layer deposition due to Fabry-Perot interference, caused by the growing layer superimposing thickness related oscillations on to the temperature related signal. In fact such pyrometric measurements have in the past been used to provide rudimentary in-situ interferrometric measurements of layer thickness, growth rate, morphology etc.14'15. 3.2. Laser Interferometry (Reflectometry) As mentioned in Section 3.1 the growing film modifies the optical nature of the wafer such that the reflectivity (or its converse the emissivity) varies in an oscillatory manner as the film grows. For the situation where light is reflected both at the top GaN surface and from the substrate/GaN layer interface as shown in Fig.8 the conditions for constructive and destructive interference are given by the well known interference equations 3 and 4 respectively. These expressions account for the difference in optical path length between directly reflected light of wavelength "A" and that traveling through the layer whose refractive index is "n" and whose thickness is "d", and also accounts for the phase change accompanying the reflectance at the top surface. For maximum reflected intensity: 2nd cos 9 = mA. (3)

Close-Coupled Showerhead MOCVD - GaN

89

For minimum reflected intensity: 2nd cos 9 = (m+ Vi) X. (4) Diode Laser

Fig. 8. Optical Interfaces caused by reflections at thin films.

In fact in the case of the interferometer used with the CCS showerhead reactor, where the probe beam and the reflected beam are normal to the wafer surface, the "cos 9" term becomes unity and the grown layer thickness corresponding to a complete oscillation of the reflectance signal is given simply by X/2n. photodiode

635nm filter beam splitter

^vs-4-v^

diode laser showerhead

m

M

m

^^__- wafer

susceptor

Fig. 9. Schematic of laser interferometer on CCS reactor.

90

Thrush, Boyd

As the layer grows the beams reflected from the substrate layer interface and the layer surface evolve through states of successive constructive and destructive interference (as detailed earlier) to produce the interferograms used to monitor the progress of the growth run. The diode laser interferometer set-up used on our CCS reactors is shown schematically in Fig.9. Light from a horizontally mounted 635 nm diode laser is directed onto the surface of the wafer via one of the standard pyrometer ports using a cube beam splitter. For this essentially "free space" application the light guide described in Section 3.1 is dispensed with and the fitting is closed with a polished quartz plug constituting the input and output window. Light reflected from the wafer surface passes back through the optical port and beam splitter to be collected by a photodiode detector located directly above the latter. A gated DC measurement is employed in which the laser is turned on as it passes over a particular part of a particular wafer (sampled area amounts to 7% of a complete revolution). In this arrangement the reflection signal is electronically sampled for the periods when the laser is turned off and is updated each revolution. Using such a strategy a measurement of the "dark signal" (when the laser is turned off) can be made and this is subtracted from the reflectance signal to provide immunity of the system to variations in any stray light entering the detector and to thus enhance the quality of the reflectance signal. In order to synchronise the signal processing with the movement of the wafers a timing pulse is provided by an encoder on the susceptor rotation drive mechanism. Using this the position of the sampling point on the susceptor can be accurately specified. A facility is also provided to (electronically) select the wafer to be studied and to permit the position which is monitored on the wafer to be adjusted. It has been found that the use of this "real time" monitoring facility has contributed greatly to the development and tracking of the growth process. For instance it is reported16'17 that the crystallographic quality of the thick GaN layers, or templates (on which device structures are subsequently deposited) can be enhanced by the use of a growth regime which promotes significant islanding of the low temperature GaN nucleation layer (roughening), during its subsequent anneal, and the use

Close-Coupled Showerhead MOCVD - GaN

91

(of procedures which extended the duration of the coalescence phase of the epi-layer proper. The interferogram shown in Fig. 10 (a) is characteristic of such templates, which are the preferred type for use in optical devices such as LEDs. On the other hand it is sometimes necessary to use templates that are of very low conductivity, such as in the buffer layers of HEMT devices.18 It has been shown that such electrical properties can be realised by managing the structure used in the initial stages of growth and the reactor conditions used for the recrystallisation of the low temperature nucleation layer and subsequent coalescence of the GaN grains so produced. Such growth conditions limit the roughening of the nucleation layer during its annealing phase and promote a rapid transition from a three dimensional to two dimensional, step flow, growth mode as shown in Fig. 10 (b). 60-

= bOty

a

40

«

tr

S 30 .y T-,

M •

m 10-

IX.

0-| 1800

, 2800

, 3800

, 4800

, 5800

, 6800

, . 7800

Time (sees)

Fig. 10. Interferograms from growth of (a) conducting and (b) semi-insulating GaN.

3.3. Emissivity Corrected Pyrometry As mentioned in Section 3.1 there are complications to the use of simple "optical power" based pyrometry during the growth of GaN layers. These difficulties are circumvented when using a scheme which simultaneously measures both the interference caused by the layer growth, and the incandescence power delivered to the detector, and corrects the latter for

92

Thrush, Boyd

the perturbations caused by the former. In other words as reflectivity is equal to one minus the emissivity, interference effects during growth lead to oscillations of the reflectance and hence to oscillations of the emissivity. This complicates pyrometric measurements of temperature by superimposing an oscillating "Fabry Perot" signal onto the incandescence signal. By the use of quasi-simultaneous reflectivity and incandescence measurements, with exact matching of the centre wavelength and bandwidth, the two effects can be separated and emissivity corrected pyrometry measurements yielding the real temperature can be performed. Tungsten lamp

950 nm (50 nm bandwidth) filter

Stepper motor driven shutter

,,J

OP 1

Input fibre

Tungsten lamp radiation 635 nm (5nm bandwidth) filter

"Beam splitter

Si photodiode i

Susceptor incandescence r showerhead t + t t • t

'- t wafer

I

susceptor

Fig. 11. Schematic diagram of emissivity corrected pyrometer (EpiTT).

An instrument based upon this scheme has been developed by LayTec, GmbH and is marketed under the trade name of EpiTT. In collaboration between LayTec and Thomas Swan the adaptation of such a system to work within the contstrained optical access afforded by a CCS system has been successfully carried out.19 A schematic diagram of this instrument is shown in Fig. 11. Light from a tungsten lamp is delivered to an optical head mounted above a penetration in the showerhead that is closed with a polished quartz plug (as for the reflectometer). From this

Close-Coupled Showerhead MOCVD - GaN

93

unit the light is directed onto the wafer via a beam splitter. Light reflected from the wafer, together with incandescence from the hot susceptor, is directed by the beam splitter into an output fibre and thence to a detector unit. This detector unit splits the light into two wavebands with the aid of band-pass optical filters centred at 950 nm and 635 nm. The longer waveband detection arm is used to effect the pyrometry measurement by measuring the incandescent power in a 50 nm wide waveband together with the interferogram, necessary to perform the emissivity correction. The shorter, narrower, waveband arm is supplementary to the temperature measurement function and provides a "familiar" interferogram for people used to red diode laser based instruments. During its operation data is collected on a cycle corresponding to 4 revolutions of the susceptor. During the first revolution the shutter of the optical control unit is open and light from the tungsten lamp is focussed onto the wafer. A reflectance signal corresponding to the average reflected intensity, gathered whilst the beam traverses a limited sector of the wafer, is taken. During the second revolution the lamp shutter is in the process of closing and the software ensures that no data is taken during this time. The shutter is then fully closed during the third revolution allowing the incandescence signal to be measured in the absence of the reflectance signal. The time for the fourth revolution is used to open the shutter again in order that the cycle can be repeated. Using appropriate algorithms in the control computer the transient "Fabry-Perot" modified reflectance signal is interpreted in terms of emissivity and this is used to correct the temperature signal derived from the incandescence.

3.4. Spectral Reflectometry A further derivative of the system described in 3.3 that allows reflectance measurements to be made across a broad waveband can provide further insight into the growth process. Such spectrally resolved data can be taken if the two channel filter based detection system is replaced by a broad-band spectrometer and photodiode array and illumination is again

94

Thrush, Boyd

provided by an intense white light source. This allows the collection of reflectance data as a function of both time and wavelength. Lay Tec markets a commercial variant of such a system under the trade name EpiR DA/UV/TT. This instrument is compatible with use on a CCS reactor and yields data over a wavelength range of 280-750 nm. By suitable manipulation of this data set reflectance transients can be displayed at chosen wavelengths allowing different features of the growing material to be analysed with greater precision. For instance if a shorter wavelength is chosen a greater proportion of the incident light is absorbed by the layer than would be the case at longer wavelengths and the form of the interferogram is dominated by the reflectance from the top surface with little influence from Fabry-Perot oscillations caused by layer thickness evolution. Such a plot is optimal for studying changes in surface morphology but does not provide information on layer thickness or growth rate. On the other hand if it is chosen to display the reflectance transient taken at a longer wavelength then the Fabry-Perot oscillations are enhanced making it a more suitable scheme for the measurement of growth rate at the expense of complicating the perception of surface morphology. The complete spectral data can also be studied in a more holistic manner by the use of three-dimensional graphing routines. These represent the reflected intensity by way of false colour contrast on a two dimensional surface formed by time on one axis and wavelength on the other. Inspection of such three dimensional plots is helpful in deciding how to "cut" the data for more quantitative analysis using two dimensional representations or can be used as a graphic "fingerprint" for monitoring the reproducibility of the growth of complex structures.

4. Reagent Concentration Monitoring Compositional control of indium containing III-V compounds have posed problems for some considerable time20'21. This is because simply defining the carrier gas flow rate through a metal organic source together with its temperature and pressure does not define the molar delivery rate of the reagent with sufficient accuracy for state-of-the-art MOCVD. This is

Close-Coupled Showerhead MOCVD - GaN

95

particularly true for solid sources such as trimethyl indium and cyclopentadienyl magnesium where "gas channelling" routes through the reagent bulk and the development of low vapour pressure "slow etching" crystal facets affects the pick-up efficiency leading to considerable variability during the life of the source. Even for bubbled liquid sources such as trimethyl gallium, a significant fall in output is noticed (for the same operating conditions), well before the source is fully depleted. In response to this problem a scheme for accurately monitoring the concentration of trimethyl indium in hydrogen was developed in the late 1980s22'23 and has subsequently been commercialised as the Epison monitor. This device operates by measuring the velocity of sound in the gas mixture leaving the metal organic source (carrier gas plus reagent) and using this data to compute the composition of the mixture, on the basis that the velocity of sound is inversely proportional to square root of the mean molecular weight of the gas mixture, as expressed equation (1).

v- M

(i)

V M where "V" is the velocity of sound, "y " is the weighted average of the ratio of the principal specific heats of the gases in the mixture, "R" is the universal gas constant, "T" the absolute temperature and "M" the mean molecular weight of the gases in the mixture. As the values for these constants which are averaged in the equation are fixed, known, or estimatable, (but again fixed) parameters it means that a measurement of the sound velocity can be used to extract the composition of the mixture by suitable algebraic manipulation. A schematic diagram of the ultrasonic cell that constitutes the heart of the Epison system is shown in Fig. 12. The gas mixture flows through the hermetically sealed and heated cell that is fitted in the pipe run from bubbler to gas switching manifold. This is accurately temperature controlled to prevent condensation of reagent and to stabilise the measurement, which is extremely temperature sensitive. An ultrasonic pulse is applied to the transmitter transducer and the transit time delay before this pulse generates a corresponding electrical signal in the receiver transducer is measured by an appropriate

96

Thrush, Boyd

electronic timing circuit, thus enabling the velocity of sound in the mixture to be determined. The efficacy of this system was proven at an early stage when feedback control from the Epison reagent concentration monitor allowed the growth of high quality InGaAs under circumstances where gross changes in bubbler pick-up efficiency were deliberately induced. Since that time the instrument has evolved through 4 iterations to yield a component with a footprint compatible with that of a mass flow controller. Also refinement of the electronics of the latest version (Epison 4) has considerably improved the signal to noise ratio compared to previous marques and now offers sufficient precision to enable the effective monitoring of reagents, delivered in dopant quantities, to be undertaken. In a recent test the measured reagent concentrations from four runs in which hydrogen carrier gas was passed at a flow rate of 450 seem through a bubbler of "solution" cyclopentadienyl magnesium were compared. In the first pair of runs the bubbler was held at 40°C whilst in the second pair it was held at 32°C. The runs were carried out within a short time frame such that any real variations in bubbler output would be expected to be very low. Excellent reproducibility was obtained in both cases suggesting that Epison 4 is capable of making competent measurements of this reagent down to (at least) 0.005%. -heater ultrasonic receiver transducer

temp, sensor

ultrasonic sender transducer measurement cell Fig. 12. Schematic of Epison reagent concentration monitor.

Close-Coupled Showerhead MOCVD - GaN

97

5. Growth Results

5.1. GaN Template Growth in Close-Coupled Showerhead Reactors Using the proven "two temperature" technique, which deposits GaN or A1N low temperature nucleation layers, prior to the growth of thick GaN epilayers at a higher temperature, specular GaN templates have been prepared on sapphire and silicon carbide substrates. Also, using a high temperature A1N nucleation layers, GaN templates have been prepared on silicon. Typically these exhibit a standard deviation in thickness below 2% across a 2" wafer, although best case values below 1% have been noted. Such uniformities have been achieved on 3x2, 6x2" and 19x2" reactors. The final surface morphologies of these layers (as indicated by Atomic Force Microscopy) exhibit a "step-flow" pattern indicative of monolayer two-dimensional growth. These surfaces have RMS roughness values down to ~ 0.13 nm over an area of 4 u,m2, as indicated by Figs. 13 (a) and 13 (b) which show the topologies of GaN template material grown on sapphire and silicon respectively, in a 6x2" reactor. As discussed briefly in Section 3.2 two different regimes have been investigated for the growth of the high temperature epilayers, one in which the isolated GaN crystallites, which result from the annealing of the nucleation layer coalesce quickly, and the other where a protracted period of 3 dimensional growth follows the growth and annealing of the nucleation layer. The "grow out" and the control of the type of threading dislocations is influenced by the rate at which the GaN coalesces, by the surface specular reflectivity and by the shape of the coalescing islands. This has a profound effect upon the conductivity of undoped material. Transmission Electron Microscopy (TEM) analysis of these materials has shown that the semi-insulating layers grown using the "rapid recovery" regime contain more edge type dislocations than do the conducting layers grown using a "slow recovery" regime25. Such studies have also indicated that the coalescence of GaN grains with facets angled with respect to the (001) direction are more effective at bending over threading dislocations and preventing them reaching the top surface than are grains with vertical

Thrush, Boyd

98

side walls26. GaN grown using conditions that promote rapid coalescence is found to be semi-insulating (up to 109 £l/sq.), whereas GaN grown in such a way that coalescence occurs gradually is invariably conducting. Hcighl (nm)

2.0

Mi'inht

Mi

I

2.0

dim i

i

1.0

•(1.0

1.0

(hi

(a) 1.0

u

(a.m

1.0

20° iU

p:m

Fig. 13. Atomic force microscopy of GaN on (a) sapphire (b) silicon.

Real-time monitoring of the evolution of the surface morphology, using the in-situ laser based interferometer described in Section 3.2, has provided insight into, and afforded improved control of, the mechanisms responsible for these different growth regimes. Fig. 10 shows the different interferograms that result when growth conditions are adjusted to produce templates that are conducting (a) or semi-insulating (b). There are a number of strategies for managing the rate at which coalescence is approached. Bougrioua et al.18 adjusted the growth rate, growth temperature and nucleation layer annealing time to achieve this end, whilst work at Cambridge University has used the ammonia partial pressure during the annealing of the nucleation layer as the main controlling parameter. Higher partial pressures are found to promote a faster recovery of the reflectivity. It is important in terms of system flexibility that both types of template can be grown. Conductive templates have superior crystalline quality and are chosen for the buffer layers of device structures requiring ambipolar current injection such as LEDs and lasers, whilst a high resistivity buffer layer is mandatory to obtain good pinch-off characteristics for Field Effect Transistors (FETS).

Close-Coupled Showerhead MOCVD - GaN

99

In general growth rates for GaN of between 2-3 u.m/ hour are used for most applications, however rates up to 12 u.m/hour have been realised in a 6x2" CCS reactor using high TMG molar flow rates. Whilst the surface morphology of such rapidly grown layers were found to be good they have not been extensively characterised in other respects. 5.2. Quantum Wells in Close-Coupled Showerhead Reactors

Fig. 14. Bright field and lattice image TEM cross sections of InGaN/GaN MQWs.

Currently the most commercially significant sector for MOCVD grown GaN based structures is that of LEDs. Consequently we have devoted effort to the study of quantum wells that form the active region of such devices. We have noted that both InGaN/GaN and GaN/AlGaN quantum wells grown in a 6x2" system are characterised by abrupt interfaces as perceived by TEM, (see Fig. 14) and that their X-Ray Diffraction rocking curves exhibit both Pendelosung fringes and large numbers of satellite peaks for multi-quantum well (MQW) structures27 again indicating abrupt interfaces and precise replication of the well and barrier structure. Fig. 15 shows results for InGaN/GaN structures. Similar results can be obtained for GaN/AlGaN quantum well structures. Reciprocal space maps for both types of quantum well structures indicate that they are fully pseudomorphic if grown correctly. An example of such a map taken from an InGaN/GaN MQW structure is given in Fig. 16. The technique used to

Thrush, Boyd

100

construct this type of map employs the asymmetric (105) reflection and shows the spread both normal to (y- axis), and in the plane (x-axis) of the layers, with the zero and +1 order satellite peaks straddling that of the bulk GaN peak. The precise alignment of the peaks on the x-axis indicates that the structures are fully strained. Similar results are obtained for GaN/AlGaN quantum well structures grown within the critical thickness constraints. To characterise the optical emission quality of the InGaN/GaN MQWS variable temperature Photoluminescence (PL) measurements were made as a means of determining the internal quantum efficiency (IQE). For the best samples high room temperature efficiencies io6 •

10 3 -

1 8m satellite fringe r . . A o j A w ' W ' **^ '^#f^A#VrrW" w 14

15

Obtained using beam selection diffractometer. (Courtesy P. Fewster, Pananylitical Research Centr 16

17

co/26 (degrees)

Fig. 15. X-Ray rocking curves of InGaN/GaN quantum wells.

up to 43% were obtained. In order to study the optical quality as a function of well composition (and therefore emission wavelength) a series of InGaN/GaN single quantum wells were grown covering the wavelength range 370-570 nm (at 6K). The indium content of the wells, which is a prime parameter in determining the emission wavelength, was controlled by varying the well and barrier growth temperatures and determined by X-ray diffraction measurements. At 6K the spectra of these samples were dominated by strong localised exciton emission, with attendant phonon replicas, across the whole of the wavelength range as shown in Fig. 17.

Close-Coupled Showerhead MOCVD - GaN

101

Qy*1000(rlu)

Qx*1000(rlu)

Fig. 16. Reciprocal space map of InGaN/GaN quantum wells. Energy (eV) 3.6

3.2

2.8

-

2.4

, Tg:800 "C, ln:5%

Tg:770°C, ln:12% /

.

Tg:750°C, ln:15%

,/

/ /

350

400

450

Tg:730°C, ln:19% /

500

Tg:710°C, ln:25%

550

600

650

700

Wavelength (nm)

Fig. 17. 6K PL spectra from series of single QW samples.

The growth temperatures and indium contents of the quantum wells are also shown on this figure. The line widths of the zero phonon lines are very narrow with the FWHM of the peak at 374 nm being 3.1 nm (28 meV), indicating quantum wells of very high quality. Further details are given in an earlier publication28. Measurements of the internal quantum efficiency have also been made on MQW samples by comparing the PL intensity at 6K with that obtained at room temperature. Making the assumption that the low temperature recombination process is wholly radiative, (i.e. that the internal quantum efficiency is 100%),

102

Thrush, Boyd

allows the value at room temperature to be estimated by simply rationing the intensities. It is found that the fall in intensity with temperature is very dependent upon the structure, particularly the indium content of the wells. Wells containing -16% indium have been found to have the best performance at room temperature giving the 43% IQE mentioned earlier. Although many of the customers of Thomas Swan are concerned with the manufacture of LEDs there has been a relative paucity of such data available which the company has been able to directly relate to its own materials measurements. For this reason we have undertaken some basic measurements on 5 mm LED lamps, professionally fabricated from material grown on a 6x2" reactor at Cambridge University. Some of the best devices were selected for measurement of their output characteristics and their reliability. For LEDs operating at -503 nm output powers up to 4.7 mW were obtained. A small number of these devices have also been subjected to real-time life testing. For operation at a constant 30 mA an initial degradation of -23% was noted over the first 4 months of operation. From that point to the current time, (~ 3.3 years into the experiment), an additional fall in output of - 5 % has been noted. 5.3. HEMT Structures in Close-Coupled Showerhead Reactors Using high resistivity GaN buffer layers, as described in Section 5.1, AlGaN/GaN High Electron Mobility Transistor (HEMT) structures have been grown on sapphire in a 6x2" CCS reactor. A generic structure is shown in Fig. 18. The conducting channel in such devices is a twodimensional electron gas (2DEG) at the AlGaN/GaN interface, arising from a combination of spontaneous and piezoelectric polarisation at this interface. The transport properties of this 2DEG are a sensitive measure of interfacial perfection with the mobility (for a given sheet carrier concentration) being degraded if the interface is not abrupt and the AlGaN is crystallographically relaxed, rather than fully pseudomorphic. X-Ray diffraction studies of HEMT structures with AlGaN thicknesses up to - 40 nm thick and aluminium contents of up to - 30% have shown very good agreement between experimentally derived rocking curves and

Close-Coupled Showerhead MOCVD - GaN

103

simulations based upon perfect interfaces and zero relaxation in the AlGaN. Qualitatively this implies that the structures are of high quality in these two respects. Additionally XRD reciprocal space maps of such structures have indicated that the AlGaN is fully strained. Using such structures state-of-the-art mobilities of greater than 1400 cmV's" 1 , for a sheet carrier concentration ~1013cm"2 have been obtained on sapphire and a maximum mobility of 1947 cir^V'V1 has been achieved at a sheet carrier concentration of 9.4xl012cm"2 when supplementing the structure shown in Fig. 18 with a thin A1N layer between the AlGaN cap and the GaN buffer. The purpose of this layer is to limit the penetration of the wave functions of the 2DEG electrons into lower mobility AlGaN layer and effectively confine them to the higher mobility GaN channel layer.

15-40 nm, 15-40%, AlGaN GaN nucleation „ , „ . „ .. layer 2 - 4 u m S.I. GaN

24 nm, 26% AlGaN AIN nucleation layer

2 urn S.I. GaN

'/,W///////sj;/s;;//;///;s/77777Z Sapphire Fig. 18. Generic HEMT structure.

Silicon Carbide Fig. 19. HEMT structure on SiC.

HEMT structures have also been prepared on SiC which is the preferred substrate for power devices owing to its higher thermal conductivity than sapphire. In this case the low temperature GaN nucleation layer used for sapphire growth is replaced by an AIN nucleation layer prior to the growth of the GaN buffer. Using such a structure with the characteristics indicated in Fig. 19 sheet resistance and carrier concentrations of 455 Ohms/square and 9.6x10' cm" respectively have been combined with a mobility of 1460 cmV's" 1 . In addition further experiments on SiC in which the thickness and aluminium content of the AlGaN layer was raised to 30 nm and 35% respectively, and combined with an AIN interlayer, it was possible to reduce the sheet resistance to the very low value of ~ 300 Ohms/square. Dense mapping

Thrush, Boyd

104

measurements performed on these latter structures showed that the component layers were of good uniformity in terms of both thickness and composition and that these characteristics were reflected in terms of the uniformity of the sheet resistance values. Fig.20 shows thickness maps of GaN layers grown on two 2" SiC wafers, one 4H and the other 6H, in the same run. They are constructed using white light interferometry. This technique effectively samples the whole layer structure and thus primarily indicates the uniformity of the GaN buffer layer (which dominates the thickness). As can be seen the standard deviation is ~2%.

x

\

» 5, ""..«>««!*!»•*' ' ^ "'

r.;xcr100.0X

Betaw

0.0*

Run #2on4H-SiC Thickness (pm) Std.Dev.

/'

' A»go 1 97.'Mfidian • 1 SF. SWDev. 2.21J* |[|0«SAhnvs

0(1X

iSpoclDD.0*

Betow

00X

Abnve

ODS;

: sickness (pm) | Std.Dev

2.0 2.2%

2.0 2.0%

Fig. 20. Thickness maps of GaN buffer layers on 4H and 6H SiC.

•

I •IV I.I . ••

I)

I- .-"- • . 1 . ;

>e*

'. 1\

.'t

Run ft 2 on 4H-SiC AI content "x" for AlxGa(1-x)N Std.Dev. of "x"

0.35 2.2%

l-r»r°r "•" 1 . .cr* i.u .-, r-t" ^ 5 '. Run # 2 on 6H-SiC AI content "x" for AlxGa(1-x)N 0.35 Std.Dev. of "x" 2.3%

Fig. 21. Aluminium content map of HEMT layers grown on 4H and 6H SiC.

, r \ \ "••

Close-Coupled Showerhead MOCVD - GaN

-20

-10 0 10 X position (mm)

Run#2on4H-SiC AIGaN thickness (nm) Std.Dev,

20

30.0 2.9%

20

105

-10 0 10 X position (mm)

Run#2on6H-SiC AIGaN thickness (nm) StdDev.

20

30.7 2.7%

Fig. 22. Thickness maps of AIGaN layers of HEMT structures grown on 4H and 6H SiC. Fig.21 shows aluminium content maps constructed from photoluminescence wavelength mapping on the same two wafers. There is a small rise in the aluminium content towards the outside of the susceptor (towards the flat) although this is less easy to appreciate in this monochrome illustration than in the coloured original. The average composition for the layers grown on both types of wafer is 35% and again the standard deviation across both wafers is ~ 2.2%. Fig. 22 shows contour plots depicting the thickness of the AIGaN layers of the two HEMT wafers constructed from a series of XRD measurements made across each of them. These data suggest that the thickness of this layer varies on average between 30.0 and 30.7 nm for the layers grown on the 4H and 6H wafers respectively and that the standard deviation is between 2.7 and 2.9% for the two wafers. X-ray measurements were also used to provide a measure of the crystalline quality of the GaN buffer layers, as reflected by the FWHM of the (002) rocking curves, and the variability of that cystallinity

106

Thrush, Boyd

across the surface of the wafers. This is seen to be an important parameter as the 2DEG is located in the upper region of the GaN buffer layer and the local transport properties of that layer will be influenced by its extended defect density and reflected in the characteristics of transistors fabricated upon it. The average FWHM was found to be 230 arc seconds and the standard deviation was 4%. Finally the uniformity of the electrical properties was assessed using a Lehighton contactless resistivity probe to map the sheet resistance across the wafers. The result of this exercise is shown in Fig. 23 indicating that the mean sheet resistance values on the 4H and 6H wafers was 291 and 306 Ohms/square and that the standard deviations were 2.2 and 2.1% respectively.

Run # 2 ort 4H-SiC Average Sheet resistance 291 Ohms/square Std.Dw. 2.2%

Run#2on6H-SiC Average Sheet resistance 306 Ohms/square Std.Dev. 2.1%

Fig. 23. Sheet resistance maps of DEGs of HEMT structures on 4H and 6H SiC.

Close-Coupled Showerhead MOCVD - GaN

6.

107

Mass Production of GaN Based Devices

6.1. Cost of Ownership Dependence on System Size The TMG on-wafer growth efficiency when growing a GaN epilayer is dependent on the process conditions used, including total flow rate, reactor height, wafer temperature, reactor pressure, NH3 flow rate, and rotation speed. The process conditions can be set, on all system capacities to achieve a high TMG growth efficiency above 20%. As was also discussed earlier the process conditions in the reaction zone and the growth rate are similar across the susceptor radius resulting in inherent uniformity. It follows therefore that these process conditions and the growth rate are independent of the susceptor size assuming the delivery gas flow per unit area and gas mixture admitted to the reactor are the same. To scale a process between systems of different capacities we aim to make the process conditions in the reaction zone similar and apply scaling rules that keep the flow rate per injector approximately the same. It follows therefore that the efficiency and mass of TMG per wafer consumed during a run should be relatively similar. For instance the weight of TMG per wafer required to deposit a 4.2 urn GaN layer in 3x2", 7x2" and 19x2" CCS systems is 0.6, 0.47 and 0.40 grams respectively. There is a modest reduction of TMG consumption per wafer moving from the 3x2" system to the 19x2" which is primarily due to an increase in wafer area relative to showerhead area. For commercial wafer production the requirement of MOCVD is to grow structures of the specified performance with a low enough cost of ownership (COO) to maintain profitability per device, or wafer, at a suitable capacity. Because the COO will vary depending on the wafer supplier's circumstances, and will differ between a chip supplier and a wafer supplier a complete model is beyond the scope of this publication. For this discussion we will show unit-less values from a model based on the production of LEDs on 2" sapphire wafers that are accepted or rejected on the basis of their performance yield passing a qualifying specification. Assuming this the profitability of the system is the epiwafer sale price minus the total COO per epiwafer, multiplied by the volume of wafers sold. The relative materials cost per wafer and total

Thrush, Boyd

108

COO per wafer are shown in table 1. These were derived assuming that the LED process cycle involved baking in-situ between runs, a typical 4.2 (im thick blue LED demonstration run, and estimated costs for the metal-organics, NH3, SiH4, N2, H2, electricity and wafers. The total COO per wafer includes facility costs, manpower, spare parts and system book price depreciation. The materials cost per wafer reduces by around one third on scaling from the 3x2" to the 19x2" capacity system. This difference is less than the reduction in TMG consumption since all gas consumptions per wafer are reduced on scaling, but the substrate cost remains constant. The lowest COO per wafer is not automatically achieved by having the largest capacity and highest throughput system. It is conceivable that a large number of smaller capacity systems even with low throughput, could be more profitable than a small number of large capacity systems, if their running and fixed costs were low and the yields were high. Table 1. Relative materials cost per wafer and total COO per wafer for the three CCS multi-wafer system sizes.

System capacity (x2" wafers) 3 7 19

Relative materials cost per accepted wafer 1.34 1.14 1.00

Relative total COO per accepted wafer 6.78 3.62 2.03

However, our data on using CCS reactors for blue LEDs supports a dramatic 3.5x reduction in COO per wafer on scaling production from the 3x2" to a 19x2" capacity system, as shown in Table 1. This is mainly due to the facility costs, manpower, spare parts and system book price depreciation showing only a small increase per system as the system capacity is increased and thus a large reduction in cost per wafer due to higher throughput. In formulating Table 1 identical performance and yield is assumed for the different capacity systems, and in the real case of slightly improved yield with the larger systems the improvement in COO per wafer on increasing system capacity will be even greater. The dramatic fall in total COO per wafer on scaling to larger capacity systems

Close-Coupled Showerhead MOCVD - GaN

109

has facilitated a drop in epiwafer and chip price since their introduction. The epiwafer price has fallen to a point that for typical blue LEDs it is likely that the 19x2" system is currently the only system that is profitable. Considerations of risk, flexibility and development cost cannot account for the difference in COO per wafer. It is because of COO per wafer that epiwafer suppliers were typically quicker than chip suppliers to upgrade their production to the larger capacity systems since they were more exposed to the epiwafer cost than chip suppliers. It is also worth noting that the materials costs are more important for the financial viability of the 19x2" system than is the case for the smaller systems, and that the materials costs are now of a similar magnitude to the fixed costs. This implies that future increases in system capacity will not yield a similar benefit in terms of profitability unless the materials costs per wafer can be further reduced. It should be pointed out that the foregoing analysis was predicated on the manufacture of low cost devices using a mature technology. However a model for the growth of high cost devices such as blue lasers on free standing GaN substrates may yield a smaller relative reduction in the COO per wafer for increases in system capacity.

6.2. Process Cycles The methodology of running the process cycle, in particular the conditioning of the reactor between process runs, can have a significant impact on the cycle time and the yield. It follows that it also has a large effect on the COO in production. A method is required to remove the coating, which is deposited during the process onto the susceptor and showerhead, which would lead to process instability and particulate contamination of the wafers, both impacting on the yield. There are currently a number of methods in use for the mass production of GaN devices, listed in Table 2 with throughputs for blue LED production based on a demonstration process neglecting downtime or maintenance, and a TSSE recommendation on compatibility with AlGaN containing devices.

110

Thrush, Boyd

From Table 2 continuous running looks an attractive proposition for the production of blue LEDs due to the improved throughput, but in reality the stability of the process is not as good as when baking or bake coating is used and if the susceptor is baked in the reactor after a batch of runs this bake-out needs to be long and reduces the throughput advantage. Thus most production is based on methods 2 and 3, baking the susceptor or baking followed by coating the susceptor in the reactor. Both give good reproducibility and performance (results using such baking are shown in section 6.3). The advantage of coating with around 0.25 jam of GaN after the baking is that it stabilises the bulk GaN quality, but good quality can be achieved without coating. Baking with HC1 yields no benefit for structures not including AlGaN, but since cleaning the showerhead by brushing with a nylon brush attached to a vacuum cleaner within the glovebox takes only around 15 minutes cleaning the showerhead every run is feasible. It yields slightly better run-to-run stability but for most products this gain isn't sufficient to outweigh the throughput loss. Recently a number of producers have switched to exchanging the susceptor every run, and cleaning externally. This cleaning can be by baking, baking with HC1, or etching. The variation due to reloading different susceptors isn't seen as a significant issue. For structures incorporating AlGaN in-situ susceptor baking and baking followed by coating is less successful since the baking temperatures (1100°C) used for GaN are insufficient to remove AlGaN deposits at a significant rate. Therefore some groups use continuous running with ex-situ cleaning of the susceptor after a number of runs, or exchanging the susceptor every run, which is the better option for both throughput and stability. It is possible to remove the susceptor, brash it with nylon brush and then bake it in-situ if adequate ex-situ cleaning facilities are unavailable. Finally one could use an in-situ bake with some HC1 gas in the carrier, which would remove the AlGaN deposit. But TSSE has little experience of this due to concerns of exhaust component lifetime if moisture is allowed to enter the system during maintenance.

Close-Coupled Showerhead MOCVD - GaN

111

Table 2. List of process cycles with their typical throughput (based on 24 hour operation excluding maintenance) and compatibility with AlGaN based processes.

No.

1 2

3

4

5

6

Process Cycle Description Running batches of runs with periodic baking of the susceptor and cleaning of the showerhead. Baking the susceptor between every process run. Baking followed by coating of the susceptor loaded with sacrificial wafers between every process run. Baking the susceptor with HCI in the system between every run. Baking or baking followed by coating of the susceptor combined with showerhead brushing between every process run. External Baking / cleaning of the susceptor between every run and exchanging between a batch of susceptors.

Throughput (runs/day)

Recommended for use with AlGaN.

5.4

Yes

3.8

No

3.7

No

3.6

Yes

3.3

No

5.3

Yes

6.3. Growth Data from 19x2" Production System The CCS GaN 19x2" system has proven itself to produce state-of-the-art devices with excellent uniformity and reproducibility in production29. The performance of this system is exemplified by results from a growth campaign undertaken after commissioning at a customer's site. This work, aimed at establishing a capability for blue LED structures, was based upon a 5 period InGaN/GaN MQW structure grown on an n-type silicon doped template, capped by a p-type magnesium doped layer. The structure is shown schematically in Fig.24. The InGaN/GaN quantum wells and barriers were grown at different temperatures (InGaN wells

112

Thrush, Boyd

~720°C, barriers ~870°C) with appropriate pauses between the two to allow for the adjustment of the temperature. |

0.2|xm GaN:Mg ([p]~3x1 0 1 7 cnr 3 )

2\im GaN:Si ([n]~3x1 0 18 cm- 3 )

1nm Undoped GaN 25nm Nucleation Layer

430nm Sapphire Substrate

Fig. 24. LED structure used for the presented production data.

Fig. 25. Susceptor/wafer key.

To determine thickness and optical emission uniformity, PL mapping of the full complement of 19 two-inch wafers was undertaken. This data was collected by an Accent RPM2000 PL mapping system fitted with a white light source to analyse the wavelength dependence of reflection due to optical interference. The thickness data is summarised for the individual wafers in columns 2 and 3 of Table 3. The positions of the individual wafers on the susceptor are identified in Fig.25. The average intra-wafer thickness uniformity was 1.55%, whilst the interwafer standard deviation of the mean wafer thicknesses was 0.89%.

Close-Coupled Showerhead MOCVD - GaN

113

Table 3. Uniformity data from the LED process run. 1

2

1

3

4

5

6

7

8

a a Avge. Avge. PL intensity Avge sheet thickness Wafer Avge. A aA.(nm) thickness intensity res. (%) position (nm) 2 mm excl. (%) 3mm (pm) (Ohms/sq.) (a.u.) 2mm excl. excl. 0.58 125.3 464.3 3.32 0 3.63 11.12 28.1 1 1.27 125.0 6.33 457.6 1.49 27.6 3.55 2 3.54 1.29 131.8 2.72 456.1 1.35 27.7 3 1.15 125.9 2.38 458.7 1.35 27.7 3.55 4 125.0 2.88 459.7 1.75 27.6 3.57 1.30 5 1.35 119.7 27.6 3.58 4.60 460.0 1.90 1.21 127.3 457.1 27.6 6 3.90 1.26 3.58 7 1.99 124.3 3.20 458.2 1.98 27.6 3.60 3.54 1.72 125.6 3.18 457.7 27.6 8 2.30 27.7 9 123.9 4.08 458.6 2.03 3.52 1.80 122.4 458.2 1.79 27.7 10 1.71 2.76 3.53 11 1.91 125.1 3.88 457.4 1.97 27.6 3.53 12 27.6 1.70 120.0 2.89 459.9 2.01 3.55 13 3.61 1.86 119.4 2.86 1.77 27.6 460.1 14 124.9 459.2 2.19 27.5 1.83 5.01 3.60 1.71 118.4 3.93 461.3 1.97 27.5 15 3.60 16 1.69 122.4 459.3 1.95 27.4 5.10 3.60 17 1.77 116.0 5.29 460.2 2.23 27.6 3.58 18 1.65 121.9 2.32 458.0 1.67 27.5 3.58 123.4 Avge. 3.57 1.55 4.13 459.0 1.91 27.6 2.94 (%) a wafer to wafer 1.81 (nm)

9 a sheet res. (%) 1.22 0.62 0.60 0.54 0.48 0.46 0.56 0.60 0.71 0.63 0.56 0.68 0.57 0.58 0.58 0.52 0.50 0.48 0.65 0.61 0.5 (%)

Mapping of the dominant PL emission wavelength of all wafers gave a mean value of 459 nm and indicated an average standard deviation of 1.9 nm on a wafer (with a 2mm edge exclusion) as shown in column 7 of Table 3. The same exercise indicated that the standard deviation of the mean wavelengths of the wafers constituting the fully populated run was 1.81 nm. An indication of doping uniformity was obtained from sheet resistance measurements made on the structures using a Lehighton contact less profiler. This data predominantly samples the conductivity of the Si doped buffer layer. The average values of sheet resistance and the standard deviation of sheet resistance across the wafers are presented in columns 8 and 9 respectively of Table 3. On average the sheet resistance exhibits a standard deviation of around 0.6% within the wafers, whilst the

Thrush, Boyd

114

standard deviation of the mean values of all the wafers is around 0.5%.The run-to-run reproducibility in terms of the optical properties was evaluated by collecting peak PL wavelength data over a series of 22 identical LED production runs. An in-situ susceptor baking procedure was used between every run. The data is shown in Fig. 26. These 22 runs gave a mean global average wavelength of 466.6 nm and a standard deviation of those global averages of 1.18 nm.

480

62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83

Run Number Fig. 26. Run-to-run peak wavelength reproducibility of a blue LED process.

7. Acknowledgements The authors are indebted to members of the Metallurgy and Materials Department of the University of Cambridge with whom we have collaborated to provide much of the data contained in this chapter. We would particularly wish to recognise the contributions of Dr. Menno Kappers, Ms Mary Vickers, Mr. Clifford McAleese, Mr. Matt Charles, Mr. David Rayment and Professor Colin Humphreys in this respect. We also wish to acknowledge the input of those customers of Thomas Swan Scientific Equipment Ltd. with whom we have valuable interchange and in a particular to Dr. Hyeongsoo Park, of EPIPLUS Co. Ltd., Korea for the provision of the 19x2" growth data. We are also grateful to Elsevier

Close-Coupled Showerhead MOCVD - GaN

115

for permission to reproduce Figs. 1, 3 and 13, originally used in Optoelectronic Devices: III-VNitride, M. Henini and M. Razeghi (Eds). References 1. H.P. Maruska and J.J. Tietjen, "The preparation and properties of vapor-deposited single crystalline GaN", Appl. Phys. Letts. Vol. 15, 327 (1969). 2. M. Manasevit, F.M. Erdman and W.I. Simpson,. "The use of Metal organics in the Preparation of Semiconductor Materials IV", The Nitrides of Aluminium and Gallium. J. Electrochem. Soc. Vol. 118, 1864 (1971). 3. S. Nakamura, M. Senoh and T. Mukai. "High-power InGaN/GaN doubleheterostructure violet light emitting diodes", Applied Physics Letters, Vol. 62, 23902392(1993). 4. EJ.Thrush, C.G. Cureton and A.T.R. Briggs. "MOCVD grown InP/InGaAs structures for optical receivers". Jnl. Cryst. Growth, Vol. 93, 870-6 (1988). 5. K.F. Jensen. "Handbook of Crystal growth 3, Thin Films and Epitaxy, Part B: Growth Mechanisms and Dynamics", North Holland, (Editor D.TJ .Hurle), 574 (1994). 6. H. Schlichting, "Boundary Layer Theory", Mc-Graw Hill, New York, 1960. 7. T.G. Mihopoulos, S.G. Hummel and K.F. Jensen. "Simulation of flow and growth phenomena in a close-spaced reactor", Jnl. Cryst. Growth, Vol. 195, 725-732 (1988). 8. R.H. Moss and J.S. Evans. "A new approach to MOCVD of Indium Phosphide and Gallium-Indium Arsenide", Jnl. Cryst. Growth, Vol 55, 129 (1981). 9. A. Thon, T. F. Kuech, Appl. Phys. Lett. 69, 55 (1996). 10. T. Mihopoulos, "Reaction and transport processes in OMCVD: Selective and group-Ill Nitride growth", PhD thesis, MIT, (1999). 11. J.R. Creighton, G.T. Wang, W.G. Brieland, M.E. Coltrin. "Nature of parasitic chemistry during AlGalnN OMVPE", Jnl. Cryst. Growth, Vol 261, 204 (2004). 12. M. Dauelsberg. Private Communication, (2000). 13. R.P. Pawlowski, C. Theodoropoulos, A.G. Salinger, T.J. Mountziaris, H.K. Moffat, J.N. Shadaid, E.J. Thrush. "Fundamental models of the metalorganic vapor-phase epitaxy of gallium nitride and their use in reactor design", Jnl. Cryst. Growth, Vol. 221, 622-628 (2000). 14. S. Nakamura. "In Situ Monitoring of GaN Growth Using Interference Effects", Japan J. Appl. Phys. 30, 1620 (1991). 15. L. Considine, E.J. Thrush, J.A. Crawley, K. Jacobs, W. Van der Strict, I. Moerman, P. Demeester, G.H. Park, S.J. Hwang, J.J. Song. "Growth and in-situ monitoring of GaN using IR interference effects", Jnl. Cryst. Growth, Vol 195, 192-198 (1998). 16. S. Figge, T. Bottcher, S. Einfeldt, D. Hommel. "In Situ and Ex Situ Evaluation of the Film Coalescence for GaN Growth on GaN Nucleation Layers", Jnl. Cryst. Growth, Vol 221, 262 (2000).

116

Thrush, Boyd

17. R. Datta, M.J. Kappers, M. Vickers, J. Barnard, C.J. Humphreys. "Growth and characterisation of GaN with reduced dislocation density", Superlattices and Microstructures, Vol. 36, 393-401 (2004). 18. Z. Bougrioua, I. Moerman, N. Sharma, R.H. Wallis, J. Cheyns, K. Jacobs, E.J. Thrush, L. Considine, R. Beanland, J.-L.Farvaque, C. Humphreys. "Material Optimisation for AlGaN/GaN HFET applications", Jnl. Cryst. Growth, Vol 230, 573-578 (2001). 19. K. Haberland J.T. Mullins, T. Schenk, T. Trepk, L. Considine, A. Pakes, A. Taylor, J.T. Zettler. "First Real-Time True Wafer Temperature and Growth Rate Measurements in a Close-Coupled Showerhead MOVPE Reactor During Growth of InGa(AsP) ", Proceedings of the International Conference on InP and Related Materials, 2003, Santa Barbara, IEEE Catalog Number: 03CH37413, ISSN: 10928699, 44-47 (2003). 20. Knauf, D. Schmitz, G.Strauch, H. Juergensen, M. Heyen, A. Melas. "Comparison of Ethyldimethylindium (EDMIn) and Trimethylindium (TMIn) for GalnAs and InP growth by LP-MOVPE", Jnl Cryst. Growth, Vol. 93, 34 (1988). 21. P.K. York, K.J. Beerlink, J. Kim, J.J. Alwan, J.J. Coleman, CM. Wayman. "Metalorganic chemical vapor deposition of InGaAs-GaAs-AlGaAs strained-layer quantum well lasers", Jnl Cryst. Growth, Vol. 107, 741, (1991). 22. E.J. Thrush, C.G. Cureton, J.M. Trigg, J.P. Stagg, B.R. Butler. "Reactor design and operating procedures for InP based MOCVD". Chemtronics, Vol. 2, 62-68 (1987). 23. J.P. Stagg,. "Reagent concentration measurements in metal organic vapour phase epitaxy (MOVPE) using an ultrasonic cell", Chemtronics, Vol. 3, 44 (1988). 24. J.P. Stagg, J. Christer, E.J. Thrush, J.A. Crawley. "Measurement and control of reagent concentrations in MOCVD reactors using ultrasonics", Jnl. Cryst. Growth Vol. 120, 98-102 (1992). 25. S. Hubbard, G. Zhao, D. Pavlidis, E. Cho, W. Sutton. "Optimisation of GaN Channel Conductivity in AlGaN/GaN HFET structures Grown by MOCVD", Paper El 1.11 Presented at MRS Fall conference, Boston MA, (2004). 26. P. Gibart. Metal Organic Vapour Phase Epitaxy of GaN and Lateral Overgrowth. Reports on Progress in Physics 67 (5): 667-715 (2004). 27. E.J. Thrush, M.J. Kappers, P. Dawson, M.E. Vickers, J. Barnard, D. Graham, G. Makaronidis, F.D.G. Rayment, L, Considine, C.J. Humphreys. "GaN/InGaN quantum wells grown in a Close-Coupled showerhead reactor. Jnl. Cryst. Growth", Vol. 248, 518-522(2003). 28. E.J.Thrush, M.J. Kappers, P. Dawson,, D. Graham J.S. Barnard, M.E. Vickers, L.Considine, J.T. Mullins, C.J. Humphreys. "GaN-InGaN quantum wells and LED structures grown in a close coupled showerhead (CCS) reactor", Phys. Stat. Sol. (a) 192, No.2, 354-359 (2002). 29. A.R. Boyd, L. Considine, O. Feron, J.T. Mullins E.J. Thrush, M.E. Vickers, C. McAleese, P. Dawson, D.M. Graham, H. Park. "Performance of a 19x2" GaNbased MOCVD System", Technical Digest of CS-MAX 2002, 246-248 (2002).

CHAPTER 4 MOLECULAR BEAM EPITAXY FOR III-N MATERIALS

Haipeng Tang1 and James B.Webb2 ;

Institute for Microstructural Science, National Research Council Canada M-50 Montreal Rd. Ottawa, Canada K1A 0R6, E-mail: [email protected] Steacie Institute for Molecular Science, National Research Council Canada, 100 Sussex Drive, Ottawa Canada K1A 0R6 Although current commercial production of GaN based LEDs and LDs is dominated by the MOCVD technique, the emerging market demand for GaN electronics could see GaN-MBE technology becoming an important production tool for future generation RF and microwave electronic components. This chapter looks at the recent advances in molecular beam epitaxy of III-N materials for high-speed electronics applications. Two variations of the nitride MBE technique, ammoniaMBE and plasma-assisted MBE, are covered with reference to key growth challenges such as the growth of semi-insulating GaN. The advantages of the GaN MBE technology for producing high performance GaN/AlGaN HFETs such as the controlled carbon doping of semi-insulating GaN, high material purity and electron mobility, good uniformity and device scalability, and excellent small and large signal RF performance are reviewed.

1.

Introduction

Molecular Beam Epitaxy (MBE) has now reached a high level of sophistication in terms of ultra-high vacuum capability, material purity, in-situ monitoring and control. The technology of MBE has been developed over many decades and today the technique has been successfully applied to the growth of a broad range of materials, particularly group III-V semiconductor compounds. It is still a tool of 117

118

Tang, Webb

choice for many researchers for developing new material systems due to its extensive "in-situ" monitoring capabilities. Many reviews are available describing the technique of MBE1'2 and thus this chapter will be limited to a discussion of the MBE technique as applied to group IllNitride semiconductors for high electron mobility transistor(HEMT) structures. The first part of the chapter deals with MBE material growth while the latter sections describe the recent advances in HEMT structure and performance. In MBE, thermal effusion cells are typically used to evaporate high purity elements to form the compound semiconductors, although variations of MBE are now used to take advantage of liquid and gaseous sources for the group III and group V components, as well as for various dopants. Two variations of MBE specific to the Ill-Nitrides are ammonia-MBE(Am-MBE) and plasma assisted molecular beam epitaxy (PAMBE), although terms such as chemical beam epitaxy (CBE), gas source-MBE and RF-MBE are also used to describe similar techniques. Am-MBE as the name implies, uses gaseous ammonia (NH3) as the group V source and solid sources of Ga, In and Al for the group III components. Solid or gaseous dopant sources of Si and Mg are also used to achieve controlled n and p-type conductivity. PAMBE on the other hand uses N2 as the group V source. This necessitates the use of an RF plasma to crack the N2 to create highly reactive atomic N since the N2 molecule (unlike NH3) has a very high thermal stability. For many years, the GaN MBE research community was divided into two camps according to these two different techniques based on the nitrogen sources. Communication between the two camps has been difficult because the growth phenomena for the two techniques are indeed very different. However, with recent advances in both ammoniaMBE and plasma-assisted MBE (PAMBE), the specific advantages (and shortcomings) of each technique can potentially open new opportunities. 2. MBE Technology Many of the early attempts to use ammonia-MBE in conventional GaAsbased MBEs were less than successful due to the restrictions placed on the range of available growth parameters required to avoid degradation

Molecular Beam Epitaxy for III-N Materials

119

of the MBE hardware components. For instance, ammonia, is a highly corrosive gas when in contact with moisture and can also react with various metals at high temperature. As a result, ammonia-MBE requires pumping systems, substrate heaters, effusion cells and gas control systems specifically engineered to handle this. Metallic surfaces near the substrate heater are typically made from molybdenum while the heater itself is made from pyrolytic graphite which is capable of reaching the high growth temperatures used in ammonia-MBE, and as well, has good chemical stability when exposed to NH3 at high temperatures. The pumping system is typically an oil free combination of cryogenic and turbomolecular pumping followed by dry backing pumps. In addition LN2 cryoshrouding provides a tremendous increase in pumping speed for NH3 allowing flow rates of up lOOccm while still maintaining growth pressures in the range of 10"5 torr. However, a high build-up of ammonia on the cryoshroud requires desorption after several hours of continuous operation. This can be safely accomplished by blanking off the cryopump and using turbomolecular pumping on the growth chamber while the cryoshroud is warmed to RT. This procedure increases the length of time between cryopump regeneration while maintaining the integrity of the system vacuum. However, particular attention is needed to ensure safe operation of the Ga and Al effusion cells, particularly at high ammonia flow rates. In the presence of ~10"5 torr of ammonia, aluminum has a tendency to "creep" from a conventional effusion cell and typical lifetime before failure of these cells is measured in one or two short growth runs. However well designed cold-lip cells have essentially eliminated this problem and extremely long lifetimes are now obtained. A similar, although less severe problem exists for gallium, but this can also be controlled with a suitable hot lip effusion cell. Although PAMBE does not suffer from these problems, the use of N2 as the group V source restricts the maximum available flux of nitrogen at the substrate since N2 is not as readily pumped by the cold shroud as is ammonia and thus low growth pressures are more difficult to maintain at high flow rates. The associated reduced nitrogen flux, requires a lower growth temperature, but is in itself a distinct advantage when indium based compounds are required, since a lower growth temperature is needed to limit the desorption of indium from the substrate during

120

Tang, Webb

growth. The reduced cracking efficiency of NH3 at low temperatures however means that ammonia-MBE is not as well suited for the deposition of indium based compounds. Both techniques however have distinct advantages over MOVPE in that a host of in-situ monitoring techniques are available. Typical MBE systems are equipped with reflection high energy electron diffraction to monitor the surface reconstruction and growth modes of the epilayer in real time while optical pyrometry and reflectometry are used to control temperature and monitor the growth rate and epilayer thickness in real time. However, despite these in-situ techniques, one of the most difficult parameters to control is the substrate temperature particularly at high growth temperatures where thermal radiation losses are high. Uniformity across the wafer is difficult to achieve due to radiation loss at the edge of the wafer and by variations in the refractory metal coatings applied to the back of the substrate. Unlike Si and other III-V substrate materials, the substrates commonly employed in GaN epitaxy (namely: sapphire and SiC) are essentially transparent and thus directly heated wafers require a refractory metal coating such as molybdenum, sputter deposited onto the back surface of the wafer to ensure adequate absorption of the heater radiation. Relatively minor changes in coating properties can result in changes in wafer emissivity, which in turn results in changes in the temperature determined using optical pyrometry. Temperature control during growth is excellent, but run-to-run reproducibility is often compromised since the temperature "window" for optimal material quality is quite narrow. Initially nitrogen plasma sources were very inefficient at producing a sufficient group V flux to stabilise the surface of the epilayer at high growth temperatures and thus PAMBE growth temperatures were limited to lower growth temperatures and relatively low growth rates. AmmoniaMBE however did not suffer from a low group V flux and as a result much higher growth temperatures (closer to the 1100°C used in MOVPE) could be employed while still maintaining relatively high growth rates (>lum/hr). This basic difference between the two techniques has important consequences for the growth of GaN-based materials. In PAMBE growth is carried out at temperatures close to 700°C and in Garich conditions (growth rate controlled by the group V flux - unlike

Molecular Beam Epitaxy for III-N Materials

121

conventional MBE of GaAs for instance) while for ammonia-MBE growth is carried out at temperatures around 900°C under nitrogen-rich conditions(growth rate controlled by the Ga-flux- similar to conventional MBE of GaAs). These two growth regimes have important consequences with regards to surface morphology (ammonia-MBE exhibits rougher surfaces) and dopant incorporation (see next section on carbon doping). 2.1. Ammonia vs Plasma Assisted MBE The PAMBE technique offers growth at lower temperatures (550°C 800°C) and can achieve atomically flat surfaces under Ga rich growth conditions. Surface and interface smoothness are critical for many device structures such as quantum well LEDs, HBTs, and resonant tunneling superlattice structures. The plasma nitrogen source has proved to be very effective for indium incorporation into GaN, a marked advantage for the growth of InGaN based optoelectronic and electronic devices. However, PAMBE performs best when high quality GaN templates are employed. Direct growth of GaN on foreign substrates often shows crystallinity and impurity concentrations that are very sensitive to substrate treatment and a surface polarity that is sensitive to the nucleation layer preparation conditions.4'5 In contrast, the ammoniaMBE technique grows at higher temperatures (850°C - 920°C) and readily obtains Ga-polarity, low-defect-density, and high-electronmobility GaN layers on sapphire and SiC substrates.6'7 In addition, the quality of the ammonia- MBE GaN layers shows little sensitivity to substrate and buffer layer conditions. However, the ammonia-MBE GaN layers typically exhibit significant surface roughness as a result of strong surface faceting inherent to the ammonia-MBE process. The ammonia source is also inefficient for InGaN growth, as the latter requires low growth temperature, which severely limits the ability to thermally crack the ammonia molecules. Tang et.al.8 recently demonstrated that PAMBE grown GaN can be significantly smoother than the underlying ammonia-MBE grown GaN templates. This is an important observation since Am-MBE, unlike PAMBE, is excellent for growing semi-insulating GaN template layers

122

Tang, Webb

(see next section). Fig.l shows the surface morphology of the starting template layer and the surface after the PAMBE growth under Ga rich conditions. The ammonia-MBE template shows significant roughness (RMS roughness of 29nm) with a pronounced faceted grain structure. However, the surface becomes very smooth (RMS roughness of 0.78nm) with clearly resolved atomic terraces after the PAMBE grown GaN layer is grown. The III/V ratio for the PAMBE grown layer was Ga rich, with gallium droplets visible on the surface of the film. Previous studies showed that under gallium droplet forming conditions, the growth of GaN by PAMBE is predominantly via a two-dimensional growth mode, leading to smooth surfaces with clearly defined atomic-steps.3 What was surprising from Tang et al's study was that the diffusion of Ga adatoms was efficient enough to fill in and smoothen out the large facetted valleys on the surface of the ammonia- MBE GaN template.

Fig. 1. AFM micrograph of a carbon-doped GaN layer (l|im thick) grown by plasmaassisted MBE on top of a carbon-doped GaN template grown by ammonia-MBE.8

Using a combination of ammonia and plasma assisted MBE, high quality AlGaN/GaN superlattices were grown on insulating templates. Good surface and interface quality is observed by AFM and x-ray diffraction studies. Fig.2 shows the x-ray diffraction spectrum around the (0002) reflection peak for a 10 period Al0.3Gao.7N (lOnm) / GaN (lOnm) superlattice grown by PAMBE at 700°C. Satellite peaks up to fifth order are clearly seen, indicative of good interface quality. The smooth

Molecular Beam Epitaxy for III-N Materials

123

interfaces are of benefit for growth of AlGaN/GaN HEMTs which will be discussed later. 105 -

Intensity

&

10

-

Mo

10 3 -

-1|

o

zj

GaN (0002)

1 1

-2

o

^

4

+1

+2

/ \J\

-3 -5 -4

10° 10-' -

\

i

i

i

1

16

i —r-

~!

1

I

I

,

17

,

I

18

1

,

I

1

,'

19

6(degree)

Fig. 2. XRD data for a 10 period AlGaN/GaN superlattice grown by plasma-assisted MBE on an in situ GaN/sapphire template prepared by ammonia-MBE.8

3. Semi-insulating GaN

3.1. Growth of Semi-insulating GaN AlGaN/GaN high electron mobility transistors (HEMTs) require a highly resistive GaN base layer to isolate the 2DEG channel from any parasitic leakage. Good isolation greatly improves the pinch-off characteristics of the transistors and greatly improves their R.F. performance. Growth of semi-insulating GaN, however, has not been straightforward since most GaN epilayers grown by techniques such as metalorganic vapour phase epitaxy (MOVPE), molecular-beam epitaxy (MBE), and hydride vapor phase epitaxy(HVPE), have a significant n-type residual carrier density. This background n-type doping arises from both native defects, which are plentiful in GaN films grown on substrates with large lattice

124

Tang, Webb

mismatch, and from residual impurities such as silicon and oxygen. Compensation or removal of these background n-type dopants is essential in order to obtain insulating or highly resistive layers. A number of approaches have been tried to compensate the residual donor levels such as, varying the growth conditions to change the profile of the structural defects and/or residual impurities in the films, and introducing various dopants to provide compensation.911 For example, in MOVPE processes, semi-insulating GaN could be obtained for a growth pressure of 76 torr at low NH3 flow rates.9 However the observed increase in resistivity for decreasing NH3 flow rates, was not consistent with the expected increase in shallow donors-resulting from the increase in N vacancies at low Willi ratios. In contrast, highly resistive GaN layers were obtained using reactive MBE by varying the III/V ratio and the growth temperature.11'12 Unlike the MOVPE results, the highest resistivity material was obtained at high nitrogen fluxes (high Willi ratio). Based on stoichiometry arguments and the electrical and PL data, the residual shallow donors, shallow acceptors and deep centres were attributed to the N vacancy, Ga-antisite/Ga-vacancy complex and the Gaantisite respectively. However, both MOVPE and PAMBE methods placed severe constraints on the range of available growth parameters, provided only limited control over the native defects and residual impurities, and thus reproducibility was difficult to achieve. 3.2. Carbon Doping Clearly a method to reproducibly grow semi-insulating GaN was needed. In an early paper by Boguslawski et.al.13 it was reported that their quantum molecular dynamics calculations showed that carbon in GaN and A1N would be amphoteric giving essentially similar concentrations of CN and Ccatjon, but with a small preference for the incorporation of carbon on the nitrogen site. In 1999, Webb et.al.14 demonstrated that carbon could be used as a dopant to obtain semi-insulating GaN. GaN grown by ammonia-MBE was shown to become highly resistive when doped with carbon using a gaseous methane ion-beam source. In the absence of cracking via the ion beam gun, it was found that methane was essentially inert and no carbon incorporation was observed [see Fig.3].

Molecular Beam Epitaxy for Hl-N Materials

125

Fig. 3. SIMS carbon profile as a function of depth from the surface of the epilayer (a) for a sample grown with fixed carbon ion energy and with the methane flow and the ion gun turned on and off during the growth; (b) for a sample grown with various anode voltages of the ion source. Note that film growth is from right to left.14

However precisely controlled concentrations of carbon could be introduced into the epilayers when the CH4 was cracked in the ion gun plasma, with the result that resistivities >108Q-cm could easily be obtained [Fig.4]. Of particular note was the lack of any memory effect in the MBE chamber from the dopant source, and the demonstration that a high mobility 2DEG could be grown on the C-GaN template with no observable parasitic conduction paths. Recently additional compounds such as CC14, CS215 and CBr416 were found to be effective sources for the introduction of controlled amounts of carbon into PAMBE GaN layers. PAMBE GaN doped with carbon using these sources resulted in insulating material, however the dependence of the resistivity with increasing carbon incorporation was unpredictably different from that prepared by ammonia-MBE. This will be discussed in later. The electrical properties of the carbon doped layers have been studied by a number of authors.15'17'18 Despite the high resistivity of the layers-

126

Tang, Webb

which has made electrical measurements difficult to performconductivity activation energies of 0.3eV to 0.7eV were observed for the PAMBE prepared samples (activation energy increasing with increasing carbon incorporation).15 Tang et.al.17 observed an activation energy of 0.9eV for highly resistive ammonia-MBE layers >108Q-cm, however in a later report for samples prepared under similar conditions, activation energies EA ranging from 0.74 to 1.2eV were observed, dependent on carbon concentration.18 In addition, TSC measurements on these samples indicated a number of clearly defined trap levels that correlated well with DLTS measurements. Traps designated as Ai (~0.90-1.0eV) dominated in GaN with lower concentrations of [C] but as [C] increased, traps A (0.67eV) and B (0.58eV) reduced, and a new trap Bx (0.50eV) appeared. It was suggested that traps A b A and B, and B x were related to Nh VGa, and CGa (or CGa-CN), respectively.

1.4

1.6

1.8

1000T(K"1) Fig. 4. Temperature dependence of resistivity and Hall-effect data for a carbon-doped GaN layer of 2 urn thickness grown by ammonia-MBE.17

Seager et.al.19'20 examined both experimentally and theoretically the role of carbon in MOVPE grown epilayers of GaN. The PL data observed for both low and high concentrations of carbon are shown in Fig.5 They observed that when the carbon concentration was less than

Molecular Beam Epitaxy for III-N Materials

127

the background level of Si in the layers, Nd-Na was close to the difference between the silicon concentration minus the carbon concentration as determined from SIMS measurements. However when the carbon concentration exceeded that of Si, the GaN layers were almost entirely compensated and Nd-Na dropped more than two orders of magnitude below that of either carbon or silicon and the layers were highly resistive. It was noted that the peak in the PL spectrum at 415nm (~3eV Figure 5) was only observed when the carbon concentration exceeded that of Si. A similar peak was also observed near 3eV in carbon doped PAMBE GaN layers,1618 and followed the same behaviour as that reported by Seager et.al. (i.e. an increase with increasing carbon concentration). In contrast, this peak was suppressed for increasing carbon concentrations for the ammonia-MBE layers, [see Fig.6] 2.5

r

n

2

r

!!

-°

;

_

_

_

J

? •f

1.C

m o 0.5

0 . 0 " * " • » « - ' " ——S' • " • " '.'.-.• ••*>•'' •' • - * * * — . . . . 1 300 400 600 600 700 400 WAVELENGTH (nm)

Fig. 5. Cathodoluminescnece intensity vs wavelength for low carbon concentration (dashed Line) and for a carbon concentration greater than the background Si density (solid line). The dashed curve has been multiplied by a factor of 20. (Reprinted with permission from C.H.Seager, A.F.Wright, J.Yu, and W. Gotz, Journal of Applied Physics, 92, 6553(200) Copyright 2002, American Institute of Physics).

Based on Density-functional-theory (DFT) calculations, Seager et.al. were able to provide a straightforward explanation for the observed behaviour. Consistent with the work of Boguslawski et.al.13 the

Tang, Webb

128

incorporation of carbon on the nitrogen site CN is found to be favoured under all growth conditions when the Fermi level was in the upper half of the band gap (when [C]« [Si]) and thus carbon will compensate the Si donors. As the carbon concentration approaches that of Si, the Si donors become increasingly compensated and the Fermi level shifts toward the middle of the gap resulting in an increase in electrical resistivity. When the carbon concentration exceeds that of Si, the material becomes semiinsulating due to self compensation of the carbon species. They noted that the self-compensation depended on growth stoichiometry. When growth occurs under N-rich conditions (as in the case of ammonia-MBE grown GaN), CN and CGa incorporate in roughly equal numbers which results in a Fermi level pinned near midgap, while for Ga-rich conditions (PAMBE) the CGa is preferred and increasing amounts of carbon results in a decrease in resistivity. !

20mT «mT

SmT 1mT NoCBr,

'• ' \

*•*».,

W*A '

%«L'

1,5 2,0 2,5 3.0 3.5 4.< Energy (eV)

Energy (eV)

Fig. 6. PL spectra for C:GaN samples prepared by (a) PAMBE and (b) ammonia-MBE with increasing concentrations of C (SI to S4: l-20mT). (a)Reprinted with permission from D.S.Green, U.K.Mishra and J.S.Speck, J.of Appl.Physics,95,8456(2004).Copyright 2004, American Institute of Physics, (b) Ref.18.

Tang et.al.17 proposed a grain boundary model to explain their carbon doped material, although the more recent data of Fang et.al.18 concluded that a model similar to that proposed by Seager would be consistent with their results. The DFT calculations would also explain the anomalous preferred behaviour of carbon incorporation as donors

Molecular Beam Epitaxy for IIJ-N Materials

129

when the carbon concentration increases above the background Si concentration, as observed by Armitage et.al.13 for PAMBE grown GaN using CC14 and CS2. 3.3. Other Dopants for Semi-insulating GaN Although carbon has been more widely studied than any other dopant for compensation of residual n-type carriers in GaN, researchers have also examined the use of Mg (used as a p-type dopant)21 as well as Zn22 and more recently Fe.23'24'25 In MOVPE grown GaN, doping with Mg results in high resistivity GaN epilayers. Annealing after growth activates the Mg dopant resulting in p-type conduction. Thus Mg doped epilayers can be effectively used in HEMT devices provided the temperature range of operation is below that of the dopant activation temperature. Unfortunately this process is not effective in MBE grown epilayers, since post deposition annealing to activate the Mg acceptors is not required. Zinc has also been used to deposit insulating GaN (1012fi-cm at 300K) on SiC substrates by HVPE. Resistivity measurements of the layers showed two distinct thermally activated regions with energies of 0.18eV and 0.7 eV. The 0.7 eV activation energy observed for temperatures below 300K was attributed to the Zn acceptor level while the origin of the 0.18eV level remained unclear. In the case of MBE prepared GaN, no reports on the use of Zn for this application have been published, however the high vapour pressure of Zn would certainly require increased precaution to limit cross contamination of other sources and components in the MBE environment. Iron doping has also been studied, although very little new data has emerged since the first report by Heikman et.al.23 on the growth of insulating GaN prepared by MOVPE using ferrocene (Cp2Fe, bis(cyclopentadienyl)iron). Samples grown with an Fe concentration of 1.3 x 1019cm"3 had a resistivity of 7 x 109Q-cm and excellent crystallinity of the layers was maintained. However the Fe precursor showed memory effects similar to Mg, making abrupt doping difficult to achieve. Vaudo et.al.24 reported the use of iron chloride in HVPE grown epilayers. The resistivity and PL data for these layers is shown in Fig.7 A single activation energy of 0.5 leV was observed while the PL data showed a

Tang, Webb

130

clear "green" band associated with the Fe impurities(as opposed to the normal "yellow" and "blue" bands observed in C:GaN). MOVPE grown Fe doped layers were also reported by Polyakov et.al.25 but no mention of the Fe precursor used to dope the layers was given. Micro cathodoluminescence data showed a behaviour different from that of Vaudo et.al. and indeed a "blue" emission band near 3eV was observed, suggesting possible carbon incorporation in the layers. At present, Fe doping has not been reported in MBE grown GaN. However, the very effective use of carbon doping in preparing these insulating layers by MBE, has essentially made the search for an alternative dopant unnecessary.

2.6 3,0 &wj)r(»V)

2.0

2.2 2.4 26 1000/T (K1)

2.8 3.0

Fig. 7. Temperature dependence of resistivity and the observed PL spectra from a semiinsulating iron-doped GaN sample at 6 K Ref.24.

4.

Electrical Characteristics

4.1. Polarization Induced Two Dimensional Electron Gas The two dimensional electron gas associated with a GaN/AlGaN heterostructure is largely induced by the polarization charges at the interface.

Molecular Beam Epitaxy for I1I-N Materials

131

Wurtzite GaN and AlGaN crystals are polar along the c-axis. For a GaN/AlxGai_xN heterostructure grown with Ga polarity (i.e. terminating with Ga or Al atoms), the total polarization charge at the interface is the sum of the spontaneous polarization charge and the piezoelectric charge: AP = APspon APpiezo x[Pspon v(AIN)Pspon (GaN)\ spon + "-" piezo =-~l •" / * sport' + 2x

a(AlN)-a(GaN)( a(GaN)

'31

c

" 13 ^ C 33

r

(4.1)

where a, e, and C denotes the lattice constants, the piezoelectric coefficients and the elastic constants, respectively. The polarization charge is immobile charge. Eq. 4.1 shows that this charge increases almost linearly with the Al concentration x in the barrier. Note that e31, £33, C13, C33 are functions of the Al concentration. Calculations show that a free electron gas of similar charge density to that of the fixed polarization charge is induced at the interface, nearly independent of the doping level of the AlGaN barrier [26]. The strong polarization effect is a crucial factor in the design of GaN/AlGaN HFET structures. Figure 8 (a) shows the sheet carrier density versus the Al concentration for a series of GaN/AlGaN HFETs grown by the ammoniaMBE technique. A strong correlation between the carrier density and the Al content is clearly revealed. Therefore, the 2DEG density is almost entirely dictated by the polarization effect, and is increased by increasing the Al content in the barrier. However, the AlGaN barrier is under a tensile strain. Catastrophic structural relaxation occurs if the strain exceeds a critical threshold. Figure 8 (b) shows that the carrier mobility drops abruptly at carrier densities exceeding 1.9 xlO!3cm"2. This threshold density corresponds with an Al concentration of about 35% for barrier thickness from 15nm to 25nm. Strain relaxation in the form of cracks or micro-cracks was observed in samples with larger than threshold strain. The sheet carrier density is an important parameter relevant to the current and power handling capability of the devices. While this parameter should be maximized by using a high Al concentration, it has to be balanced with the increased alloy scattering and risk of structural

132

Tang, Webb

instability associated with the high Al concentration. As illustrated in Fig. 8b, the carrier mobility shows a clear decline with increasing carrier density, indicating the effect of alloy scattering and interface roughness scattering. Catastrophic failure following thermal annealing was observed in devices with Al content very close to the threshold value. The design rule, therefore, is to maximize the carrier density but well within the relaxation limit.

(A

E o

CM

CO

E o

100 0.0

(a)

0.1

0.2

0.3

1

0.4

Al concentration (%)

(b)

2

A/ S (10 1 3 crrf 2 )

Fig. 8. (a) sheet carrier density versus Al concentration in the barrier, and (b) carrier mobility versus sheet carrier density, for a series of GaN/AlGaN HFET layers grown on sapphire substrates by ammonia MBE (Unpublished data by authors).

4.2. Reproducibility and Uniformity The application of GaN/AlGaN HFETs in microwave power amplifiers and modules requires large periphery devices. To build practical modules and circuits, it is important to be able to scale up the small gate width HFET devices with high power density to high power large periphery devices. The material uniformity across the wafer and the wafer-to-wafer reproducibility is essential for achieving these applications. In principle, it is easier to obtain good wafer uniformity and reproducibility with the MBE technique than with the MOCVD technique because of simpler, more precise control of the supply and incorporation of the material species.

133

Molecular Beam Epitaxy for III-N Materials 75%

wo CM

S» 1000 1000

l

I

25%

100 0 (a)

i

1

1

r

10

20

30

40

50

(b)

Sample Number #

tt 15 Run number

Intentional variation of sheet resistance

VV»

0

(c)

Run number

Fig. 9. run-to-run reproducibility of electrical characteristics for growth of GaN/AlGaN HFETs by (a) ammonia MBE27 © 2 0 01 IEEE, (b) and (c) plasma-assisted MBE, Ref.28.

The MOCVD technique is more prone to irregularities in the more complex gas flow dynamics and gas phase reactions, which can cause microscopic non-uniformity and compromise the reproducibility. Molecular beam epitaxy has been historically the preferred technique for the commercial production of AlGaAs/InGaAs/GaAs HFETs epi-wafers. Although the MOCVD technique has been the overwhelming production tool for GaN-based LEDs and laser diodes, the MBE technique has proved to be a highly competitive technique for the production of GaNbased electronic devices. Both the ammonia-MBE and plasma-assisted MBE techniques demonstrated good uniformity and reproducibility for producing GaN/AlGaN HFETs.

134

Tang, Webb

Figure 9 (a) shows the statistics of room temperature mobility of all 50 GaN/AlGaN HFETs grown on sapphire wafers by ammonia- MBE over one year period. About 75% of the growth experiments yielded mobilities higher than 800cm2/Vs. The yield of high mobility structures is quite satisfactrory considering that the statistics include samples of various structures grown under different conditions. Figure 9 (b) shows the run-to-run reproducibility of GaN/AlGaN HFETs grown by plasma-assisted MBE on SiC wafers.27 For 23 consecutive growth runs, a room temperature Hall effect mobility of approximately 1200cm2/Vs was consistently obtained. Figure 9 (c) shows the run-to-run reproducibility of the sheet resistance of the HFETs grown on SiC wafers by plasma-assisted MBE.28 The yield and reproducibility is excellent for a total of 113 wafers.

30

20

10 -

see

tear

1S0O

Fig. 10. Maximum drain current density across a quarter of a 2in. GaN/AlGaN HFET sample.30

GaN/AlGaN HFETs with good electrical characteristics have also been successfully grown on resistive silicon substrates using the ammonia-MBE growth technique. The mobility of the HFET-on-silicon wafers is consistently higher than 1000cm2/Vs, with the highest reported value reaching 1600cm2/Vs.29 Good uniformity of the electrical characteristics across the wafer suitable for fabrication of microwave integrated circuits (MICs) has been

Molecular Beam Epitaxy for III-N Materials

135

demonstrated using the MBE technique. The uniformity was characterized by comprehensive measurement of the maximum drain current of devices across the wafer. The characterization results for an HFET wafer grown by plasma-assisted MBE are presented in Fig. 10 as a histogram of the measured drain currents.30 The variation in /maXis less than 5% . This level of uniformity is comparable to what is found in the III-V (GaAs and InP) device technologies. 4.3. Low Temperature Mobility The room temperature mobility of a high density GaN/AlGaN 2DEG is limited predominantly by the polar optical phonon scattering, and shows little impact from defects and impurities. The screening effect due to the high density of a 2DEG further reduces the influence of the defects. Therefore, the room temperature mobility is not a good indicator of the material quality of the HFET structures grown by the various techniques. At extreme low temperature and for a low-density 2DEG, however, the lattice vibration is no longer the dominant carrier scattering mechanism. Instead, the electron interactions with various structural imperfections (dislocations, impurities, alloy randomness, interface roughness etc.) emerge as the important mobility limiting scattering sources. The low temperature mobility is a more accurate reflection of the material purity and perfection that is attained by any particular growth technique. In this respect, the molecular beam epitaxy technique has proved to be the leading technology for producing extremely high quality GaN/AlGaN 2DEG structures. In fact, the only GaN/AlGaN 2DEGs with low temperature mobility exceeding 50000cm2/Vs reported in the literature have all been grown by MBE.31,32 The advantages of the MBE technique include: low residual impurity incorporation, layer thickness uniformity and precision, and sharp interface. Figure 11 shows the temperature dependent Hall measurement data for a GaN/AlGaN 2DEG grown by plasma-assisted MBE on an MOCVD grown GaN/sapphire template.31 The mobility of this structure was 24000cm2/Vs at 77K, and increased to 51700cm2/Vs at 13K.

136

Tang, Webb

25000

T

|"ri'"y»»|™if|-|f"1,f-,,|

j"'"'j|l"-'j.~

£T j

|

• y i - j j r . - - j j ~ |nm-<|«w.-|

-J|,„-

20000 0)

>

\ 15000

d = 31 nm T=13K

.IT 10000

o 5000 0 0.05

0.1

, i X 0.15 0.2 0.25 Alloy composition x

_L 0.3

0.35

Fig. 11. Temperature dependence of the Hall mobility and the sheet carrier concentration in the Al 0.o9 Ga 0 9 1 N/GaN heterostructure with 16 nm AlGaN barrier grown by PAMBE on an MOCVD GaN/sapphire template Ref 31.

The low temperature mobility was found to be a strong function of the Al concentration in the barrier. The ultra high mobility was achieved only in samples with small carrie density and Al concentration. As shown in Fig. 12, the low temperature mobility drops sharply with increasing Al concentration in the barrier. Such a behavior rules out the importance of scattering by charged impurities or dislocations in these high quality structures as the field of these charged centers would be increasingly screened out at higher 2DEG densities. Instead, it indicates that scattering by alloy disorder potential and interface roughness is the most important scattering mechanism. As the 2DEG density increases, the electron wavefunction is closer to the interface, with a portion extended into the AlGaN barrier. The mobility decreases as a result of the increased alloy scattering and roughness scattering. For the sample in Fig. 12, the measured mobility of 51700 cm2/Vs is approaching the calculated low temperature mobility limit of 80000 cmVVs due to alloy scattering.33 This means these MBE grown GaN/AlGaN structures are of excellent material purity, interface quality and low defect density.

Molecular Beam Epitaxy for III-N Materials

100 M~*

25 * . J1

m

z

•

,

10 610*

,,,

— —

j:f'"

,,J3: 510* :

i/

-

y—+ ,!.!*'

\

;'

,,c*

\

,n'

5 -:

*

/

•

310* ~

-! 2 10* 5

:

* - yfcti'

: 110*

«

-/*-,., .

^*""'-*-Hfr™»_».^-#—# ••....,.,».i„,.._^

-JJtt-.'.

410*

\

../

X

CD

i

y

ts 10

I

|

t

x = o,o@ d = 16nm

' *. \ 1g • \ •

8

Tm

—p-TT

137

•• • • - - • - •

•

10 100*0/UK"1)

-

J

' '•••"

100°

Fig. 12. Low-temperature electron mobility in the Al x Ga !_x N/GaN 2DEG structures as a function of alloy composition x.31

Using the plasma-assisted MBE technique, an AlGaN/GaN 2DEG with a record mobility of 75000cm2/Vs at a sheet carrier density of 1.5xl012cm"2 has been grown successfully on a HVPE GaN template with a dislocation density of 2xl08cm~2.34 At such low dislocation density, the scattering by the dislocations is no longer a dominant scattering effect. The measured mobility approaches the calculated mobility limit for alloy scattering. It should be noted that such high mobility 2DEG structures grown by plasma-assisted MBE were obtained uniquely under Ga rich growth conditions and exhibited atomically smooth surface and interface. Growth under nitrogen rich conditions, however, leads to rough surface morphology and significantly reduced mobility. Nearly defect free AlGaN/GaN 2DEGs have been grown by ammonia- MBE on semi-insulating bulk GaN crystals. A mobility of 60000cm2/Vs at 1.5K for a sheet carrier density of 2.4x1012cm~2 has been reported.32 The dislocation density was lower than 2xl08cm"2 in this case, again showing that the mobility was not limited by dislocation scattering in these high quality structures.

138

Tang, Webb

Compared with growth using GaN templates as substrates, direct growth of AlGaN/GaN 2DEGs on sapphire or SiC substrates by MBE is subject to a higher degree of interface roughness and dislocation density. An AlGaN/GaN 2DEG with mobility of 14800cm2/Vs at 1.2K for a carrier density 4xl012cm"2 has been grown on a sapphire substrate by ammonia-MBE.35 Interestingly, this sample exhibited a huge surface roughness (RMS value of 60nm) and very large grain sizes. The large grain size correlates with a small dislocation density. The sample was grown under conditions favorable for dislocations reduction but at the expense of surface smoothness. The defects are major sources of scattering in these structures. 4.4. Magnetotransport Properties Important physical parameters for AlGaN/GaN 2DEG structures such as the carrier concentration, the scattering times, and the effective mass can be extracted from low temperature magnetotransport measurements. The observation of Shubnikov-de Haas (SdH) oscillations in the longitudinal magnetoresistance 7?^ and quantum Hall effect in the transverse magnetoresistance R^ is also an important proof of the existence of a quantized two-dimensional electron gas. The oscillatory part of the longitudinal resistance is expressed as.36 AR^

= 4R0X

(r)exp(

- i / «

rJcos(

ITTF

where rq is the quantum scattering time, (Oc -eBIrn'is

JC )

(4.2)

the cyclotron

2

frequency, e = 7Jh nlm* is the Fermi energy, and X(T) a temperature damping factor given by: X(T) = (2n2kT /fi
likS. v5y

Molecular Beam Epitaxy for III-N Materials

139

The quantum lifetime is usually derived from the so-called Dingle plot. If AR/4RoX(T) is plotted in logarithm against 1/B, the slope gives l/xq with an intercept of 1 at 1/B=0. The electron effective mass can be determined by measuring the oscillation amplitude as a function of temperature. The logarithm of the measured amplitude A divided by temperature T is plotted as a function of temperature. The data are then fit to the following equation derived from Eq. 4.2 using the effective mass as a fitting parameter: In

U^

= C - l n sinh

V J

2n kBTm ehB

(4.5)

where C is a constant at constant field B. AlGaN/GaN 2DEGs grown by MBE on GaN template substrates show excellent material and interface quality, allowing the observation of pronounced quantum effect in the magnetotransport measurements. 4.0. 0.1 j

3,5

of

1J

I

fy = 7

3.0-

I"' a:

""'"•i.j ,

'

•0,11

0

. 1

2

BC0

• ./10

3

/..

>1211

"s

B
Fig. 13. Longitudinal (R^ ) and transverse (R^ ) magnetoresistance vs magnetic field at 50mK. Shubnikov de Hass oscillations begin at 1.8T. The inset shows RJR =(/fxc-/?0)//?o —the low field part of R^ after normalization by the zero-field resistance value R0 . The best fit of the SdHOs amplitude yields the value of rD=3K. Ref.31.

Fig. 13 shows the magnetoresistance (R^ and i?^) data for an AlGaN/GaN 2DEG grown by ammonia-MBE on an insulating bulk GaN crystal substrate.31 Strong SdH oscillations showing spin splitting as well as quantum Hall plateaus are observed. From the period of the SdH

Tang, Webb

140

oscillations, a carrier density of 2.4 xl0 12 cm 2 is extracted, which is in good agreement with the low field Hall measurement value. The Hall mobility reaches 60100cm2/Vs at 1.5K, which corresponds to a transport scattering time of rt - jum* I e = &.2ps . A quantum scattering time tq=0Aps was extracted from the Dingle plot, which is about 20 times shorter than the transport scattering time. Fig. 14 shows the magnetoresistance measurement results for an AlGaN/GaN 2DEG grown by plasma-assisted MBE on a semi-insulating HVPE GaN template. Integer quantum Hall effect was observed, indicative of the lack of parallel conduction in this sample. The values of the quantized Hall resistance at v=7, 6, 5 are labeled and correspond well with the measured quantized resistance.34 The physical parameters extracted from magnetoresistance measurements of several representative high-quality AlGaN/GaN 2DEG samples grown by MBE on GaN templates are listed in Table 1. 800

60OQ

v ~ t9,0W cm* / V *oc 60fl

'/"

,v

400

, A I

im

6000

4QOQ

3M0 S

i

Sill:

/

A

/

4.27kn; v=6

im\nt v«7 £

/

5. i«»; v s .

cc!

ft

- ioao

y4m

1000

8

10

1?

S(T)

Fig. 14. Magnetotransport data for an Alo.03Gao.97N/GaN heterostructure at T=1.7K. The onset of Shubnikov-de Hass oscillations occurs at 2T. The integer quantum Hall states are well resolved by 8T. Reprinted with permission from M.J. Manfra, N.G. Weimann, J.W.P. Hsu, L.N. Pfeiffer, K.W. West, S. Syed, H.L. Stormer, W. Pan, D.V. Lang, S.N.G. Chu, G. Kowach, A.M. Sergent, J. Caissie, K.M. Molvar, L.J. Mahoney, R.J. Molnar, J. Appl. Phys. 92, 338 (2002). Copyright 2002, American Institute of Physics.

Molecular Beam Epitaxy for III-N Materials

141

Table 1 Transport property parameters extracted from magnetoresistance measurements Growth

Template

% (ps)

"^(ps)

m

Ref

2.43

0.5

0.215

31 34

Technique Plasma MBE

MOCVD GaN

Plasma MBE

HVPE GaN

2.35

0.55

-

Ammonia MBE

Bulk GaN

8.2

0.4

-

32

Ammonia MBE

MOCVD GaN

1.65

0.25

0.20

37

The quantum lifetime is given by the total scattering rate, while the transport lifetime is weighted the scattering angle &. — = jp{0)d£l

(4.6)

— = {{l-cos 0)p{0)dQ

(4.7)

The quantum lifetime is found to be significantly smaller than the transport lifetime in these high quality AlGaN/GaN 2DEG samples, with T/Tq ratio ranging from 5 to 20. This means the dominant scattering mechanism is small-angle scattering due to long-range interaction with remote ionized impurities in the barrier or channel, which limits the quantum lifetime but has little impact on the transport lifetime. The transport lifetime, on the other hand, is mainly limited by short-range potentials such as the alloy disorder potential and ionized centers inside the 2DEG region. The surface morphology of the AlGaN/GaN structures grown on the GaN template substrates are very smooth due to lack of lattice mismatch. High mobility AlGaN/GaN heterostructure can also be grown directly on sapphire substrates by ammonia MBE using a high growth temperature around 900°C. The high growth temperature allows achieving large grain size and low dislocation density of 108-109cm*2. However, the epilayers exhibit large surface roughness due to habitual faceting of the textured surface. It is found that such surface morphology has a more severe impact on the quantum lifetime than on the transport lifetime.37

142

Tang, Webb

Fig. 15 shows the effect of surface morphology on the SdH quantum oscillation amplitude by comparing the the SdH data of a very rough AlGaN/GaN 2DEG sample (RMS roughness of 63nm) and a smooth one (RMS roughness of 2nm) grown by ammonia MBE. The smooth sample was grown at 800°C on an MOCVD GaN template, whereas the rough sample was grown at 910°C directly on a sapphire wafer. Despite of the huge difference in surface roughness, the two samples exhibited almost identical Hall mobility of 14500cm2/Vs at 4K, indicating that the transport lifetime is not sensitive to the large surface roughness. However, the SdH oscillations are basically washed out in the rough sample.

(a)

BfTesIa) 150 0.1 -

a _

(b)

1 0 0 -- [c o.o-

<

a

5 6 7 8 9 / BfTeslal/

50

4.

—-^ 1 0
2

4

i 6

i 8

"i 10

BfTesIa)

Fig. 15. Magnetoresistance data measured at 1.2K and respective AFM surface micrograph for (a) an AlGaN/GaN 2DEG grown by ammonia-MBE on an MOCVD GaN/sapphire template, and (b) an AlGaN/GaN 2DEG grown by ammonia- MBE directly on sapphire.

Molecular Beam Epitaxy for III-N Materials 5.5 • •

measured roughness scattering

5.0; 0.5 Q.

D

ma 0.0

—\

0

10

a A

1

1

1

1

r

20

30

40

50

60

70

RMS roughness amplitude A (nm)

Fig. 16. A comparison between the measured quantum lifetime and the calculated roughness scattering quantum lifetime presented against the roughness amplitude for the series of AljGa^N/GaN 2DEG samples studied. The measured Ns and A values were applied in the calculations for each sample.37

The roughness scattering is modeled using a scattering potential given by Ando:38 U*=xA2A2

'•{" depl + 0.5N .)"

exp(-qzAz

I A)

(4.8)

£ £

L0

where A is the amplitude and Athe lateral correlation length of the roughness. Since the roughness is caused by the grain texture, the average grain size is a natural measure of the correlation length. The calculation results show that scattering by roughness potential with a large correlation length (grain size>100nm) is predominantly small-angle scattering with little impact on the transport lifetime. However, it can become a limiting factor for the quantum lifetime. Fig. 16 compares the calculated xq due to roughness scattering and the measured xq values for a series of AlGaN/GaN 2DEG samples showing different roughness amplitude. The good agreement for samples with roughness amplitude>10nm indicates that large surface roughness limits the quantum scattering lifetime. However, for smoother samples (RMS roughness< 5nm), the measured xq falls significantly below the roughness

144

Tang, Webb

scattering Tq. Other scattering mechanisms such as the remote ionized impurity scattering set in as the dominant factor in this case. 5. AlGaN/GaN HFET Device Characteristics There has been growing demand for development of monolithic microwave integrated circuit (MMIC) power amplifiers in the frequency range of l-40GHz for commercial wireless systems and military radar systems. Currently, GaAs based pHEMTs are mostly used for building the power MMICs operating at these frequencies. However, the wide bandgap nitride semiconductors promise higher microwave power performance due to a significantly higher breakdown field of 3MV/cm (compared with 0.4MV/cm for GaAs) and a higher electron saturation velocity of 2xl07cm/s (versus lxl0 7 cm/s for GaAs). The power handling capability of a specific material system is determined by the current carrying capability and the breakdown voltage. The maximum sheet carrier density (>10I3cm"2) in AlGaN/GaN HFETs is 40% higher than that in the GaAs based HEMTs. Although the mobility of AlGaN/GaN HFETs (typically around 1200cm2/Vs) is relatively low, the saturation velocity is slightly higher than that in the AlGaAs/GaAs HEMTs. As a result, the AlGaN/GaN HFETs demonstrate a higher current carrying capability as well as good RF performance. With the rapid development in the nitride growth and processing technologies, the fT and fMax of AlGaN/GaN HFETs with submicron gates are well pushed into the >100 GHz range. The rapid development has seen the power performance of AlGaN/GaN HFETs to surpass other technologies in the S-band and KBand, and to continuously test the potential for higher frequency operation. The nitride MBE technique has proven to be a powerful tool in the development of the AlGaN/GaN HFET devices. Using the two variations of the technique (ammonia MBE and plasma-assisted MBE), AlGaN/GaN HFETs with excellent characteristics have been fabricated on all types of the commonly used substrates: sapphire, SiC and silicon. SiC is the most suitable substrate, possessing the highest thermal conductivity and smallest lattice mismatch with GaN, but is currently subject to high wafer cost. Sapphire is the most widely used substrate at a

Molecular Beam Epitaxy for III-N Materials

145

medium wafer cost. The key drawback of sapphire is the poor thermal conductivity. Flip-chip bonding solution is needed for high power limit operation. Silicon has the largest lattice mismatch with GaN. Special growth schemes involving strain engineering using multiple layers must be used to avoid cracking of the epilayer. However, the low wafer cost, compatibility to silicon and good thermal conductivity are important advantages. The device characteristics for MBE grown HFETs on these different substrates are presented in this section. 5.1. DC and Small-Signal RF Characteristics

5.1.1 Devices Grown on Sapphire and SiC by Ammonia MBE The AlGaN/GaN HFET layers typically consist of a two micron thick, semi-insulating, carbon-doped GaN, a 0.2 micron thick undoped GaN channel layer, and an AlGaN barrier with thickness in the range of 15nm to 25nm and Al concentration in the range of 25% to 35%. Mesa isolation was achieved by CAIBE etching into the insulating carbondoped GaN layer with an etch depth of about 0.3 micron. Annealed Ti/Al/Ti/Au (20/100/45/5 5nm) was used for the source and drain contacts, and Pt/Au (100/200nm) for the gate contact. Various gate dimensions were achieved using photolithography and e-beam lithography. Fig. 17 (a) and (b) show the DC drain output characteristics and the measured fT and fMax over temperature respectively, for a device grown on sapphire substrate with a gatelength of 1 |im and a two-finger total gate width of 100u.m.39 The off-state drain to source breakdown voltage is 33V. A maximum drain current of 949mA/mm and a peak transconductance of 160mS/mm were measured. The negative output conductance at high Vds is attributed to heating, which is accentuated by the poor thermal conductivity of the sapphire substrate. At maximum drain current, the channel temperature is estimated to increase to 282°C, which causes the carrier mobility to reduce to 240cm2/Vs from its room temperature value of 1053cm2/Vs. Room temperature RF

146

Tang, Webb

characterization with Vds at 10 volts indicate device fT and fMax to be 15.6GHz and 49.4GHz respectively.

0

2

4

(a)

-50

6 Vds (volts)

(b)

0

50 100 150 200 Chuck Temperature (C)

250

Fig. 17. (a) Measured Id Vd characteristics for a 2 gate lm x 50m wide device. Vgs from +1V to -6V step =-1V. From linear region, Rs=Rd~2-mm. (b) Measured fT and fMAx o v e r temperature for the same device. Vds is 10 volts.39

0 a

4M

UPG

30CO

c '(5

a

20-

10V^IOV

^

\

V 9 =2.5V

oJ 0

5

10

15

Source-drain voltage V^ (V)

20

10

100

Frequency (GHz)

Fig. 18. DC drain output characteristics and small-signal RF characteristics for a two finger 0.15iim x 40um AlGaN/GaN HFET grown on SiC by ammonia MBE.40

Molecular Beam Epitaxy for I1I-N Materials

147

Fig. 19. SEM micrograph of the T-gate device in Fig. 18, showing a foot length of about 0.15m.40

The DC and RF characteristics of devices grown on SiC substrates with e-beam lithography defined T gate (0.15um foot) are shown in Fig. 18. The total (two-finger) gate width is 80 um in these devices. Themaximum drain current exceeds 1.2A/mm and shows negligible much smaller negative output conductance than the devices on sapphire substrates. This attests to the superb thermal conductivity of the SiC substrates. Rf measurements yielded an fT of 103GHz and fMax of 170GHz. The current-gain cutoff frequency fT scales nicely with the inverse gatelength of the devices with various gatelength values. Increasing the operation frequency of the AlGaN/GaN HFETs depends on deep submicron processing using e-beam lithography. Fig. 18 shows an SEM image of the T-gate device in Fig. 19, processed using threelayer resist, e-beam lithography. 5.1.2 Devices Grown on Resistive Silicon by Ammonia MBE Remarkable progress has been achieved in the effort to grow AlGaN/GaN HFET on silicon wafers using the ammonia-MBE technique.

148

Tang, Webb

The use of AlN/GaN interlayers balances the strain and allows to obtain crack free layers with reasonably low dislocation density (109cm"2). The highly resistive silicon substrates (p=20000ncm) minimize the capacitive charge coupling through the substrate during device operation, making it possible to obtain high cutoff frequencies. The mesa patterns were defined by reactive ion etching using a BCI3/CI2 plasma. Ti/AI/Ni/Au ohmic contacts were evaporated and alloyed at 875 °C for 30s. This resulted in an ohmic contact resistance of Rc= 0.4 0 mm. Fig. 20 shows the DC drain output characteristics for a 50(im wide single finger device with a 0.25um gate defined by e-beam lithography.41 The sheet carrier density and carrier mobility for this wafer are 9.6x1012cm"2 and 1440cm2/Vs respectively. The maximum drain current reaches 1.1 A/mm nd shows negligible decline due to thermal effect, attesting to the improved thermal dissipation with a silicon substrate. Small signal RF measurements yielded a current gain cutoff frequency fT=27GHz and a maximum oscillation frequency fMax=81GHz for a 2x50p,m device biased at Vds=15V and Vgs=-4.5V. These devices also exhibited a breakdown source-drain voltage>50V.

120©

ViaB=1V.4Va*»-1V

1000

f eoo #00-

6: Q

,

.

,

.

,

,

2

4

0

8

10

12

.

. 14

VoftV Fig. 20. IV characteristics of 50m-wide single-finger AIGaN/GaN HEMT on silicon substrate.41

Molecular Beam Epitaxy for III-N Materials

149

5.1.3 Devices Grown on Sapphire and SiC by PAMBE AlGaN/GaN HFETs with excellent DC and RF characteristics have been grown on sapphire substrates as well as SiC substrates using the plasmaassisted MBE technique. Fig.21 shows the DC drain output data and small signal RF measurement data for a device grown by PAMBE on a sapphire substrate42. The layer structure consists of 2p,m undoped GaN, 30nm undoped Al0.24Ga0.76N and a 5nm undoped GaN cap layer. Hall measurements showed a sheet carrier concentration of 1.5 x 1013cm"2 and an electron mobility of 1170cm2/Vs. The device was fabricated using inductively-coupled-plasma reactive ion etch (ICP-RIE) for mesa isolation. Ohmic contacts were formed by rapid thermal annealing of evaporated Ti/AI/Ti/Au at 860°C for 30s. The ohmic contact resistance was typically measured to be 0.35Qmm. Mushroom-shaped gates (Ni/Au) with gatelength of 0.25|Lim were fabricated using electron-beam lithography. The devices had a gate width of 100p.m.

10'

frequency, GHt

|fa|

Fig. 21. (a) DC characteristics of 0.25 x 100m AlGaN/GaN HEMT on sapphire substrate, gate bias was swept from 2 to -8V in steps of -IV; (b) short circuit current gain and maximum power gain versus frequency for typical 0.25 x 100m AlGaN/GaN HEMT on sapphire substrate, device was biased at VDS= 10V and V o s = -6.7V.42

The device exhibited a peak transconductance of 216mS/mm and a maximum drain current density of 1.39A/mm. This is the highest current value ever reported for GaN-based HEMTs on a sapphire substrate. The

150

Tang, Webb

low contact resistance might be responsible for the reduced heating effect at small drain drive voltages. Small signal S-parameter measurements against frequency derived a fT of 67GHz and fMax of 136GHz for this device. Fig.22 shows the DC and small signal RF characteristics of a device grown by PAMBE on a semi-insulating 4H-SiC substrate.43 The epitaxial films consist of a nucleation layer, 2u,m of GaN, and a 24nm AlGaN Schottky barrier layer. The epitaxial layers exhibited room temperature 2DEG sheet carrier density of 1.1 x 1013cm"2and Hall effect mobility of 1100cm2/Vs. The device is a 50nm x 200u.m device fabricated with two-finger, T-shaped gates. The source-drain breakdown voltage was measured to be >50V at a gate current density of ImA/mm. The degraded pinch-off at high drain voltages is due to the short channel effects. This deep submicron device exhibited record value of fT=110GHz with fMax>140GHz. The excellent high frequency response indicate that the AlGaN/GaN HFETs have potential for power amplification at frequencies > 20GHz. ' " " » ' ' • * *•>»•* < " > ' y f

.,-ty

MsamB *•.,,

nit

< i l.i.li,itn1uMti i.

,II,IIII«II. I .I,I,HI,.IIII.IIIM

io (b

Fig. 22. (a) Common source I-V characteristics of typical 0.05 x 200m gate AlGaN/GaN HEMT, with the gate bias starting at -7V in IV steps; (b) Short circuit current gain I h2i 12 and maximum stable/available gain (MSG/MAG) of typical 0.05 x 200m gate HEMT biased at VDS=7V and VGS=-4V.43

Molecular Beam Epitaxy for III-N Materials

151

5.2. Large-Signal RF Performance The microwave power limit can be predicted from the DC currentvoltage characteristics by the relation:

p =-i out

Q

an

v max

max

W-1/

The maximum voltage is of course limited by the source drain breakdown voltage. Breakdown voltages up to 400V have been reported for GaN devices due to the high breakdown field of 3MV/cm of GaN.44 The off-state breakdown voltage is linearly dependent on the effective gatelength (the gatelength plus twice the barrier thickness). But bulk and surface defects, doping concentration and processing conditions can also have direct impact on the breakdown characteristics. State-of-the-art AlGaN/GaN power devices typically exhibit off-state breakdown voltages in the range of 50-80V or higher. The maximum DC channel current typically exceeds lA/mm in high performance devices reported. However, in many devices, the measured RF output power is significantly lower than predicted from the DC IV characteristics. The channel current is compressed at microwave frequencies. The RF-DC dispersion is commonly observed in GaN HFETs. Elimination or reduction of the dispersion by surface passivation or improving growth quality has been extensively investigated with various degrees of success. The large-signal microwave power performance of MBE grown AlGaN/GaN HFETs has seen very dramatic progress in the past five years or so. High power performance at different frequencies has been reported from various groups working on MBE grown GaN HFETs. The best results of microwave output power from small periphery devices include: 8.2W/mm at 2GHz, 6.1W/mm at 7GHz, and 3.2W/mm at 25GHz for unpassivated devices grown by PAMBE on 6H SiC wafers[45-47]; 7.3W/mm at 10 GHz for SiN passivated devices grown by PAMBE on 4H-SiC wafers;48 8.2W/mm at 10GHz and 6.6W/mm at 20GHz for unpassivated devices grown by PAMBE on 4H-SiC wafers;28,49 6.6W/mm at 2GHz and 1.9W/mm at 10GHz for SiN passivated devices grown by ammonia-MBE on resistive silicon wafers.50'51 Obviously, the effort to grow the devices on SiC substrates by plasma-assisted MBE has produced the most remarkable progress. It

152

Tang, Webb

is also noteworthy that such high RF power was achieved on unpassivated devices. This is in contrast to the MOCVD grown devices for which surface passivation is usually required to obtain high microwave output power. Fig.23 shows the DC characteristics for a 2x0.15|xm x lOO^in Alo.3Gao.7N/GaN device and large-signal RF characteristics for a 2x0.15u,m x 50u.m device grown by PAMBE on semi-insulating 4H-SiC substrate.49 The CW load pull measurement at 20GHz yielded a maximum power density of 6.6W/mm with a power added efficiency (PAE) of 35% and gain compression of 4.4dB. The device exhibited an fT of 85GHz and fMax of 140GHz in small-signal measurements The device was not passivated.

MS

S>|NldBm

^ 3

Fig. 23. (a) Common source I-V characteristics of 0.15 x 200m AlGaN/GaN HEMT, gate voltage swept from-7 to 2V in IV steps, (b) Large signal characteristics of 0.15 x 100m AlGaN/GaN HEMT at 20GHz. Device was biased at VDS=20V and IDS=80mA [ref.49].

Fig.24 (a) and (b) show the power performance of devices grown by ammonia-MBE on resistive silicon substrates for measurement frequencies of 2GHz and 10GHz respectively.50'51 These devices were passivated with SiN on the surface. The 2x0.25xl25|im device exhibited an output power density of 6.6W/rrrrn at 2GHz with a PAE of 49%. Small signal measurements yielded an fT of 27GHz and fMax of 81 GHz.

Molecular Beam Epitaxy for III-N Materials

153

The 0.3x300|J.m device exhibited an output power of 1.9W/mm at 10GHz. Small signal measurements yielded an fT of 30GHz and fMax of 72GHz for this device.

-20 3N

-15

-10

-5 6 Pm.iBm

5

10

15

0 ,,,

5

10 P„(dBm>

15

20

Fig. 24. (a) Class AB power measurement at 2GHz and VDS=30V for a two-finger 2x0.25xl25(im AlGaN/GaN device on Si(lll) substrateRef 50. (b) Power characteristics 10GHz and VDS=30V for a 0.3x300iim device on Si(l 11) substrate51 © 2004 IEEE.

The microwave power performance of AlGaN/GaN HFETs is often hampered by the phenomenon of RF-DC drain current dispersion, also known as RF current collapse or premature gain compression under large signal excitation. The dispersion has been attributed to charge trapping at the AlGaN surface, within the AlGaN barrier or within the GaN buffer , trapping in SI SiC substrates and insufficient confinement of the channel charge. For MOCVD grown AlGaN/GaN HFETs, surface trapping has been found to be the prevalent mechanism causing severe dispersion in most devices. As a result, proper surface passivation to stabilize the surface states is invariably necessary for achieving high microwave output power in the MOCVD grown devices. It is quite remarkable that excellent microwave output power has been demonstrated in the AlGaN/GaN HFETs grown by the plasmaassisted MBE technique without surface passivation. The results indicate that surface effects are not the dominant source of dispersion in these MBE-grown layers. Compared with the remarkable results of large signal power performance for the devices grown on SiC wafers by the PAMBE

154

Tang, Webb

technique, results on the power performance of the AlGaN/GaN HFETs grown on sapphire or SiC by the ammonia-MBE technique has been unsatisfactory so far. Although these ammonia-MBE grown devices exhibited excellent DC and small-signal RF characteristics, large signal measurements only yielded up to 1.5W/mm output power in the 4 to 10GHz frequency range. The devices show quite significant RF current collapse and insufficient breakdown voltages. One probable problem is that these devices employed epilayers with high sheet carrier densities >1.5xl013cm"2, which cause a higher degree of interface roughness scattering as well as a higher electric field under the gate. Work is underway to optimize the growth structure and material quality for high power performance, and to apply surface passivation to the ammoniaMBE grown devices. In general, current dispersion and effects of surface passivation in MBE grown AlGaN/GaN HFETs have been much less investigated than in the MOCVD grown devices. Future studies in this area should prove further the potential of the MBE grown AlGaN/GaN devices for high power microwave applications. 5.3. Device Scalability and Large Periphery Performance For practical microwave power IC applications, the high power densities demonstrated by the small gatewidth devices must be scalable to obtain high power of large periphery devices and circuits. The nominally identical components within microwave integrated circuits must also possess matching electrical characteristics. The MBE grown AlGaN/GaN HFET epitaxial structures demonstrated excellent reproducibility and uniformity that allowed device scaling and successful fabrication of large periphery devices and integrated circuits. Continuous-wave, load-pull power measurement of 0.1mm, 1mm and 2mm wide devices with 0.25|i,m gatelength was performed at 10GHz on an AlGaN/GaN HFET-on-SiC wafer grown by PAMBE. The maximum output power of 0.65W, 6.3W, and 10.5W measured for the devices with 0.1mm, 1mm, and 2mm wide gates, respectively, scales almost linearly with device size.27 A single stage X-band power amplifier

Molecular Beam Epitaxy for HI-N Materials

155

by power combining four 1mm devices was designed and successfully fabricated. The power performance of this amplifier at 9GHz is shown in Fig.25. Continuous wave output power of 22.9W was measured at 9GHz with associated PAE of 37%. The maximum power density at the output of the amplifier was 5.72W/mm. After correcting for output matching network insertion loss, the power density is very close to the maximum density of 6.3W/mm measured for an individual 1mm device. This is clear evidence that uniformity of the wafer is suitable for fabrication of microwave integrated circuits. Using the same AlGaN/GaN HFET wafers grown by PAMBE, GaN digital circuits, 31-stage ring oscillators using 217 transistors, have been constructed. 52 Successful operation was demonstrated at test temperature of 265°C. This again demonstrates the MBE GaN HFET technology meets the uniformity requirement for practical IC applications. I- 4 mm Total Periphery V -32V

22. 9 W

CW Operation

40

S

s 5

30

1

30

16

6

25

as

35

P warn)

Fig. 25. Power performance of a single stage GaN HFET power amplifier. The amplifier was built by power combining of four discrete l mm devices. The peak output power of the amplifier is 22.9W27 © 2001 IEEE.

156

Tang, Webb

6. Conclusion After more than a decade of persevering effort by researchers in government, university and industrial laboratories, the development of MBE technologies for Ill-nitride materials has finally made great strides in achieving a high level of material quality and application potential. It is clear that the MOVPE technique is the dominant commercial tool for nitride-based LEDs and laser diodes. However, MBE as well as MOVPE are currently competing techniques for nitride based electronics, a projected multi-billion dollar market over the next 5 to 10 years. The review in this chapter of the progress in the nitride MBE technologies, though far from exhaustive, clearly finds unique advantages and real potential of the MBE techniques for commercial production of GaN high speed, high power devices. The particular merits can be summarized: (1) high purity and interface quality: the highest low-temperature mobility values exceeding 50000cm2/Vs were all achieved with the MBE techniques; (2) controlled carbon doping to obtain highly resistive GaN layers essential for isolation of AlGaN/GaN HFETs; (3) excellent wafer uniformity and device scalability; (4) Small and large signal highfrequency performance matching or exceeding the MOCVD grown devices; (5) two variations of the MBE technique (PAMBE and Ammonia- MBE) offer a wide range of growth parameters. However, it should be noted that there has been much less work on surface passivation and optimized processing of the MBE GaN HFETs than the enormous effort put into maximizing the performance of the MOCVD or MBE GaN HFETs. Since the elimination and/or reduction of the surface trapping by passivation or field-plate schemes is a threshold condition for the commercial applications of the GaN HFET technology, the MBE researchers will have to dedicate or derive from their industry partners more resources towards the device processing challenges. Current trends predict the MBE technique will become an important player in a future world that will be dominated by the dream-come-true GaN electronics. References 1. C.T. Foxon, "Three decades of molecular beam epitaxy ", J of Crystal Growth 251, 1 (2003).

Molecular Beam Epitaxy for III-N Materials

157

2. M. A. Sanchez-Garcia, J. L. Pau, F. Naranjo, A. Jimenez, S. Fernandez, J. Ristic, F. Calle, E. Calleja and E. Munoz, "Plasma-assisted MBE growth of group-Ill nitrides: from basics to device applications", Mat. Sci. andEng. B93, 189 (2002). 3. B. Heying, R. Averbeck, L.F. Chen, E. Haus, H. Riechiert, and J.S. Speck, "Control of GaN surface morphologies using plasma-assisted molecular beam epitaxy", J. Appl. Phys. 88, 1855 (2000). 4. K. Xu, N. Yano, A.W. Jia, A. Yoshikawa, and K. Takahashi, "Kinetic Process of Polarity Selection in GaN Growth by RF-MBE", phys. stat. sol. (b) 228, 523 (2001). 5. X.-Q. Shen, T. Ide, S.-H. Cho, M. Shimizu, S. Hara, H. Okumura, S. Sonoda, and S. Shimizu, "Essential Change in Crystal Qualities of GaN Films by Controlling Lattice Polarity in Molecular Beam Epitaxy", Jpn. J. Appl. Phys. 39, LI 6 (2000). 6. H. Tang and J.B. Webb, "Growth of High Mobility GaN by Ammonia- Molecular Beam Epitaxy", Appl. Phys. Lett. 74, 2373 (1999). 7. J.B. Webb, H. Tang, J. A. Bardwell, and P. Coleridge, "Growth of High Mobility GaN and AlGaN/GaN HFET Structures on 4H-SIC by Ammonia-Molecular-Beam Epitaxy", Appl. Phys. Lett. 78, 3845 (2001). 8. H. Tang, S. Rolfe, M. Beaulieu, S. Haffouz, and J. Webb, "Plasma-assisted MBE Growth of GaN on GaN/sapphire Templates Grown in situ by Ammonia- MBE", State-of-the-Art Program on Compound Semiconductors XLI and Nitride and Wide Bandgap Semiconductors for Sensors, Photonics, and Electronics V. The Electrochemical Society Proceedings Vol. 2004-06, 215 (2004). 9. K.C.Feng, J.Y.Lip and H.X.Jiang, "Optical properties of a high-quality insulating GaN epilayer", Appl. Phys. Lett. , 74, 3821 (1999). 10. Jae-Hoon Lee, Myoung-Bok Lee, Sung-Ho Hahm, Yong-Hyan Lee, Jung-Hee Lee, Young-Ho Bae and Hyan Kyung Cho, "Growth of Semi-insulating GaN Layer by Controlling Size ofNucleation Sites for SAW Device Applications", MRS Internet J. Nitride Semicond.Res. 8, 5 (2003). 11. D. C. Look, D. C. Reynolds, W. Kim, O. Aktas, A. Botchkarev, A. Salvador, and M. Morkoc, "Deep-center hopping conduction in GaN", J. Appl. Phys. 80, 2960 (1996). 12. D.C.Look, D.C.Reynolds, R.L.Jones, W.Kim, O.Aktas, A. Botchkarev, A.Salvador and H.Morkoc, "Electrical and optical properties of semi-insulating GaN", Mat.Sci. and Eng. B44, 423 (1997). 13. P.Boguslawski, E.L.Briggs and J.Bernholc, "Amphoteric properties of substitutional carbon impurity in GaNandAIN", Appl. Phys. Lett. 69, 233 (1996). 14. J.B.Webb, H.Tang, S.Rolfe and J.A.Bardwell, "Semi-insulating C-doped GaN and High Mobility AlGaN/GaN Heterostructures Grown by Ammonia- Molecular Beam Epitaxy", Appl.Phys.Lett. 75, 953 (1999). 15. R.Armitage, Q.Yang, H.Feick and E.R.Weber, "Evaluation of CCl4 and CS2 as carbon doping sources in MBE growth of GaN", J. Cryst. Growth 263,132 (2004). 16. D.S.Green, U.K.Mishra and J.S.Speck, "Carbon doping of GaN with CBr4 in radiofrequency plasma-assisted molecular beam epitaxy", J.Appl.Phys. 95, 8456 (2004).

158

Tang, Webb

17. H.Tang, J.B.Webb, J.A.Bardwell, S.Raymond, Joseph Salzman and C.Uzan-Saguy, "Properties of carbon-doped GaN", Appl.Phys.Lett. 78, 757 (2001). 18. Z-Q. Fang, D.C. Look, B. Claflin, S. Haffouz, H. Tang, and J. Webb, "Thermally stimulated current spectroscopy and photoluminescence of carbon-doped semiinsulating GaN grown by ammonia-based molecular beam epitaxy", phys. stat. sol. (c) 2, 2757 (2005). 19. C.H.Seager, A.F.Wright, J.Yu and W. Gotz, "Role of carbon in GaN", J.Appl.Phys. 92, 6553 (2002). 20. C.H.Seager, D.R.Tallant, J.Yu and W.Gotz, "Luminescence in GaN co-doped with carbon and silicon", J. Luminescence 106, 115 (2004). 21. Y.K.Su, S.J.Chang, T.M.Kuan, C.H.Ko, J.B.Webb, W.H.Lan, Y.T.Chemg, S.C.Chen, "Nitride-based HFETs with carrier confinement", Mat. Sci. Eng. B 110, 172 (2004). 22. N.I. Kuznetsov, A.E.Nikolaev, A.S. Zubrilov, Yu V. Melnik and V.A. Dimitriev, "Insulating GaN:Zn layers grown by hydride vapor phase epitaxy on SiC substrates", Appl. Phys. Lett, 75,3138 (1999). 23. Sten Heikman, S.Keller, St.DenBaars and U.K.Mishra, "Growth of Fe doped semiinsulating GaN by metalorganic chemical vapor deposition", Appl. Phys. Lett. 81, 439 (2002). 24. R.P.Vaudo, X.Xu, A.Salant, J.Malcarne and G.Brandes, "Characteristics of semiinsulating, Fe-doped GaN substrates", phys. stat. sol. (a) 200, 18 (2003). 25. A.Y.Polyakov, N.B.Smimov, A.V.Govorkov and S.J.Pearton, "Electrical and optical properties of Fe-doped semi-insulating GaN templates", Appl. Phys. Lett. 83,3314(2003). 26. O. Ambacher, J. Smart, J.R. Shealy, N.G. Weimann, K. Chu, M. Murphy, W.J. Schaff, L.F. Eastman, R. Dimitrov, L. Wittmer, M. Stutzmann, W. Rieger, and J. Hilsenbeck, "Two-dimensional electron gases induced by spontaneous and piezoelectric polarization charges in N- and Ga-face AlGaN/GaN heterostructures", J. Appl. Phys. 85, 3222 (1999). 27. M. Micovic, A. Kurdoghlian, P. Janke, P. Hashimoto, D.W.S. Wong, J.S. Moon, L. McCray, C. Nguyen, "AlGaN/GaN Heterojunction Field Effect Transistors Grown by Nitrogen Plasma Assisted Molecular Beam Epitaxy", IEEE Transactions on Electron Devices. 48, 591 (2001). 28. M. Micovic, J.S. Moon, T. Hussain, P. Hashimoto, W.S. Wong, L. McCray, "GaN HFETs on SiC substrates grown by nitrogen plasma MBE", phys.stat. sol. (a) 188, 31 (2001). 29. F. Semond, P. Lorenzini, N. Grandjean, J. Massies, "High-electron-mobility AlGaN/GaN heterostructures grown on Si (111) by molecular-beam epitaxy", Appl. Phys. Lett. 78, 335 (2001). 30. N.X. Nguyen, C. Nguyen, D.E. Grider, "High performance GaN/AlGaN MODFETs grown by RF-assisted MBE", Electron. Lett. 34, 811 (1998).

Molecular Beam Epitaxy for IH-N Materials

159

31. I.P. Smorchkova, C.R. Elsass, J.P. Ibbetson, R. Vetury, B. Heying, P. Fini, E. Haus, S.P. DenBaars, J.S. Speck, U.K. Mishra, "AIN/GaN and (Al,Ga)N/AlN/GaN twodimensional electron gas structures grown by plasma-assisted molecular-beam epitaxy", J. Appl. Phys. 86, 4520 (1999). 32. E. Frayssinet, W. Knap, P. Lorenzini, N. Grandjean, J. Massies, C. Skierbiszewski, T. Suski, I. Grzegory, S. Porowski, G. Simin, X. Hu, M. Asif Khan, M.S. Shur, R. Gaska, D. Maude, "High electron mobility in AlGaN/GaN heterostructures grown on bulk GaNsubstrates", Appl. Phys. Lett. 77, 2551 (2000). 33. D. Jena, Y. Smorchkova, A. C. Gossard, and U. Mishra, "Electron Transport in IIIVNitride Two-Dimensional Electron Gases", phys. stat. sol. (b) 228, 617 (2001). 34. M.J. Manfra, N.G. Weimann, J.W.P. Hsu, L.N. Pfeiffer, K.W. West, S. Syed, H.L. Stormer, W. Pan, D.V. Lang, S.N.G. Chu, G. Kowach, A.M. Sergent, J. Caissie, K.M. Molvar, L.J. Mahoney, R.J. Molnar, "High mobility AlGaN/GaN heterostructures grown by plasma-assisted molecular beam epitaxy on semiinsulating GaN templates prepared by hydride vapor phase epitaxy", J. Appl. Phys. 92, 338 (2002). 35. H. Tang, J.B. Webb, S. Rolfe, J.A. Bardwell, D. Tomka, P. Coleridge, C.H. Ko, Y.K. Su, and S.J. Chang, "GaN/AlGaN two-dimensional electron gas grown by ammonia- MBE on MOCVD GaN template", phys. stat. sol. (b) 234, 822 (2002). 36. P.T. Coleridge, R. Stoner, and R. Fletcher, "Low-field transport coefficients in GaAs/Ga^ljAsheterostructures", Phys. Rev. B 39, 1120 (1989). 37. H. Tang, J.B. Webb, P. Coleridge, J.A. Bardwell, C.H. Ko, Y.K. Su, and S.J. Chang, "Scattering lifetimes due to interface roughness with large lateral correlation length in AlxGaj.xN/GaN two-dimensional electron gas", Phys. Rev. B66, 245305 (2002). 38. T. Ando, A.B. Fowler, and F. Stern, "Electronic properties of two-dimensional systems", Rev. Mod. Phys. 54, 437 (1982). 39. T. W. MacElwee, J. A. Bardwell, H. Tang, and J. B. Webb, "Characterization of AlGaN/GaN HEMT Devices Grown by MBE", Mater. Sci. Forum 338-342, 1647 (2000). 40. J.A. Bardwell, Y. Liu, H. Tang, J.B. Webb, S.J. Rolfe, and J. Lapointe, "AlGaN/GaN HFET devices on SiC grown by ammonia- MBE with high fT and fMAX", Electron. Lett. 39, 564-566 (2003). 41. R. Behtash, H. Tobler, M. Neuburger, A. Schurr, H. Leier, Y. Cordier, F. Semond, F. Natali, J. Massies, "AIGaN/GaN HEMTs on Si(lll) with 6.6 W/mm output power density", Electronics Letters. 39, 626 (2003). 42. V. Kumar, W. Lu, R. Schwindt, J. Van Hove, R. Chow, I. Adesida, "0.25 jjm gatelength, MBE-grown AlGaN/GaN HEMTs with high current and highfT", Electron. Lett. 37, 858 (2001). 43. M. Micovic, N.X. Nguyen, P. Janke, W.S. Wong, P. Hashimoto, L.M. McCray, C. Nguyen, "GaN/AIGaN high electron mobility transistors with fT of 110GHz", Electron. Lett. 36, 358 (2000).

160

Tang, Webb

44. E. Kohn, I. Daumiller, P. Schmid, N.X. Nguyen, C.N. Nguyen, "Large signal dispersion of GaN/AlGaN heterostructure field-effect transistors", Electron. Lett. 35,1022(1999). 45. M. Manfra, N. Weimann, Y. Baeyens, P. Roux, D.M. Tennant, "Unpassivated AlGaN/GaN HEMTs with CW power density of 3.2 W/mm at 25 GHz grown by plasma-assisted MBE", Electron. Lett. 39, 694 (2003). 46. M.J. Manfra, N.G. Weimann, O. Mitrofanov, T. Waechtler, D.M. Tennant, "High power GaN/AlGaN/GaN HEMTs operating at 2 to 25 GHz grown by plasmaassisted MBE", phys. stat. sol. (a) 200, 175 (2003). 47. M. Manfra, N. Weimann, Y. Baeyens, P. Roux, D.M. Tennant, "Unpassivated AlGaN/GaN HEMTs with CW power density of 3.2 W/mm at 25 GHz grown by plasma-assisted MBE", Electron. Lett. 39, 694 (2003). 48. S. Rajan, P. Waltereit, C. Poblenz, S.J. Heikman, D.S. Green, J.S. Speck, U.K. Mishra, "Power Performance of AlGaN-GaN HEMTs Grown on SiC by PlasmaAssisted MBE", IEEE Electron Dev. Lett. 25, 247 (2004). 49. J.S. Moon, M. Micovic, P. Janke, P. Hashimoto, W.S. Wong, R.D. Widman, L. McCray, A. Kurdoghlian, C. Nguyen, "GaN/AlGaN HEMTs operating at 20GHz with continuous-wave power density > 6W/mm", Electron. Lett. 37, 528 (2001). 50. R. Behtash, H. Tobler, M. Neuburger, A. Schurr, H. Leier, Y. Cordier, F. Semond, F. Natali, J. Massies, "AIGaN/GaN HEMTs on Si(lll) with 6.6 W/mm output power density", Electron. Lett. 39, 626 (2003). 51. A. Minko, V. Hoel, E. Morvan, B. Grimbert, A. Soltani, E. Delos, D. Ducatteau, C. Gaquiere, D. Theron, J.C. De Jaeger, H. Lahreche, L. Wedzikowski, R. Langer, P. Bove, "AlGaN-GaN HEMTs on Si With Power Density Performance of 1.9 W/mm at 10 GHz", IEEE Electron Dev. Lett. 25, 453 (2004). 52. Hussain, M. Micovic, T. Tsen, M. Delaney, D. Chow, A. Schmitz, P. Hashimoto, D. Wong, J.S. Moon, Ming Hu, J. Duvall, D. McLaughlin, "GaN HFET digital circuit technology for harsh environments", Electron. Lett. 39, 1708 (2003).

CHAPTER 5 GROWTH AND PROPERTIES OF NONPOLAR GaN FILMS AND HETEROSTRUCTURES

Yue Jun Sun* and Oliver Brandt+ Paul-Drude-Institutfiir Festkorperelektronik, Hausvogteiplatz 5-7, D-10117 Berlin, Germany + Email: [email protected] We outline the study of the growth of GaN( 1-100) on yLiAlO2(100) by molecular beam epitaxy. Optimum growth conditions are first established empirically. Then a comprehensive investigation on Ga adsorption and desorption kinetics is presented. The optimum growth conditions with regard to surface morphology of GaN( 1-100) epilayers are those giving rise trilayer coverage at low temperature. Furthermore, growth of (In,Ga)N/GaN(l-100) multiple quantum wells is presented and a detailed investigation of their structural and optical properties is described.

1.

Introduction

Although the development of nitride-based light emitting devices has been progressed with considerable success, their luminous efficiencies, especially in the green or ultraviolet wavelength regions, are still quite low, due to the presence of electrostatic fields within the active layers. 1 These fields are generated by the spontaneous and piezoelectric polarization fields, 2 since most structures are deposited along the conventional growth direction, i. e. [0001]. These polarization fields give

Present address: Aixtron AG, Global Service Operations, ShangHai office, Kuen Yang Building Rm.1903, ZhaoJiaBang Rd. No.798, 200030 ShangHai, China; email: [email protected] 161

162

Sun, Brandt

rise to large internal electrostatic fields. Although this property can be used for certain device application, for example, high electron mobility transistors,3"5 it is not desirable for light emitting devices, as the internal electrostatic field reduces the overlap of the electron-hole wave functions [quantum confined Stark effect(QCSE)].1'6 The poor overlap of electronhole wavefunctions also results in a long radiative life time,7 and consequently low internal quantum efficiency, as there are always competing non-radiative recombination channels at elevated temperatures. One solution to get rid of these notorious internal electric fields is the growth of cubic group Ill-nitride heterostructrures.8'9 However, even after intense efforts,'011 the lack of suitable substrates and the inherent thermodynamic metastability of this phase has detained the achievement of device-quality material. Alternatively, in the thermodynamically stable wurtzite structure of cc-GaN, any direction orthogonal to the [0001] axis, such as [1-100] and [11-20], is free of spontaneous polarization. In addition, these directions also do not carry piezoelectric polarization, if shear stresses in the growth plane are absent. Therefore, due to the resulting absence of electrostatic fields, these directions constitute a distinct advantage for fabricating high-efficiency GaN-based LEDs over corresponding [0001]-oriented structures on conventional substrates such as Al2O3(0001) and SiC(0001). Waltereit et al. first demonstrated an improved internal quantum efficiency of GaN/(Al,Ga)N quantum wells grown along the [1-100] direction on y-LiAlO2(100), as a result of the absence of internal electrostatic fields across the quantum wells.12 This triggered a surge in studying nonpolar group-Ill nitride heterostructures.13-23 Most recently, LEDs based on non-polar A-plane and M-plane nitrides have been realized.24-26 In this chapter, we present an investigation of the synthesis of wurtzite (In,Ga)N(l-100) heterostructures on y-LiAlO2(100) by plasma-assisted molecular beam epitaxy (MBE). First a brief description of y-LiA102, used as substrate in this study, and its preparation prior to growth is given. Next, we examine the optimization of the growth conditions of GaN(l-lOO) epilayers. For this purpose, the adsorption and desorption

Nonpolar GaN Heterostructures

163

kinetics of Ga on GaN(l-lOO) are studied using in situ reflection highenergy electron diffraction (RHEED). Finally, the growth of nonpolar (In,Ga)N/GaN multiple quantum wells (MQWs) and a detailed investigation of their structural and optical properties is described.

2. Substrate: y-LiA102 In this section, a brief description of D-LiA102 the crystal structure and the preparation for growth of y-LiA102 is presented. Selected properties of LiA102 (LAO) related to the epilayers are pointed out. 2.1. Properties of the Substrate Crystal structure y-LiA102 has a tetragonal structure with lattice constants a=b= 0.51687 nm and c= 0.62679 nm and belongs to the space group P4A2. 2 7 The y-LiAlO2(100) plane exhibits a comparatively small lattice mismatch to GaN(l-lOO) (the M-plane), namely, [0001]GaN||[010]LAO ~ 0.3% and [ll-20] GaN ||[001] LAO ~ 1.7%, while conventional substrates, e.g. A1203, have much larger lattice mismatch to GaN. Most interestingly, and as mentioned in the introduction, [1—100]oriented GaN thus grown on y-LiAlO2(100) can prove advantageous for improving the internal quantum efficiency of light-emitting devices, as it is free of electrostatic fields along the growth direction. In addition, yLiA102 can be selectively etched with respect to the GaN epilayers, allowing the fabrication of on-chip LEDs. Figure 1 schematically displays the top view for y-LiAlO2(100), and SiC(0001) as well for comparison. Previous reports of GaN(0001) on yLiAlO2(100)28 might resilt from the near-hexagonal symmetry on the substrate surface, as indicated by the dark shaded areas in Fig. 1 (a), and are probably due to improper nucleation conditions.29 From the energetic point of view,30 GaN(l-lOO) on LAO(100) is more favorable than GaN(0001) on LAO(100).

Sun, Brandt

164

• [oio]

• p'iooj

Fig. 1. Unit cells and ball-and-stick models for the growth of GaN on y-LiAlO2(100) (a) and SiC(OOOl) (b). The light and dark shaded areas represent nucleation sites for GaN(l-lOO) and GaN(OOOl).

Disadvantage Despite the potential merits of y-LiA102 as substrate for GaN as mentioned above, it also possesses a number of drawbacks which one should bear in mind. For instance, y-LiA102 is inhomogeneously etched by a variety of acids,31 making chemical polishing an intricate task. Worse, y-LiA102 is hydrolytic,32 and care must be taken to limit the exposure to H 2 0 to a brief dip. Furthermore, y-LiA102 is thermally much less stable than, for example, A1203 or SiC, despite its high melting point of about 1700°C.33 Lee et al.34 pointed out that the dissociation of yLiA102 at the typical metal organic chemical vapor deposition (MOCVD)4 growth temperature of ~1040°C might restrict its use and indeed the incongruent evaporation of Li 2 0 occurs at high temperature.35 However, we did not observe any dissociation of this material up to 1300°C in vacuum by quadrupole mass spectroscopy.36 The y-LiA102 substrates used in this study were grown by a modified Czochralski method at the Institut fiir Kristallzuchtung and chemomechanically polished by Crystec. The peak-to-valley (P/V) roughness and RMS roughness of the polished substrates have values of 1.6 nm and 0.12 nm, respectively, across a 3x3 u.m2 area. Although these are excellent values, a damage layer still exists in contrast to state-of-the-

Nonpolar GaN Heterostructures

165

art SiC or A1203 substrates. Further improvement of the surface quality of LiA102 substrate would be highly desirable. Substrate preparation Substrate preparation should deserve the most attention in order to obtain reproducible high-quality epilayer. Usually, a chemical preparation before loading into the growth reactor is performed. In the MOCVD technique, sapphire is commonly treated by a flux of H2 prior to deposit of a low-temperature buffer. For SiC substrates, a dramatic improvement of the surface quality is attained after annealing the wafers to about 1600°C in flowing H2.37 Etched SiC wafers exhibit a well ordered stepped surface. Since LAO is attacked by various acids and even water, the well-established cleaning process for other substrates used for GaN growth is obviously not applicable for this case. In fact, we have not yet found any effective method, such as thermal or chemical etching to improve the surface morphology of as-received LAO. The asreceived substrates were thus only degreased sequentially in trichloroethylene (10 min), acetone (5 min) and methanol (5 min) in an ultrasonic bath. Finally, the substrate was dipped briefly (30-60 s) in deionized water. The LAO substrate was then fixed with In onto a Si wafer clipped to a Mo holder. Prior to growth, the substrate was outgassed in the load-lock chamber for 1 hour at 200-300°C. During mounting and loading, the substrate surface is again exposed to air and thus recontaminated. Although the outgassing procedure can remove a part of these contaminants, especially water, it would be desirable to have a method for further cleaning substrate inside the MBE system. For SiC(OOOl), Kaplan proposed Ga adsorption and desorption for removing O related contaminants,38 which was later confirmed by Strite et al..39 The procedure is to deposit Ga on the surface where it reacts with contaminants to form a volatile compound, e. g., Ga 2 0. An annealing step up to 750°C can desorb these products and excess Ga remaining on the surface. Analogously, such a Ga treatment was adopted for LAO substrates. However, neither the in situ RHEED pattern after this polishing procedure nor the ex situ characterization of epifilms shows any improvement.

166

Sun, Brandt

50 urn * •

s»

^

i .,j:\. -**.•.

•

•

*

—

» •*•

•

#

» *•"

-

'

:

•

J» Ja

S&U.-' Fig. 2. Interference contrast micrographs of the surface of side A (a) and side B (b) of LAO(IOO) exposed to HC1(32%) at 60°C for 30 min.

2.2. Polarity of the Substrate Considering the polar nature of LAO(IOO), we first investigate the influence of its polarity on the growth of M-plane GaN. Chemical etching While we cannot yet provide a microscopic identification of the two polarities, it is at this stage perhaps more important to be able to phenomenologically distinguish them by simple chemical etching. Figure 2 shows the surface morphology of either face after a 30 min exposure to HC1(32%) at 60°C, recorded by a Zeiss interference contrast microscope. The most obvious feature in both these micrographs are rhomboidal etch pits which we believe to occur at locations where dislocations in LAO intersect the surface. The elongated side of the etch pits is along [010] direction. Indeed, the distinct linear arrangement of these etch pits was attributed to subgrain boundaries formed by dislocations .40 More important, however, is the fact that in between these etch pits the surface of face A remains relatively smooth, whereas face B becomes clearly corrugated. Note, however, that a longer treatment or higher temperatures tend to obscure this difference due to the resulting excessive roughening of both faces.

167

Nonpolar GaN Heterostructures

E_l

32.0

i

I

32.5

•

I

33.0

,

I

33.5

,

I

34.0

i

I

34.5

29 (cleg) Fig. 3. co-29 triple-crystal scans of two samples grown on two opposite sides but under the same conditions.

Epilayer The influence of polarity of the substrate on the epi-film are checked by growth on two opposite sides of LAO(IOO) substrate. Figure 3 shows triple-crystal co-20 scans for two films grown on each face under the same conditions. In both profiles, the peaks located at 29 = 34.682° and -32.254° are due to the LAO(200) and GaN(l-lOO) reflections, respectively. The film grown on face B exhibits an additional weak peak in the left vicinity of the LAO(200) diffraction peak, namely, at 34.5°. Although this angular position coincides with those of either wurtzite GaN(0002) or cubic GaN(l 11), the RHEED pattern of this sample shows additional reflections stemming from GaN(OOOl). However, no such peak is detected on the film grown on face A, indicating a high phase purity of this film. Note that M-plane GaN has indeed the same crystal structure as C-plane GaN, while in this chapter, we call them different phase. In fact, pure M-plane GaN epilayers can be reproducibly synthesized on face A if the nucleation conditions are appropriately chosen (see section III). Because of this finding, we will concentrate in the following on layers grown on face A.

168

Sun, Brandt

Fig. 4. SAD pattern along [0001]GaN or [010]LAO (a) and [ll-20] G a N or [001]LAO(b). a and y denote the wurtzite GaN film and the y-LiA102 substrate, respectively. The unit meshes of the epilayer and the substrate are indicated by the solid and dash lines, respectively.

2.3. In-plane Orientation Relationship As shown earlier, the high-resolution co-20 scan can indicate the outof-plane orientation relationship between the LAO substrate and GaN epilayer. To confirm the expected in-plane azimuthal correlativity as mentioned at beginning of this chapter, selected area electron diffraction (SAD) technique, for example, is presented here. Figure 4 (a) and (b) show SAD patterns along GaN [0001] and [11-20] direction, respectively. Obviously, the GaN film grows epitaxially on the substrate with the expected orientation relationship, i.e., [0001]GaN||[010]LAo and [ll-20]GaN||[001]LAo- The coincidence of the diffraction spots from the epilayer and substrate in both figures evidences the small lattice mismatch along these two azimuths. In addition, the elongated GaN spots in Fig. 4 (b) along the [0001] direction reveal that there are high-density planar defects lying in the basal plane of GaN, which are indeed demonstrated to be stacking faults.41

3. Optimization of Heteroepitaxy of GaN Films on y-LiAIO2(100) Although there are a few reports on nonpolar group-Ill nitride heterostructures,13-23 a systematic knowledge of optimizing the growth of them is still lacking. This section shows a study of the impact of

Nonpolar GaN Heterostructures

169

nucleation conditions on the growth of M-plane GaN( 1-100) films with high phase purity. The optimum growth condition is then empirically established, and high quality M-plane GaN can be reproducibly obtained. 3.1. Impact of Nucleation Conditions on Phase Purity Nucleation is generally believed to be the most critical step for obtaining high-quality C-plane GaN layers.42"44 This is thought to be also applicable for the growth of M-plane films. Growth Growth was carried out in a custom-designed, solid-source three-chamber MBE system equipped with an if plasma source. The N source is operating at a plasma power of 250-500 W with a N2-flow of 0.3-0.4 seem, yielding a GaN growth rate of 0.25-0.4 um/h. The base pressure in the growth chamber is held below 10" 10 Torr by an ion-getter pump and a Ti-sublimation pump. During growth, the chamber is cryopumped and has a pressure around 1.5 xlO 5 Torr for a typical N2 flow of 0.3 seem. The growth front is monitored in situ by RHEED using a 20 kV electron gun. The diffraction pattern is recorded using a CCD camera. RHEED First of all, it should be stressed that MBE offers an excellent tool for determining the phase purity of the layers in real time, namely, by RHEED. During nucleation, a spotty pattern appears which already allows a judgment of the success of the nucleation procedure, since the symmetries of M- and C-plane GaN are distinctly different. Further growth at Ga-stable conditions yields a streaky RHEED pattern, reflecting the progressive smoothening of the surface. Phase mixture manifests itself in the occurrence of additional reflections (typically spots), which are readily distinguishable from the "clean" M-plane pattern. In addition, RHEED patterns also reveal the in-plane orientation relationship between epilayer and substrate. XRD Triple-crystal high-resolution x-ray diffraction (HRXRD) provides an ex situ nondestructive tool to check the relationship of the

170

Sun, Brandt

corresponding out-off-plane orientation between the epilayer and the substrate, and simultaneously the phase purity, at least at the angular resolution and detection limit of the diffractometer. Hereafter, the samples' phase purity and their orientation are checked by XRD. Impact of nucleation conditions on phase purity We have demonstrated that specific nucleation conditions are required for the growth of M-plane GaN.29'36 The results of our investigations are summarized in Tab. I, including the respective nucleation condition and phase purity obtained. TABLE I. Nucleation conditions for the samples investigated. The table provides information whether nitridation was employed (yes V/no —), as well as about the stoichiometry (N-rich/Al- or Ga-rich) and the nucleation temperature. We furthermore indicate the phase purity of the layers (M: M-plane, C: C-plane) as well as the rms roughness obtained by AFM over 5x5 (im2. All films listed have a thickness of 500-600 nm.

Sample A1N buffer Nitridation Stoichiometry TN (°C) Phases rms (nm) N Al N Ga N Ga Ga Ga Ga

580 580 660 660 660 740 660 580 510

C

c M/C M/C C/M M M M M

80 15 38 23 7.5 0.8 3.3

The nucleation regime can be divided into three groups, namely, insertion of an A1N buffer, nitridation prior to growth, and direct deposition. The growth of the A1N buffer was initiated at 580°C under different stoichiometry. Then, the growth of GaN was performed under Ga-stable conditions at high temperature. Nitridation was carried out by exposing the substrate to the N plasma at 350 W and 660°C for 40 min. The subsequent nucleation was executed under either N-rich or Ga-rich conditions. The initial stage of direct deposition of GaN (nucleation) was performed at various temperatures and under different stoichiometry. In all cases, the nucleation temperature TN can be divided into four regimes:

Nonpolar GaN Heterostructures

171

high (740°C, actually the growth temperature TG), intermediate (660°C), low (580°C) and very low (510°C). The thickness of the A1N buffer or the GaN nucleation layers at very low, low and intermediate 77V was about 100 nm. For all samples, the growth temperature TG of 740°C was kept constant during the subsequent growth and the stoichiometry was set to Ga-stable. During the initial deposition regardless of nucleation conditions, a spotty RHEED pattern first appears, while further growth at Ga-stable conditions yields a streaky RHEED pattern29 reflecting the progressive smoothening of the surface. RHEED allows to judge the orientation of the growth front, since the symmetries of M- and C-plane GaN are distinctly different. Several conclusions can be drawn from this table. (i) It is occasionally reported that the use of an A1N nucleation layer has a positive impact on the properties of C-plane GaN grown on it.45,46 We thus investigated if an A1N nuclation layer would also benefit the growth of M-plane films. As seen from Tab. I, growth under either Alrich or N-rich conditions however leads to a predominant C-plane structure. To investigate the origin of this effect, 200 nm thick A1N layers deposited at the same temperature as the buffer used in sample A and B were grown and measured by XRD. It is found that the A1N layers are [0001]-oriented, while no peak pertinent to the M-plane orientation is detected.36 A1N nucleation layers (at least under our nucleation conditions) are thus clearly detrimental for the growth of M-plane GaN layers. (ii) Considering that nitridation of A1203 substrates prior to nucleation is well established to improve the crystal quality of the subsequent GaN film by the formation of an A1N template,47-50 we have studied the nitridation of LAO as a potential means to improve epitaxial growth also in this case. Apparently, nitridation of LAO prior to growth invariably induces phase mixture of C- and M-plane in our case, as seen from Tab. I. Interestingly, it was reported that nitridation of LAO is necessary for achieving the M-plane GaN by MOCVD.25 (iii) Though N-rich growth has been established to be disadvantageous for the PAMBE growth of C-plane GaN epilayers on

172

Sun, Brandt

either A1203,51 SiC,52 or GaN,53 particularly concerning the surface morphology, it is not clear whether N-rich nucleation is unfavorable for M-plane films. As demonstrated, the immediate N-rich nucleation condition indeed leads to predominant C-plane GaN.36 (iv) Clearly, the nucleation conditions are crucial to the orientation relationship of GaN on LAO(IOO). Our study shows that only immediate Ga-rich nucleation results in the growth of pure A/-plane GaN.29'36 3.2. Influence of Nucleation Temperature on Surface Quality With the knowledge of how to obtain pure M-plane GaN epilayers, the surface morphologies of M-plane GaN films with different nucleation temperatures are studied by AFM (cf. Ref. [36]) and TEM.41

IU

E c

C/3 C/3

:

'

•

•

^

LU „ „ i

^

Z 10'

• ': ,*

'"N^

IE

^' , -"

• .

-

-

•

*

"

•

_,-*

(3

=> O

.

o

0-

0 •

500

550

600

650

I CD

Z)

DC

5:10°

co

o -. 10 HI -z. •

o

10'

1

1

700

o ' .• 10" tr w : 750

or

TN(°C)

Fig. 5. P/V (solid symbols) and rms (open symbols) roughness over an area of 5x5 (am2 as a function of the nucleation temperature. The thicknesses of all samples are 500-600

Roughness vs. TN The P/V and RMS roughness of the pure M-plane GaN are plotted vs. the nucleation temperature in Fig. 5. Obviously, the impact of TN on the surface roughness is dramatic. Films grown at low nucleation temperature (580°C) are significantly smoother than those grown at higher nucleation temperature. The nucleation temperature at 580°C is apparently the optimum one. In addition, a further reduction of the nucleation temperature to 510°C does not result in a further reduction

Nonpolar GaN Heterostructures

173

of the surface roughness, on the contrary, the films are significantly rougher than those nucleated at 580°C. Chen et al.11 reported an rms roughness of 0.7 nm over 2x2 urn2 for a 4-p.m-thick homoepitaxial Mplane GaN layer grown by MOCVD. This compares to 0.6 nm for the best samples in Fig. 5 over the same area, demonstrating a virtually identical degree of roughness. This degree of roughness is close to that obtained on C-plane films and is well suited for the growth of heterostructures, such as quantum wells. ' TEM To examine the microstructure of the layers, we employ cross sectional TEM. TEM studies are performed in a JEOL3010 microscope operating at 300 kV. Figure 6 shows a two-beam bright field cross sectional micrograph of a typical M-plane GaN epilayer with g=l-100 near the [11-20] zone axis. The microstructure of the M-plane GaN is quite different from that of C-plane GaN. The density of perfect threading dislocations is found to be below lxlO 6 cm" 2. The dominant defects are basal plane stacking faults (SFs), most of which intersect the

Fig. 6. Cross-sectional g=l-100 two-beam bright-field TEM micrograph near the [11-20] zone axis of a GaN film grown on a LAO substrate.

entire thickness of the layer. Their density amounts to lxlO 5 cm -1 . The partial dislocations terminating these basal plane SFs have been determined to be of Shockley type with Burgers vectors b=l/3[10-10] or 1/3[01—10].56 A recent work suggested that coalescence of two nucleation regions with different stacking sequences is the reason for the

174

Sun, Brandt

formation of stacking faults, as three <1-100> faces have ABAB stacking and the other three have ACAC stacking. Nevertheless, it is clear that a low nucleation temperature results in a significantly improved microstructure and surface morphology.41 4. Ga Adsorption and Desorption Kinetics In the case of GaN(OOOl), several studies have established the optimum growth conditions in MBE and related them to the existence of a Ga bilayer on the growth front.58-62 The lateral mobility of the Ga atoms within the bilayer is predicted to be high, consistent with the high morphological quality of GaN(OOOl) grown under conditions giving rise to bilayer coverage. Empirically, we have found that Ga-rich conditions results in GaN( 1-100) layers with comparatively smooth surface morphology, whereas N-rich conditions lead to a roughening of the growth front. Lee et a/.18 observed a reconstruction of approximate symmetry "(4x5)" on Ga-rich grown GaN(l-lOO) on ZnO by scanning tunneling microscopy. The surface was found to be metallic in nature, and the authors proposed that it consists of >2 monolayer (ML) of Ga based on the fact that Ga-adlayer structures with coverages between 2 and 3 ML on GaN( 1-100) are energetically favorable at Ga-rich conditions. In this section, we present a study on the adsorption and desorption kinetics of Ga on GaN( 1-100) using RHEED.

4.1. Surface Reconstruction In the course of our previous work,29 we noticed that all smooth layers exhibited a complex RHEED pattern when the surface was quenched to about 150°C with a cooling rate of 40°C/min and the N flux shut off immediately at the termination of growth. The RHEED pattern observed exhibits a fourfold periodicity along all major azimuths, consistent with a (4x4) reconstruction with respect to the unit mesh of the GaN( 1-100) surface. Note that this reconstruction is not identical to the pseudo-(4x5) reported by Lee et al.,ls which gave rise to a fourfold

Nonpolar GaN Heterostructures

175

pattern along the [0001] azimuth only. In the following, we establish a relation between the adsorbed Ga coverage and reconstructions of the surface.

Fig. 7. RHEED patterns of the GaN(l-lOO) film, recorded at 100°C after a 2-step deposition of 2 ML Ga at 580°C. Note the intense half-order streaks along all azimuths.

Relation between Ga coverage and surface reconstructions The growth front was monitored in situ by RHEED at 20 kV energy and an incidence angle of 1.5°. The diffraction patterns along all major azimuths of GaN(l-lOO), i. e., [11-20] (0°), [22-43] (~39°), [11-23] (-58°), and [0001] (90°), were recorded using a CCD camera. The intensity of the specular spot and other regions of interest are recorded using a digital RHEED analysis system,63 which was set to sample data at a rate of 25 Hz. Growth was initiated by nucleation at 580°C under Ga-rich conditions. The growth temperature was then raised to 740°C and kept constant during the subsequent adsorption/desorption studies and growth. All of these studies have been done using a 500 nm thick GaN( 1-100) layer grown under standard conditions, i. e., Ga stable growth ((j)Ga= 0.475 ML/s, §N = 0.275 ML/s) and a substrate temperature of 740°C. During growth of M-plane GaN under standard conditions, RHEED shows the characteristic ( l x l ) pattern.29 The pattern is entirely streaky, but dim, very similar to the RHEED pattern observed during growth of C-plane GaN. Upon interrupting the growth for about 30 s at constant

176

Sun, Brandt

temperature and the N flux on, the intensity of the pattern increases greatly, while the ( l x l ) symmetry does not change. The RHEED pattern of the recovered surface does not exhibit any appreciable change when switching the N flux off. It does not change either when subsequently cooling the sample down to 100°C. Upon deposition of up to 4 ML of Ga on this surface at 100°C, the only noticeable change is that the RHEED pattern becomes dim similar to that observed during growth. This finding is in marked contrast to the observation of a (4x4) reconstruction for the surface quenched to low temperature immediately after growth. Apparently, the thermal energy at 100°C is insufficient for the Ga adatoms to condense into this reconstruction. To study the formation of the (4x4) reconstruction further with thissample, the excess Ga deposited at 100°C is first flashed off by heating the sample to 740°C. The sample is next cooled down to 580°C, at which the desorption of Ga is negligible. Next, we deposit 1 ML of Ga onto this surface. Upon cooling to 100°C, clear half-order streaks are observed along all azimuths except for [11-20], consistent with a (1x2) reconstruction of GaN( 1-100). Upon heating, the half-order streaks abruptly disappear at a temperature of 250°C, but the pattern stays dim in comparison to that of the recovered surface. An additional ML of Ga is then deposited at 580°C, which gives rise to a more intense (1x2) reconstruction after cooling to 100°C as shown in Fig. 7. However, we often observe the formation of 1/4-order streaks [i. e., next to the (00) streak] along the [11-20] azimuth at this coverage, which would indicate a (4x2) reconstruction. This observation might indicate a chainlike ordering of (lx2)-reconstructed domains. The sample is then heated to 580°C again. Finally, the procedure is repeated for the deposition of an additional ML of Ga, resulting in a total of 3 ML of Ga on the surface. Figure 8 shows the resulting RHEED pattern at 100°C, which exhibits an intense (4x4) reconstruction identical to that observed when quenching the sample after growth at standard growth conditions. We have checked that the deposition of 2 and 3 ML of Ga in one step at 580°C results in RHEED patterns which are indistinguishable from those displayed in Figs. 7 and 8, respectively. Deposition of intermediate amounts, such as

Nonpolar GaN Heterostructures

111

2.5 ML, often results in a superposition of both reconstructions, in that a four-fold periodicity is observed along [11-20], but a two-fold along all other azimuths. It is presumed due to the form of domains. Deposition of more than 3 ML reduce the overall intensity of the pattern, but does not change its symmetry. In all cases, the quarter-order streaks get diffuse and eventually disappear at temperatures above 250°C.

Fig. 8. RHEED patterns of the GaN(l-lOO) film, recorded at 100°C after a 3-step deposition of 3 ML Ga at 580°C. Note the intense quarter-order streaks along all azimuths.

It is clear from these experiments that two stable Ga coverages exist, which condense at low temperatures into a (Ix2)/(4x2) reconstruction at bilayer coverage and a (4x4) reconstruction at trilayer coverage. On surfaces with intermediate amounts of adsorbed Ga, both reconstructions may coexist (presumably in the form of domains), although we have no experimental evidence for such as coexistence. Furthermore, the recovery behavior of the surface at elevated temperatures provides the opportunity to study the adsorption/desorption kinetics of Ga in real time. 4.2. Ga Adsorption/Desorption Kinetics First, Ga adsorption and desorption kinetics in vacuum is studied. The sample was held at a constant temperature of 740°C. Ga

178

Sun, Brandt

adsorption/desorption isotherms were recorded by monitoring the intensity of the specularly reflected beam along the [22-43] azimuth, starting with 60 s of adsorption and allowing the surface to recover completely prior to the next experiment. • + • 0.056 ML/s

I

T s = 740 °C

- 0.073 ML/s

N == 0 M L / s

• 0.093 MUs

0.12 ML/s 0.15 ML/S

>H

is.

0.19 ML/S 0.24 MUs 0.30 ML/S

LU I-

0.38 ML/S

Q LU N

0.475 ML/S 0.59 ML/S 0.73 ML/S

rr O

0.90 ML/S 11 ML/S

100

200 300 400 T I M E (S)

500

Fig. 9. Variation of the specular RHEED intensity upon adsorption of Ga for 60 s and subsequent desorption for Ga fluxes as indicated. The arrow at t = 10 s indicates the opening, the dashed line t = 70 s the closing of the Ga shutter.

Figure 9 shows the variation of the specular RHEED intensity Is for Ga fluxes between 0.056 and 1.11 ML/s. For the two lowest fluxes, no appreciable change in the RHEED intensity is observed, reflecting that Ga desorption at this temperature inhibits the adsorption of significant amounts of Ga on the surface. At higher fluxes (0.12 ML/s), the RHEED intensity exhibits a clear drop once Ga deposition is initiated, and an immediate recovery once the Ga supply is stopped. For
Nonpolar GaN Heterostructures

179

recovers without (0.15 ML/s) or with a very short (0.19 ML/s) delay. Apparently, this situation corresponds to a balance between impinging and desorbing Ga flux, thus establishing an equilibrium Ga coverage at the surface independent of deposition time. For a Ga flux higher than this value, the recovery is progressively delayed. Furthermore, the RHEED intensity continues to decrease after the initial, rapid drop, albeit at a much slower rate.

t' 00

Z

LLI

<

o -z. "

0

50

100

150

200

250

300

350

TIME (S) Fig. 10. The four stages of Ga adsorption and desorption for the example of ()>Ga = 0.73 ML/s.

We interpret these observations according to the following four-stage scenario as depicted in Fig. 10. Initially (stage I), Ga adsorbs in a layerby-layer fashion, resulting in the initial, rapid drop of the RHEED intensity until full coverage is established. If the impinging Ga flux exceeds the desorbing flux, Ga accumulation takes place, eventually leading to the formation of droplets which manifests itself in a further but slower decrease of the RHEED intensity (stage II). Once the Ga supply is stopped, the Ga adlayer starts to desorb. Excess Ga, however, acts as reservoir, thus stabilizing the Ga adlayer and delaying recovery (stage III). Once the excess Ga is depleted, the Ga adlayer desorbs, accounting for the rapid final recovery (stage IV). An approximate analysis of these transients indicate an adsorbed amount of Ga at the end of stage I slightly below 2 ML for 0/15.64

180

Sun, Brandt

Next, Ga adsorption and desorption in the presence of an active N flux is investigated. The sample was held at a constant temperature of 740°C and exposed to an active N-flux of 0.275 ML/s. Ga adsorption /desorption isotherms were recorded exactly as described earlier. Ga fluxes range from 0.19 to 1.66 ML/s. The behavior observed is analogous to that shown in Fig. 9, and we thus interpret it in the same way. Quantitatively, the adsorbed amount of Ga at the end of stage I was found to be 3 ML, i. e., higher than for adsorption in vacuum.64 To confirm the four-stage scenario mentioned earlier, we develop a quantitative model. The model is required to account for adsorption and desorption of Ga as well as for the formation of excess Ga interacting with the Ga adlayer, and for GaN growth in the case of the simultaneous presence of Ga and N. Our model, in units normalized to the maximum Ga adlayer coverage, is described by the equationswhere 9 denotes the Ga adlayer coverage, n the amount of excess Ga, jga and jN the Ga and N delivery rate, dGa the diffusion rate of excess Ga adatoms impinging onto 0, and jg( y„) the rates of Ga adlayer (excess Ga) desorption. dd_ jGa(1 - 6 ) + dGa{\-d)ndt dn JGa0-dGa(}-e)n-Ynn dt

yed-

jNi6

(1) (2)

The first term of each equation accounts for adsorption of Ga, building up the Ga adlayer #and subsequently excess Ga n. In much the same way, the second terms account for diffusion of excess Ga impinging onto the Ga adlayer. The third terms describe the desorption of the 6 and n. The last term in Eqs. (1), finally, stands for the incorporation of Ga and N adatoms into the crystal, i. e., actual growth. Note that the model described by Eqs. (1) and (2) is a linearized, and highly simplified version of microscopic theories of the surface kinetics such as developed in Ref. 65. This simplification is necessary to reduce the number of free parameters and thus to guarantee stable fits, but also means that there certainly are a variety of more refined models which

Nonpolar GaN Heterostructures

181

could equally well describe the data. Finally, we have to relate the surface coverage calculated by means of Eqs. (1) and (2) to the experimentally observed RHEED intensity. Since the Ga adlayer at growth temperature does not condense into a reconstruction, but rather represents a liquid, it serves merely to dampen the RHEED intensity. We thus follow the well-established relation between surface coverage and intensity in Auger attenuation experiments,66 which reads: I{6,n) = e~e-bn (3) Both the adlayer coverage q and the excess coverage n are assumed to damp the intensity exponentially. The ripening of excess Ga to clusters and droplets leads to a different rate of decrease with respect to the Ga adlayer coverage, as expressed by the prefactor b.

Fig. 11. Experimental (symbols) and simulated (solid lines) RHEED transients for three different Ga fluxes as indicated in the figure (a) in vacuum, and (b) under an active N flux of 0.275 ML/s (the transients are vertically offset for clarity).

Figure 11 (a) shows three experimental Ga adsorption/desorption isotherms in vacuum together with simulations based on Eqs. (l)-(3). Except for the Ga flux, which was taken to be equal to the experimentally determined flux as indicated in the figure, we used the same parameters for these simulations, namely: a diffusion coefficient of 2xl0" 13 cm2/s, a Ga adlayer desorption flux of 0.137 ML/s, an excess Ga desorption flux of 0.003 ML/s, and a prefactor b of 0.06±0.003. Evidently, the agreement between the experimental transients and the

182

Sun, Brandt

simulations is very good. Most importantly, this agreement can only be obtained with the above set of parameters, since the slope of the transient in stage III is mainly determined by yn, while ye determines the slope of the transient in stage IV. Furthermore, the onset of stage IV depends essentially on the sum of y„ and yg, resulting in unique fits to the experimental data. This result reflects that our general understanding of the adsorption/desorption processes is essentially correct. The coexistence of 9 (1.64 ML) and n (0.42 ML) deduced from the simulations at the end of stage I is a direct result of the finite diffusion rate of the excess Ga. If the Ga adlayer is not completed (9 < 1), it will not form a continuous film but two-dimensional islands. This inevitable fact has several important consequences. First of all, our model treats 9 as a continuous variable, and neglects its inherently discrete nature at low coverage. Second, the desorption kinetics of two-dimensional islands is known to be dominated by desorption from the islands' edges, which changes the first-order (monomolecular) desorption kinetics assumed here to half-order. Third, the desorption rate ye will be enhanced in this case, essentially resulting in a coverage-dependent desorption rate which is no longer a constant but a function of G. In fact, for Ga fluxes lower than 0.19 ML/s, we had to increase ye significantly to account for the early and comparatively rapid recovery observed. In view of that said above, it is clear that this simple measure is purely phenomenological and only approximates the actual processes. We next turn to a comparison between experimental Ga adsorption/desorption isotherms under active N flux and simulations based on Eqs. (l)-(3). As above, we take the experimentally determined Ga flux for the simulations, and, of course, the additional N flux. For a Ga flux larger than 0.475 ML/s, the parameters required to get a good agreement between simulated and experimental transient are identical to those given above, except for yg, which turns out to be slightly larger (0.155 ML/s). For lower Ga fluxes, the Ga adlayer is not completed and, increases drastically as discussed above for the case of desorption in vacuum. Figure 11 (b) shows the three RHEED transients at the

Nonpolar GaN Heterostructures

183

transition between N-stable and Ga-stable conditions. The desorption fluxes required to fit them are 0.155, 0.21, and 0.28 ML/s from low to high Ga flux. The good agreement between experimental and simulated transients shows that the adsorption/desorption kinetics is not fundamentally different in the presence of a N flux compared to that under vacuum. Rather, the N flux seems merely to consume Ga and thus to shift the adsorption/desorption balance toward higher Ga fluxes. For sake of completeness, we determine again the Ga adlayer and excess Ga coverage predicted by the simulation at the end of stage I. This coverage is found to be independent of Ga flux for <j)Cai 0:475 ML/s, and amount to d- 1:94 ML and n = 0:88 ML, respectively. Note that these values are in fact significantly larger than those obtained for adsorption in vacuum. At this point, it is instructive to examine the predictions of our model for prolonged growth using the parameters used for the simulations in Fig. 11(b). '' *N

= 0.275 ML/s

f

52 t.s

\ S

0.475 ML/s

f

0.38 ML/s

r^—n 1 (Gaoff 0.59 ML/s

4N

= 0.275 ML/S

Gaoff

LU

O D?10 1 LU

oc

LU >1.6 O

/ - " " " 0 . 5 9 ML/S

>

O,o°

CC 1 4 LU

<

3 1-10 2

•""

0.475 ML/S

LU10" 1 I

I

1 50

,

!_,./_. , 100 600 700

TIME (s)

(a)

EXC

2.0 .

0.38 ML/S 0

(b)

100 200 300 400 500 600 700 800

TIME(S)

Fig. 12. Simulated Ga adlayer and excess Ga coverage during growth. The parameters used are those deduced from the simulations shown in Fig. 11, but the Ga supply is now terminated at t = 700 s.

Figure 12 (a) shows the Ga bilayer coverage vs. growth time for the three different Ga fluxes at the transition between N-stable and Ga-stable conditions. As expected, full bilayer coverage is not reached for the lowest Ga flux. In contrast, the Ga adlayer coverage slowly converges towards a full bilayer for the intermediate Ga flux, and is obtained rapidly for the highest Ga flux. However, as shown in Fig. 12 (b), the price one pays for this rapid bilayer completion is a diverging Ga excess

184

Sun, Brandt

coverage, which increases without bound for prolonged growth. In fact, layers grown under these conditions exhibit a high density of Ga droplets. In contrast, the Ga excess coverage saturates at 2 ML for the intermediate Ga flux, reflecting that a dynamic equilibrium is established under these conditions. Note that the desorption flux for this case is 0.21 ML/s, which in fact is almost exactly equal to the difference of Ga and N flux. The total Ga coverage during growth is here thus 4 ML, which is consistent with the observation of the Ga-trilayer related (4x4)reconstruction observed for this growth conditions after quenching the sample to 100°C (cf. Fig. 8). For the lowest Ga flux, the excess coverage is (as expected) negligible.

2 pm

Fig. 13. AFM micrograph of a 500 nm-thick GaN(l-lOO) layer grown under standard conditions. The vertical stripes are oriented along [11-20]. The P/V and rms roughness of this layer are 6.6 and 0.6 nm over an area of 5x5 u,m2, respectively.

Finally, Fig. 13 shows an AFM micrograph of a GaN(l-lOO) layer grown under our standard conditions, which has been deemed previously to be most suitable on empirical grounds.29 In particular, these conditions result in droplet-free and microscopically smooth surfaces. The surface of the layer exhibits the characteristic highly anisotropic, stripe-like morphology of GaN( 1-100). The orientation of these stripes, investigated by Raman scattering, is found to be along , i. e., perpendicular to the c-axis. The origin of this surface might be a

Nonpolar GaN Heterostructures

185

consequence of growth kinetics. The migration length of adatoms most likely differs along [11-20] and [0001], which have distinct step structures. An alternative explanation of this peculiar morphology are the stacking faults intersecting the layers, which lie on basal planes and thus have the same orientation as the stripes on the surface. 5. M-plane (In,Ga)N/GaN Multiple Quantum Wells Nonpolar wurtzite Ill-nitride MQWs are expected to be free of electrostatic fields along the growth direction. To experimentally confirm this presumption, one usually relies on the results of spectroscopic studies. The correct interpretation of these results, however, requires an accurate knowledge of the structural parameters of the samples under investigation. This statement is particularly true for the case of (In,Ga)N/GaN quantum wells, in which the surface segregation of In may severely distort the intended composition profile as indeed found for the case of C-plane structures.67"69 In this section, growth of M-plane (In,Ga)N/GaN MQWs on LAO(IOO) and a detailed investigation of their structural and optical properties are described.

28 (cleg)

26 (deg)

Fig. 14. Experimental (•) and simulated (—) triple-axis co-29 scans across the (a) GaN(l-lOO) and (b) GaN(2-200) reflection of a 20-period M-plane (In,Ga)N/GaN MQW grown on LiAlO2(100) (sample MQW-A). The superlattice satellites are denoted by SL±n.

186

Sun, Brandt

5.1. In Incorporation and Surface Segregation To study the impact of surface orientation on the In incorporation during the MQWs growth, a set of 20-period (In,Ga)N/GaN MQWs was grown on y-LiAlO2(100) substrates. Growth For all samples, growth was initiated by a 100 nm thick low temperature (580°Q GaN nucleation layer, after which the temperature was increased to 740 °C for the growth of a 500 nm thick GaN buffer layer, which has an rms roughness of only 0.6 nm as shown in section III. For the subsequent growth of the (In,Ga)N/GaN MQWs, the substrate temperature was decreased to 600°C. In, Ga, and N fluxes were identical for all MQWs, which differ only due to the different growth times for the well, namely, 25 s for samples A and B, and 40 s for samples C and D, respectively. The RHEED patterns stayed streaky during the entire growth of the MQWs. In fact, the rms roughness of the samples as determined by AFM is between 1.2 and 1.7 nm. The samples are further characterized by HRXRD and secondary-ion-massspectroscopy (SIMS). HRXRD To determine the structural parameters of these M-plane MQWs by HRDXD, we follow the same procedure as employed in Refs. [70]. Symmetric x-ray co-29 scans were taken with a Philips X'Pert PRO triple-axis diffractometer with a Cu K OCi source, a Ge(220) hybrid monochromator and a Ge(220) three-bounce analyzer crystal. We first analyze the angular positions of the satellite peaks kinematically based on the nominal growth times to obtain the structural parameters of the sample, implicitly assuming that segregation does not occur. Next, we employ simulations based on the dynamical diffraction theory70 and vary the kinematically obtained parameters until the intensity of the satellites matches the experiment in addition to their position. The simulations are done for perfect interfaces and are convoluted with the instrumental resolution (25") only. As an example, Fig. 14 shows co-29 scans for sample MQW-A across the (a) GaN(l-lOO) and (b) GaN(2-200)

Nonpolar GaN Heterostructures

187

reflection. Several important conclusions can be drawn from these scans: (i) No additional signal is detected at the position of either GaN(0002) or in between GaN(0002) and InGaN(0002), demonstrating that the entire structure has the expected orientational relationship, i. e., [l-100](in.Ga)N|| [100]LAO- Additional resonant Raman measurements (not shown here) demonstrate that the structure also fulfills the expected in-plane orientation relationship, namely, [0001](In;Ga)N||[010]LAO and [ll-20](In;Ga)N||[001]LAo. (ii) The angular separation between the LAO(200) and GaN(l-lOO) reflections or LAO(400) and GaN(2-200) indicates a lattice dilation of 0.6% along the growth direction, i. e., [1-100], for the GaN buffer. This fairly high strain is at least partly a result of the coherent growth of GaN( 1-100) on LAO(100) along the [0001] direction,30 where the lattice mismatch amounts to only 0.3%. The mismatch in thermal expansion, of course, is also compressive and enhances this effect, (iii) Superlattice satellites up to the 2nd order are observed for all samples, demonstrating the high periodicity of the structures and allowing us to determine their structural properties. In fact, the agreement between the experimental and simulated profiles is quite satisfactory, whereas simulations with the kinematically obtained parameters (not shown here) exhibit satellite intensity ratios clearly inconsistent with those experimentally observed. The results of this analysis are summarized in Tab. II, and indicate a very significant In surface segregation for all samples under investigation. Interestingly, the In content obtained from this analysis is -7%, while C-plane samples grown under identical conditions exhibit an In content of about -15%. The broadening of the satellites in the experiment with respect to the simulation, which is particularly evident in Fig. 14 (b), gives rise to concern about the validity of our approach. Indeed, if this broadening would be caused by interface roughness, the intensity ratio between the satellites would be changed and could not be described correctly by the present model. However, the widths of the 0th and the -\th satellites are virtually identical for all samples, reflecting that the broadening arises from heterogeneous strain rather than from interface roughness. A Williamson-Hall plot71 of the width of the SL0 reflection results in an

188

Sun, Brandt

estimated heterogeneous strain along the growth direction of 1.2xl0~3, larger than that obtained typically for bare GaN layers. This result is not unexpected, as compositional fluctuations in the ternary (In,Ga)N alloy, which are believed to be largely driven by In surface segregation,72 will inevitably result in a certain degree of inhomogeneous strain. SIMS In order to independently confirm the actual In content and structural parameters, we performed high-resolution SIMS measurements. SIMS was carried out with a CAMECA IMS 4f-E6 system, employing Cs+ primary ions with an impact energy of 5.5 keV and InCs+ ions as species for detection. The depth resolution under these conditions is about 1 nm/decade and 2 nm/decade at the leading and trailing edge, respectively. TABLE II. Parameters of the M-plane (In,Ga)N/GaN MQWs under investigation as determined by the kinematical analyses based on growth time (kin), the dynamical simulations taking the real depth profile of In into account (dyn), and SIMS. The In content and the well width are denoted as x and dw (nm), respectively. The lower and upper margin for the well thickness in the SIMS column corresponds to the value obtained directly from the In content and that obtained from the FWHM of the In profile, respectively. dyri

Kin Sample MQW-A MQW-B MQW-C MQW-D

x 0.14 0.16 0.19 0.20

dw 1.35 1.47 2.43 2.48

X

0.068 0.070 0.066 0.068

SIMS

i

dw 2.7 3.3 7.0 7.4

X

0.070 0.071 0.072 0.071

dw 2.7-3.1 3.3-4.3 6.4-7.2 7.1-7.6

As an example, the In depth profile of sample MQW-A is shown in Fig. 15. The leading and trailing edges of this depth profile are 1.4 and 2.2 nm/decade, i. e., close to the instrumental resolution. The progressive damping and broadening of the signal with depth (cf. inset of Fig. 15) is caused by roughening from the sputtering process. The actual well width can be deduced in two ways from this profile. First, we may take the In content as measured by SIMS for granted (note that InCs+ detection is

Nonpolar GaN Heterostructures

189

indeed known to be quite accurate in this regard) and combine this value with the average In content as measured by HRXRD. Second, we may take the FWHM of the In peaks to reflect the actual well width if the instrumental broadening is Gaussian and is significantly smaller than the well width itself (otherwise the FWHM will overestimate the well width). The In contents and well thicknesses thus determined for the samples under investigation are listed in Tab. II. CO

'E ,0, Z10

:

o

£E rr -z. 10: LU

o z o o £10' 0

10

20

30

40

50

DEPTH (nm) Fig. 15. SIMS depth profile of the top 50 nm of sample MQW-A. The depth resolution is indicated. The entire depth profile is shown in the inset.

The good agreement with those deduced from the HRDXD simulations is evident. This result signifies that In surface segregation is present on the M-plane as well as on the C-plane of GaN. The lower In content found for the former even suggests that it is stronger in the present case. In other words, the In incorporation efficiency on the Mplane, phenomenologically lumping together processes such as segregation and desorption, seems to be lower when compared to the Cplane. An interesting point is the symmetrical "top-hat" shape of each QW in Fig. 15, which we observed for all samples under investigation. This observation is analogous to the case of C-plane (In,Ga)N/GaN MQWs68 and indicates a Oth-order In segregation process. This process implies an In incorporation limit—here 7%—at a given growth temperature, and has the consequence that the incorporation of In does not depend on the actual amount of In floating on the growth front.

190

Sun, Brandt

5.2. Recombination Mechanism The spontaneous emission of C-plane (In,Ga)N quantum wells is determined by both e-h wavefunctions separation due to the built-in internal electrostatic field (QCSE) and exciton localization caused by potential fluctuations.73-76 The realization of M-plane (In,Ga)N/GaN MQWs allows us to investigate the impact of excitons localization on radiative recombination without the interference of internal electrostatic fields. To study the recombination mechanism of M-plane (In,Ga)N/GaN MQWs, continuous-wave photoluminescence (cw-PL) spectroscopy and time-resolved (TR) PL were carried out. cw-PL cw-PL spectra at various temperatures were recorded using a 325 nm He-Cd laser with an excitation intensity of 10 W/cm2. Figure 16 (a) shows the evolution of the cw-PL spectra from 5 to 300 K for an Mplane (In,Ga)N/GaN MQW (MQW-A). The dominant emission centered close to 3 eV stems from the quantum wells. Interestingly, we also observe emission near 3.36 eV, originating from excitons bound to stacking faults in GaN41 and at 3.497 eV from the (D°,X) transition in GaN. The energy position of the latter indicates a large compressive strain,77'78 which is consistent with the XRD measurements described above. However, the observation of luminescence from GaN alone is interesting, since we never detected any signal from GaN in the case of C-plane (In,Ga)N/GaN MQWs with a total thickness, as the present sample, exceeding the penetration depth of the laser. The lack of GaN luminescence in this case is plausible: the internal electrostatic fields in the barriers of C-plane (In,Ga)N/GaN MQWs rapidly separate electrons and holes and direct them to the wells, thereby enhancing the capture efficiency. The observation of rather strong luminescence from the GaN barriers in the present sample may thus be taken as a manifestation of the absence of internal electrostatic fields. Figure 16 (b) shows Arrhenius plots of the integrated cw PL intensity of MQW-A and MQW-D. The total PL intensity of these two M-plane MQWs is reduced by a factor of 17 and 44, respectively, when the temperature is increased from 5 to 300

Nonpolar GaN Heterostructures

191

K. In fact, all samples under investigation exhibit a strong and broad PL band between 400 and 420 nm at room temperature. For all of our Mplane MQWs, the luminescence quenches by a factor of 15-50 when the temperature increases from 5 to 300 K. The PL of these MQWs is still detectable up to 500 K. This finding is in contrast to C-plane MQWs of similar well widths as, e. g., MQW-D (~7 nm), where the emission intensity decreases drastically with well width and is hardly detectable at room temperature.

(b)

go B

8

8

Oa

4 •i <J

a 8 9

O •

! —i

2.8 3.0 3.2 3.4 P H O T O N E N E R G Y (eV)

3.6

0

i

i

50

i

i

i

100 1000/T

1

150 (K

MQW-A MQW-D 1

i

200

-1,

Fig. 16. PL spectra of the MQW-A in the temperature range from 5 K to 300 K. All spectra are normalized and vertically offset for clarity (a). Arrhenius representations of the integrated PL intensity of MQW-A and MQW-D (b).

Anomalous temperature-dependent PL width In both C-plane 79-81 (In,Ga)N/GaN MQWs and cubic (In,Ga)N/GaN MQWs,82'83 exciton localization caused by In compositional fluctuation has a crucial influence on the radiative recombination process. A clear indication for the presence of localized states in the M-plane (In,Ga)N/GaN MQWs is obtained by the anomalous temperature dependence of the full-width at half maximum (FWHM) of the PL spectrum, as shown in Fig. 17, for MQW-A. The narrowing of the spectrum upon increasing temperature has been observed in several materials systems in which strong localization occurs, such as InAs/GaAs quantum dots,84 (In,Ga)As/GaAs QWs85 and (In,Ga)N QWs.73,86 At low temperatures, excitons are randomly trapped by the localizing centers, and the PL line-width

192

Sun, Brandt

basically corresponds to the width of the distribution of local potential minima. With increasing temperature, excitons are thermally activated and become mobile within the band of localized states, allowing them to relax further into deeper potential minima. This thermal activation will thus result in the distinct narrowing of the Hnewidth. At even higher temperature, the thermal distribution of excitons will force them to populate also higher-energy states. As a result, the PL band broadens and its width once again approaches the bandwidth. 330 -

0

a

100

200

300

400

''

500

TEMPERATURE (k)

Fig. 17. FWHM of the PL spectrum as a function of temperature for the M-plane MQWA. The dashed line is a guide for the eye.

PL transition energy as a function of well thickness With the accurate structural parameters obtained as described in section VA, a quantitative comparison of the transition energy as a function of the well thickness can be done. Figure 18 shows a comparison between the experimental room-temperature PL peak positions and the calculated transition energies, rigidly shifted down by 110 meV for clarity, versus the well widths. The calculated values of the transition energy are obtained by a self-consistent Schrbdinger-Poisson calculation assuming a constant In content of 7%, a band gap for Ino.07Gao.93N of 3.14 eV,87 a band offset ratio of 70:30,88 effective electron and hole masses of 0.2 and l,88 a nonlinear spontaneous and piezopolarization following Ref. [89], and assuming a uniform background carrier density of lxlO 18 cm-3 which is the typical value for bare GaN layers grown on LiAlQ2. The dashed

Nonpolar GaN Heterostructures

193

curve takes the full polarization field into account, while the solid curve is obtained assuming no polarization. The experimental transition energies are seen to follow closely the prediction for this quasi-flat-band case. It is most important to realize that all of the above parameters, except for the polarization constants, will merely rigidly shift these curves up or down, but will not change their shapes. A redshift, in fact, is expected as our calculations do not include excitonic effects, which may amount to 40-60 meV in the present structures,90 and exciton localization. These results thus provide evidence for the fact that the structures under investigation are free from internal electrostatic fields as encountered in corresponding C-plane structures. _

>

3.3

~3.2 O (T

LU 3.1 Z LU

Z 3.0

o

h03 2.9 Z.

< LX 2 8 •" ' 0

2

4

6

8

10

WELL THICKNESS (nm) Fig. 18. Transition energy as a function of the well thickness for samples MQW-A to MQW-D, assuming a constant x of 7%. The symbols are experimental data, while the solid and dashed lines are calculated based on polarization charges of 0 and 9.7xl0 12 C cm -2 , respectively. Note that both curves are rigidly shifted down to lower energies by llOmeV.

Recombination dynamics To understand the dynamics of radiative recombination of the M-plane MQWs, TRPL measurements at 7 K and at various excitation fluences were performed. We use a frequency-doubled Ti:sapphire laser with a photon energy of 3.4 eV, a pulse width of about 200 fs and a repetition rate of 4.75 MHz. For detection, a Hamamatsu C5680 streak camera was used. Figure 19 (a) shows TRPL transients for a 20-period M-plane (In,Ga)N/GaN MQW, after excitation of the wells only with different excitation fluences. The initial decay is rapid, faster

194

Sun, Brandt

than that we have observed for C-plane MQWs with similar structural parameters. The transients, however, cannot be fit by either a biexponential or a stretched exponential.82'91"95 It is noteworthy that the transients exhibit an identical behavior for different excitation fluences. This finding is in complete contrast to C-plane MQWs where screening of the electric field occurs at these carrier densities (1018 cm"3), and as such provides a further sign for the actual absence of internal electrostatic fields along the growth direction. Figure 19 (b) shows the transient obtained with the highest excitation fluence plotted as ln(ln[VI(t)]) vs. ln(t), in which a stretched exponential would be a straight line. Obviously, the data cannot be represented by a stretched exponential. The key for understanding the recombination dynamics for this sample is displayed in Fig. 20. The PL band exhibits rapid spectral diffusion during the first 500 ps after excitation. For longer time, the PL band still redshifts, but significantly slower. We interpret this finding as follows: following excitation at 3.4 eV, the photogenerated excitons thermalize and initially occupy extended states in the QW. These free excitons then rapidly relax towards lower-lying localized states, which is the reason for the rapid spectral diffusion as well as for the rapid initial decay of the PL transient. Once reaching the localized states, the excitons are subject to a much slower redistribution within the band of localized states, described by a stretched exponential.91 Mathematically, we thus deal with the following situation: dnAt) nf(t) —Z±± = G(t)—^-—f— dt Tf dnb(t) dt

=

nf(t) Tr

nf(t) Tr

(4)

tMnb(t)

Tb

where G(t) is the excitation pulse, nf and nb the densities of free and bound excitons, respectively, xs and /% the corresponding radiative lifetimes, ir the relaxation time, and b is the dimensionality parameter.

Nonpolar GaN' Heterostructures

TIME (ns)

195

ln[t (ps)]

Fig. 19. (a) Integrated TRPL intensity transients at three different excitation fluences for a 20-period M-plane (In,Ga)N/GaN MQW. Open circles are the experimental data, solid lines show fits with identical parameters except for the amplitude. The apparent step in the data prior to the pulse is an experimental artifact, (b) Doublelog-log representation of one PL transient. Open circles are the experimental data, while solid lines show the fit.

This rate-equation system has no analytical solution, but we have found that the sum of an exponential [«= exp(-i/r' r ), representing the relaxation from free excitons towards localized states with l/r' r = l/Tf +l/r r ] and a stretched exponential [<=c exp(- {t/t\ *f), accounting for the slow communication between the localized states] is a sensible approximation, although the values for the individual lifetimes and the dimensionality parameter b are then modified. The fits shown in Figs. 19 (a) and (b) were done with identical parameters (except, of course, for the amplitude), namely, r' r = 0.2 ns, fb = 1 ns, ji' = 0.68. The former value is a superposition of both relaxation and radiative recombination of free excitons, and thus cannot be decomposed. The latter parameters, however, roughly correspond to % = 3 ns and /3- 0.5 as judged from our numerical simulations of Eqs. (4) and (5). These are similar values as found for nonpolar cubic (In,Ga)N.9'82 Finally, it is interesting to note that the energy shift displayed in the Fig. 20(b) is logarithmic in nature and can be described by AE = -kTac ln(vf), with the acoustic phonon temperature Tac and the attempt-to-escape rate n ~ 10 s~ . This logarithmic relaxation has been observed before, both for the case of multiple-trapping relaxation of carriers within a band of localized states and for relaxation of excitons towards localized states in quantum

196

Sun, Brandt

PHOTON ENERGY (eV)

TIME (ns)

Fig. 20. The spectral diffusion of the PL band for the first 500 nm (a) and temporal variation of the photon energy at the PL spectral peak (b). Note the different slopes for short and long times. The solid lines in the inset are fits according to the equation AE = -kTac ln(vr).

The temperatures required to fit our data are found to be significantly higher than the actual lattice temperature, namely, 50 K for the short- and 20 K for the long-time region. This behavior is identical to that observed by Gobel and Graudszus96 and has been attributed to the fact that the relaxation process itself proceeds by the generation of phonons, thus increasing the phonon population over the thermal equilibrium value. In this context, the difference of the phonon temperatures extracted from our experiments is easy to be understood: while the relaxation of free excitons towards localized states proceeds exclusively by the generation of phonons, excitons, once occupying localized states, have at least partly to rely on existing phonons to assist them to migrate within the band of localized states. 6. Conclusion and Outlook In this chapter, it has been shown that high-quality M-plane GaN films can be reproducibly synthesized on y-LiAlO2(100). The phase purity and surface quality of these M-plane GaN epilayers are very sensitive to nucleation conditions. Next studies on Ga adsorption and desorption kinetics establish optimum growth conditions of M-plane GaN films, which is crucial for a further investigation of this promising

Nonpolar GaN Heterostructures

197

material. Based on high-quality M-plane GaN buffer layers, the successful growth of (In,Ga)N/GaN(l-100) MQWs has been demonstrated. Significant In surface segregation was determined by both HRXRD and SIMS. This knowledge is of importance for realizing higher In incorporation in the future. It was then demonstrated that these heterostructures are, as expected, free of internal electrostatic fields along the growth direction. As a consequence, a high PL efficiency of these structures, even at elevated temperatures (500 K) and for wide wells, was observed. This finding indicates the potential of nonpolar (In,Ga)N/GaN as active region in light emitting devices. Future work on this promising material should concentrate on the following issues: First, the surface quality of y-LiA102 is still inferior when compared with commercially available SiC(OOOl) and Al2O3(0001) wafers. For further improving the crystal quality of M-plane GaN, it is desirable to find a way to improve the surface morphology of y-LiA102. Second, the In incorporation efficiency is quite low in our M-plane (In,Ga)N/GaN MQWs. Exploring appropriate growth conditions for higher In contents will be an interesting topic. Finally, we note that M-plane bulk A1N substrates are now commercially available.98 Regarding the chemical and thermal stability, A1N is certainly superior to y-LiA102 as a substrate. With an identical structure, M-plane A1N has the same stacking sequence as M-plane GaN. Thus, GaN films grown onM-plane A1N substrates might be free of the planar defects observed in those grown on structurally incommensurable substrates, such as y-LiAlO2(100). In addition, A1N is an O-free material, unlike LiA102. One thus does not need to worry about O contamination originating from the substrate during growth, especially for MOCVD epitaxy. Therefore, the epitaxial growth of GaN films and relevant structures on M-plane A1N substrates will be technologically and commercially interesting in the future.

Acknowledgments The authors gratefully acknowledge Klaus H. Ploog for his continuous support and Hans-Peter Schbnherr for his invaluable technical assistance.

198

Sun, Brandt

References 1. T. Deguchi, K. Sekiguchi, A. Nakamura, T. Sota, R. Matsuol, S. Chichibu, and S. Nakamura, Jpn. J. Appl. Phys., Part 2, 38, L914 (1999). 2. F. Bernardini, V. Fiorentini, and D. Vanderbilt, Phys. Rev. B 56, 10024(R) (1997). 3. A. Bykhovski, B. Gelmont, and M. Shur, J. Appl. Phys. 74, 6734 (1993). 4. S. Kaiser, M. Jakob, J. Zweck, W. Gebhardt, O. Ambacher, R. Dimitrov, A. T. Schremer, J. A. Smart, and J. R. Shealy, J. Vac. Sci. Tech. B 18, 733 (2000). 5. S. Arulkumaran, T. Egawa, H. Ishikawa, and T. Jimbo, J. Vac. Sci. Tech. B 21, 888 (2003). 6. N. Grandjean, B. Damilano, S. Dalmasso, M. Leroux, M. La'ugt, and J. Massies, J. Appl. Phys. 86, 3714 (1999). 7. R. Langer, J. Simon, V. Ortiz, N. T. Pelekanos, A. Barski, R. Andr'e, and M. Godlewski, Appl. Phys. Lett. 74, 3827 (1999). 8. S. F. Chichibu, M. Sugiyama, T. Kuroda, A. Tackeuchi, T. Kitamura, H. Nakanishi, T. Sota, S. P. DenBaars, S. Nakamura, Y. Ishida, et al., Appl. Phys. Lett. 79, 3600 (2001). 9. S. F. Chichibu, M. Sugiyama, T. Onuma, T. Kitamura, H. Nakanishi, T. Kuroda, A. Tackeuchi, T. Sota, Y. Ishida, and H. Okumura, Appl. Phys.Lett. 79, 4319 (2001). 10. O. Brandt, in Group III Nitride Semiconductor Compounds: Physical and Applications, edited by B. Gil (Oxford Univ. Press, New York, 1998). 11. Z. H. Feng, H. Yang, X. H. Zheng, Y. Fu, Y. P. Sun, X. M. Shen, and Y. T. Wang, Appl. Phys. Lett. 82, 206 (2003). 12. P.Waltereit, O. Brandt, A. Trampert, H. T. Grahn, J. Menniger, M. Ramsteiner, M. Reiche, and K. H. Ploog, Nature 406, 865 (2000). 13. H. M. Ng, Appl. Phys. Lett. 80, 4369 (2002). 14. M. D. Craven, S. H. Lim, F. Wu, J. S. Speck, and S. P. DenBaars, Appl. Phys. Lett. 81, 469 (2002). 15. M. D. Craven, S. H. Lim, F. Wu, J. S. Speck, and S. P. DenBaars, Appl. Phys. Lett. 81, 1201 (2002). 16. E. Kuokstis, C. Q. Chen, M. E. Gaevski, W. H. Sun, J. W. Yang, G. Simin, M. A. Khan, H. P. Maruska, D. W. Hill, M. C. Chou, et al., Appl. Phys. Lett. 81, 4130 (2002). 17. C. Q. Chen, M. E. Gaevski, W. H. Sun, E. Kuokstis, J. P. Zhang, R. S. Q. Fareed, H. M. Wang, J. W. Yang, G. Simin, M. A. Khan, et al., Appl. Phys. Lett. 81, 3194 (2002). 18. C. D. Lee, R. M. Feenstra, J. E. Northrup, L. Lymperakis, and J. Neugebauer, Appl. Phys. Lett. 74, 1793 (2003). 19. M. D. Craven, P. Waltereit, F. Wu, J. S. Speck, and S. P. DenBaars, Jpn. J. Appl. Phys., Part 2 42, L235 (2003). 20. H. M. Ng, A. Bell, F. A. Ponce, and S. N. G. Chu, Appl. Phys. Lett. 83, 653 (2003).

Nonpolar GaN Heterostructures

199

21. B. A. Haskell, F. Wu, M. D. Craven, S. Matsuda, P. T. Fini, T. Fujii, K. Fujito, S. P. DenBaars, J. S. Speck, and S. Nakamura, Appl. Phys. Lett. 83, 644 (2003). 22. H. Lu, W. J. Schaff, L. F. Eastman, J. Wu, W. Walukiewicz, V. Cimalla, and O. Ambacher, Appl. Phys. Lett. 83,1136 (2003). 23. W. H. Sun, J. W. Yang, C. Q. Chen, J. P. Zhang, M. E. Gaevski, E. Kuokstis, V. Adivarahan, H. M.Wang, Z. Gong, M. Su, et al, Appl. Phys. Lett. 83, 2599 (2003). 24. C. Q. Chen, V. Adivarahan, J. W. Yang, E. Kuokstis, and M. A. Khan, Jpn. J. Appl. Phys., Part 2 42, L1039 (2003). 25. Y. Dikme, P. van Gemmern, B. Chai, D. Hill, A. Szymakowski, H. Kalischl, M. Heuken, and R. H. Jansenl, Phys. Status Solidi C 2, 2161 (2005). 26. A. Chakraborty, , B. A. Haskell, S. Keller, J. S. Speck, S. P. Denbaars, S. Nakamura, and U. K. Mishra, Jpn. J. Appl. Phys. 44, LI73 (2005). 27. M. Marezio, Acta Cryst. 19, 396 (1965). 28. E. S. Hellman, Z. Liliental-Weber, and D. N. E. Buchanan, MRS Internet J. of Nitride Semicon. Res. 2, 30 (1997). 29. Y. J. Sun, O. Brandt, and K. H. Ploog, J. Vac. Sci. Tech. B 1350, 1350 (2003). 30. A. Trampert, T. Liu, P.Waltereit, O. Brandt, and K. H. Ploog, Inst. Phys. Conf. Ser. No. 169, 277 (2002). 31. C. H. Hsu, K. P. Ip, J. W. Johnson, S. N. G. Chu, O. Kryliouk, S. J. Pearton, L. Li, B. H. T. Chai, T. J. Anderson, and F. Ren, Electrochem. Solid-State Lett. 4, C35 (2001). 32. S. J. Beckerman, R. B. Ford, and M. T. Nemeth, Power Diffraction 11, 312 (1996). 33. B. Cockayne and B. Lent, J. Cryst. Growth 54, 546 (1981). 34. J. W. Lee., S. J. Pearton, C. R. Abernathy, R. G. Wilson, B. L. Chai, F. Ren, and J. M. Zavada, Mater. Res. Soc. Symp. Proc. 449, 1041 (1997). 35. A. Novoselov and A. Pajaczkowska, Cryst. Res. Technol. 6, 949 (1998). 36. Y. J. Sun, Ph.D dissertation, Humboldt-Universit'at zu Berlin, (2004). 37. V. Ramahandran, M. F. Brady, A. R. Smith, R. M. Feenstra, and D. W. Greve, J. Electron. Mater. 27, 308 (1998). 38. R. Kaplan, Surf. Sci. 215, 111 (1989). 39. S. Suite, M. E. Lin, and H. Morkoc,, Thin Solid Films 31, 197 (1993). 40. K. Xu, P. Z. Deng, J. Xu, Y. Z. Zhou, W. J. Liu, S. S. Jiang, and J. H. Jiang, J. Chin. Ceramic Soc. 26, 281 (1998). 41. Y. J. Sun, O. Brandt, U. Jahn, T. Y. Liu, A. Trampert, S. Cronenberg, S. Dahr, and K. H. Ploog, J. Appl. Phys. 92, 5714 (2002). 42. M. S. Yi, H. H. Lee, D. J. Kim, S. J. Park, D. Y. Noh, C. C. Kim, and J. H. Je, Appl. Phys. Lett. 75, 2187 (1999). 43. F. Yun, M. A. Reshchikov, P. Visconti, K. M. Jones, D. F. Wan, M. Redmond, J. Cui, C.W. Litton, and H. Morkoc., Mater. Res. Soc. Symp. Proc. 639, G3.17 (2001). 44. S. K. Davidsson, T. G. Andersson, and H. Zirath, Appl. Phys. Lett. 81, 664 (2002). 45. S. Tanaka, S. Iwai, and Y. Aoyagi, Jpn. J. Appl. Phys., Part 1 170, 329 (1997).

200

Sun, Brandt

46. S. A. Nikishin, N. N. Faleev, V. G. Antipov, S. Francoeur, L. G. de Peralta, G. A. Seryogin, H. Temkin, T. I. Prokofyeva, M. Holtz, and S. N. G. Chu, Appl. Phys. Lett. 75, 2073 (1999). 47. R. Ebel, M. Fehrer, S. Figge, S. Einfeldt, H. Selke, and D. Hommel, J. Cryst. Growth 201-202, 433 (1999). 48. G. Namkoong, W. A. Doolittle, A. S. Brown, M. Losurdo, P. Capezzuto, and G. Bruno, J. Vac. Sci. Tech. B 20, 1221 (2002). 49. C. Heinlein, J. Grepstad, T. Berge, and H. Riechert, Appl. Phys. Lett. 71, 341 (1997). 50. B. Pecz, M. A. D. Forte-Poisson, F. Huet, G. Radnoczi, L. Toth, V. Papaioannou, and J. Stoemenos, J. Appl. Phys. 86, 6059 (1999). 51. A. V. Sampath, A. Bhattacharyya, I. Sandeep, H. M. Ng, E. Ilipoulos, and T. D. Moustakas, Mater. Res. Soc. Symp. Proc. 639, G6.56 (2001). 52. O. Brandt, R. Muralidharan, P. Waltereit, A. Thamm, A. Trampert, H. von Kiedrowski, and K. H. Ploog, Appl. Phys. Lett. 75,4019 (1999). 53. B. Heying, R. Averbek, L. F. Chen, E. Huas, H. Riechert, and J. S. Speck, J. Appl. Phys. 88,1855(2000). 54. Y. J. Sun, O. Brandt, B. Jenichen, and K. H. Ploog, Appl. Phys. Lett. 83, 5178 (2003). 55. Y. J. Sun, O. Brandt, M. Ramsteiner, H. T. Grahn, and K. H. Ploog, Appl. Phys. Lett. 82, 3850 (2003). 56. T. Y. Liu, A. Trampert, Y. J. Sun, O. Brandt, and K. H. Ploog, Philos. Mag. Lett. 84, 435 (2004). 57. R. R. Vanfleet, J. A. Simmons, H. P. Maruska, D.W. Hill, M. M. C. Chou, and B. H. Chai, Appl. Phys. Lett. 83, 1139 (2003). 58. A. R. Smith, R. M. Feenstra, D. W. Greve, M. S. Shin, M. Skowronski, J. Neugebauer, and J. E. Northrup, J. Appl. Phys. 16, 2242 (1998). 59. J. E. Northrup, J. Neugebauer, R. M. Feenstra, and A. R. Smith, Phys. Rev. B 61, 9932 (2000). 60. G. Mula, C. Adelmann, S. Moehl, J. Oullier, and B. Daudin, Phys. Rev. B 64, 195406 (2001). 61. C. Adelmann, J. Brault, D. Jalabert, P. Gentile, H. Mariette, G. Mula, and B. Daudin, J. Appl. Phys. 91,9638 (2002). 62. C. Adelmann, J. Brault, G. Mula, B. Daudin, L. Lymperakis, and J. Neugebauer, Phys. Rev. B 67, 165419 (2003). 63. SAFIRE by VTS Schwarz GmbH, Hindenburgstrasse 12, D-76332 Bad Herrenalb, Germany. 64. O. Brandt, Y. J. Sun, L. D'aweritz, and K. H. Ploog, Phys. Rev. B 69, 165326 (2004). 65. H. J. Kreuzer, in Chemistry and Physics of Solid Surfaces VII, edited by R. Vanselow and R. F. Howe (Springer, Berlin-Heidelberg, 1988), vol. 7, p. 259. 66. M. Zinke-Allmang, Surf. Sci. Rep. 16, 377 (1992).

Nonpolar GaN Heterostructures

201

67. O. Brandt, P. Waltereit, S. Dhar, U. Jahn, Y. J. Sun, A. Trampert, K. H. Ploog, M. A. Tagliente, and L. Tapfer, J. Vac. Sci. Tech. B 20, 1626 (2002). 68. P. Waltereit, O. Brandt, K. H. Ploog, M. A. Tagliente, and L. Tapfer, Phys. Rev. B 66,165322 (2002). 69. 69 E. Monroy, N. Gogneau, D. Jalabert, E. Bellet-Amalric, Y. Hori, F. Enjalbert, L. S. Dang, and B. Daudin, Appl. Phys. Lett. 82, 2242 (2003). 70. O. Brandt, P. Waltereit, and K. H. Ploog, J. Phys. D 35, 577 (2002). 71. G. K. Williamson and W. H. Hall, Acta Metall. 1, 22 (1953). 72. O. Brandt, Y. J. Sun, H.-P. Sch'onherr, K. H. Ploog, P. Waltereit, S. Lim, and J. S. Speck, Appl. Phys. Lett. 83, 90 (2003). 73. F. B. Naranjo, M. A. S'anchez-Garc'ya, F. Calle, E. Calleja, B. Jenichen, and K. H. Ploog, Appl. Phys. Lett. 80, 231 (2002). 74. U. Jahn, S. Dhar, O. Brandt, H. T. Grahn, K. H. Ploog, and I. M. Watson, J. Appl. Phys. 93,1048 (2003). 75. S. Dhar, U. Jahn, O. Brandt, P. Waltereit, and K. H. Ploog, Appl. Phys. Lett. 81, 673 (2002). 76. S. Chichibu, T. Sota, K.Wada, and S. Nakamura, J. Vac. Sci. Tech. B 16, 2204 (1998). 77. B. J. Skromme, J. Jayapalan, R. P. Vaudo, and V. M. Phanse, Appl. Phys. Lett. 74, 2358 (1999). 78. K. Kornitzer, T. Ebner, K. Thonke, R. Sauer, C. Kirchner, V. Schwegler, M. Kamp, M. Leszczynski, I. Grzegory, and S. Porowski, Phys. Rev. B 60, 1471 (1999). 79. Y. Narukawa, Y. Kawakami, S. Fujita, S. Fujita, and S. Nakamura, Phys. Rev. B 55, R1938 (1997). 80. S. F. Chichibu, H. Marchand, M. S. Minsky, S. Keller, P. T. Fini, J. P. Ibbetson, S. B. Fleischer, J. S. Speck, J. E. Bowers, E. Hu, et al., Appl. Phys. Lett. 74, 1460 (1999). 81. K. O'Donnell, R. Martin, and P. Middleton, Phys. Rev. Lett. 82, 237 (1999). 82. J.-C. Hoist, A. Hoffmann, D. Rudloff, F. Bertram, T. Riemann, J. Christen, T. Frey, D. J. As, D. Schikora, and K. Lischka, Appl. Phys. Lett. 76, 2832 (2000). 83. S. F. Chichibu, T. Onuma, T. Sota, S. P. DenBaars, S. Nakamura, T. Kitamura, Y. Ishida, and H. Okumura, Appl. Phys. Lett. 93, 2051 (2003). 84. A. Patan'e, M. G. Alessi, F. Intonti, A. Polimeni, M. Capizzi, F. Martelli, M. Geddo, A. Bosacchi, and S. Franchi, Phys. Status Solidi A 164,493 (1997). 85. M. G. Alessi, F. Fragano, A. Patan'e, M. Capizzi, E. Runge, and R. Zimmermann, Phys. Rev. B 61, 10985 (2000). 86. P. Lefebvre, T. Taliercio, A. Morel, J. AlLegre, M. Gallart, B. Gil, H. Mathieu, B. Damilano, N. Grandjean, and J. Massies, Appl. Phys. Lett. 78, 1538 (2001). 87. J. Wu, W. Walukiewicz, K. M. Yu, J. W. Ager-IH, E. E. Haller, H. Lu, and W. J. Schaff, Appl. Phys. Lett. 80, 4741 (2002). 88. O. Mayrock, H.-J. Wunsche, and F. Henneberger, Phys. Rev. B 62,16870 (2000). 89. V. Fiorentini, F. Bernardini, and O. Ambacher, Appl. Phys. Lett. 80, 1204 (2002).

202

Sun, Brandt

90. O. Mayrock, unpublished. 91. X. Chen, B. Henderson, and K. P. O'Donnell, Appl. Phys. Lett. 60, 2672 (1992). 92. J. Z. Li, J. Y. Lin, H. X. Jiang, A. Salvador, A. Botchkarev, and H. Morkog, Appl. Phys. Lett. 69, 1474 (1996). 93. H. Scher, M. F. Shlesinger, and J. T. Bendler, Phys. Today 44 (1), 26 (1991). 94. A. Vertikov, I. Ozden, and A. V. Nurmikko, J. Appl. Phys. 86, 4697 (1999). 95 M. Pophristic, F. H. Long, C. Tran, I. T. Ferguson, and J. R. F. Karlicek, Appl. Phys. Lett. 73, 3550(1998). 95. E. O. G'obel and W. Graudszus, Phys. Rev. Lett. 48,1277 (1982). 96. J. H. Collet, H. Kalt, L. S. Dang, J. Cibert, K. Saminadayar, and S. Tatarenko, Phys. Rev. B 43, 6843(1991). 97. Crystal IS, Inc., 877 25th Street, Watervliet, NY 12189, USA..

CHAPTER 6 INDIUM-NITRIDE GROWTH BY HIGH-PRESSURE CVD: REAL-TIME AND EX-SITU CHARACTERIZATION

Nikolaus Dietz Department of Physics & Astronomy, Georgia State University P.O. Box 4106, Atlanta GA 30302-4106, USA E-mail: [email protected] Growth techniques that employ elevated pressures offer a pathway to overcome limitations in the epitaxy of high quality indium-rich group Ill-nitride compounds and related materials, which exhibit large thermal decomposition pressures. High-pressure chemical vapor deposition (HPCVD) opens an avenue for controlling stoichiometric single-phase surface compositions at optimal processing temperatures. The application of real-time optical monitoring techniques such as ultraviolet absorption spectroscopy (UVAS), principal angle reflectance spectroscopy (PARS) and laser light scattering (LLS) are crucial to gain insights into the gas-phase and surface chemistry during the heteroepitaxial nucleation and growth process. They are also critical for controlling the chemical vapor deposition process, which allows for the formation of heteroepitaxial layers of improved perfection with respect to stoichiometry, the elimination of extended defects, and the chemical integrity of the interface. The real-time optical monitoring techniques employed during InN growth demonstrated their utility in the optimization and engineering of the growth process, as well as providing crucial insights in gas phase decomposition dynamics and in surface chemistry processes under HPCVD conditions. The optical InN layer analysis by absorption spectroscopy shows that the shift of the absorption edge from 1.85eV down below 0.63eV is caused by a series of absorption centers, appearing as the stoichiometry ratio of indium and nitrogen is varied. This coincides with a sharpening of the Raman modes, indicating an improved crystalline structure.

203

204

N. Dietz

1. Introduction The fabrication of emerging detector and emitter structures as well as high-frequency/high-power devices operating at high temperature(s) that are based on group Ill-nitride compound alloys have attracted much interest in recent years. This is due to their wide range of applications that cannot be accomplished with present Si and group III-P/As materials technology. Of particular interest is an improved knowledge of the binary base system InN, and to which extend alloys and heterostructures can be formed for fabrication of optical electronic device structures.1'2 GaN and A1N alloys and heterostructures are the most studied group IIInitride compounds, however, InN is becoming increasing more significant due to its unique optoelectronic properties. At present, the growth of Ill-nitrides is mostly performed using lowpressure deposition techniques such as molecular beam epitaxy (MBE),3" 5 organometallic chemical vapor deposition (MOCVD - also denoted as MOVPE)6'7 or variations of both. However, low-pressure deposition processes are limited to regimes where the partial pressures of the constituents do not differ vastly and the decomposition process can be countered by off-equilibrium processing conditions. The application of these techniques for the growth of InN and related indium rich III-N alloys becomes a major challenge due to their stoichiometric instabilities and low dissociation temperatures, leading to inconsistent and process dependent materials properties. Even for the binary InN compound itself many of the fundamental properties are still debated. The considerable uncertainty over the band gap of InN,2'3'8'9 the influence of the intrinsic materials point defect chemistry on the optical and electrical properties,1 and the effect of extrinsic impurities such oxygen on the band gap,10'11 show a decided lack of understanding of this system. Controversial reviews of the present status of InN growth and characterization have been provided by Bhuiyan et.al.1 and Davydov et.al.12 These reviews imply that new approaches for the growth of In-rich group-III-nitride alloys need to be explored in order to improve the structural and optical properties of InN and related alloys. CVD growth of InN requires low growth temperature to overcome thermal decomposition pressures, thus limiting the quality of InN and

InN Growth by HPCVD: Real-time and Ex-situ Characterization

205

related group Ill-nitride epilayers.1'1315 This low growth temperature requires the application of extremely high V-III ratios in order to prevent the formation of metal droplets on the surface. Studies on the decomposition of InN layers10 have shown that oxygen is easily incorporated into the InN crystal under thermal treatments, and has been suggested as the source for the wide discrepancy in reports in the bandgap energy of InN. Recent studies of the indium - nitrogen system16 show considerable uncertainty in the p-T-x relations due to missing experimental validation. However, studies of the nitrogen pressure needed to prevent thermal decomposition of bulk InN provided a relationship given by17 P(Nj -> P 0 exp

AH

h _ J_

R

T " T

(1)

which results in the p-T"1 relation shown in Fig.l, 17 where AHr represent the heat of formation, R the universal gas constant and T the equilibrium temperature. This relation indicates that in the pressure range p £ 102 bar and for substrate temperatures < 900K the surface decomposition of InN will be effectively suppressed. Thin film growth at high pressures requires - in addition to specific reactor design considerations - careful considerations towards flow kinetics, gas phase reactions, and the diffusion through the surface boundary layer as well as altered surface chemistry.18 To minimize gas phase reactions, extract sufficient organometallic (OM) nutrients from the bubbler, and embed the precursor flow in the main reactor gas stream, a pulsed precursor injection scheme has been implemented, which is essential for • compression of precursors to reactor pressure, • minimization of gas phase reactions, • engineered nucleation kinetics and layer growth, and • analyzing the gas-phase and surface decomposition dynamics in real-time. The remainder of the article is sections as follows. Section II provides a brief summary on the HPCVD reactor characteristics. The following sections provide results on the topics as listed:

206

N. Dietz

2. 3. 4. 5. 6. 7. 8. 9.

HPCVD Reactor characteristics; Reactor flow characterization; Optical characterization of the precursors ammonia and TMI; Flow kinetics: Analysis utilizing pulsed gas injection; Precursor decomposition dynamics at higher pressures; Growth of InN: Real-time optical monitoring; Ex-situ characterization of InN layers, and Summary and Outlook. Temperature ( K )

0.2

2500

1667

1250

1000

833

0.4

0.6

0.8

1.0

1.2

Reciprocal temperature 10 / T (K~)

Fig. 1. Thermal decomposition pressure vs. reciprocal temperature for A1N, GaN and

2. HPCVD Reactor Characteristics In order to explore the growth at higher pressures, a high-pressure flow channel reactor with incorporated real time optical characterization capabilities19"21 was implemented at Georgia State University, GSU, capable of growing group Ill-nitrides at reactor pressures up to 100 bar. The implementation of the HPCVD system is schematically depicted in Fig.2, which shows the reactor cross section containing the substrates and optical monitoring plane, perpendicular to the flow direction. The flow channel is embedded in an inner reactor cylinder, while an outer reactor pressure vessel confines the overall pressure for well above 100

InN Growth by HPCVD: Real-time and Ex-situ Characterization

207

bar. The substrates are symmetrically embedded in the upper and lower part of the flow channel in order to prevent preferential material deposition. A fully machined and assembled half-part of the inner reactor system is shown in Fig.3.

n Optical

Fig. 2. Schematic cross section of the reactor containing the optical access ports and the center of the substrates. Two optical ports provide access to the flow channel and three ports in each of the two half sections of the reactor provide access to the growth surface.

Two of these inner parts together form the inner shell of the HPCVD reactor, which are inserted into the outer pressure confinement shell. Optical access ports are integrated along the center axis of the substrates allowing optical characterization of flow kinetics, gas phase reactions via LLS and UVAS. Optical access ports also provide for access to the substrate surface through the backside, enabling the use of principal angle reflectance spectroscopy (PARS) and LLS in back-scattering geometry.19,22'23 Both half parts of the inner reactor are identical with exemption of the angle of incidence for PARS, which have been chosen 28 deg and 30 deg for the upper and lower parts, respectively.

208

N, Dietz

Gas Inlvl

a-Al203

Optical Access ports

Heater Sapphire substrate

Fig. 3. Perspective view of one of the machined and assembled inner half-parts of the HPCVD reactor. The transparent Sapphire substrate allows a view on the heater element as well as to the optical ports entering from the back.

As schematically depicted in Fig. 4, PARS utilizes p-polarized light impinging the substrate-ambient interface near the principal angle (pP. The angle of total reflection, (pT, is approximately 5 deg above 9 P . The wavelength dependency of the principal angle will shift the principal angle (pP of more than 2 deg in the wavelength range of 200nm and 1700nm. The temperature shift is on the order of 0.2 deg between room temperature (RT) and 1000°C. Utilizing these two angles of incidence, two independent but complementary measurements can be performed at the same time. The entrance and exit ports for PARS, together with the normal incidence access port can be utilized for a variety of spectroscopic real-time techniques.

InN Growth by HPCVD: Real-time and Ex-situ Characterization

209

r

100

total reflection^

PARS: Interface Sapphire - Air

80 c-Sapphire: AI2O3 L__J

Pi

X^633 nm

1 J f

Pi

PARS:

1

60

40

/ •

-

•

p

20 ....,R S ---" 10

29.52°/

20 30 40 Angle of incidence (p; (deg)

50

Fig. 4. Angle dependency of reflectance for p- and s- polarized light at the interface Sapphire - ambient, Depict are the characteristic angles: principal angle (pP and total reflection angle
3. Reactor Flow Characterization Gaining an insight into the growth kinetics at elevated pressure requires an analysis of each step critical to the growth process. One crucial step is the maintenance of the laminar flow conditions in order to provide a consistent supply of precursor constituents. This allows the correlation of gas phase constituent concentrations to the diffusion processes and surface chemistry processes that drive the thin film growth process. In order to analyze the gas flow kinetics, LLS in forward geometry was applied, as schematically illustrated in the inset of Fig.5. The forward component of LLS is analyzed for laser light focused through the flow channel. Fig.5 shows the LLS intensities collected under constant pressure (9 bar) conditions and flow rates varied between 3 slm and 21 slm. The LLS intensity remains constant for flow rates below a critical point of approximately 7 slm. For flow rates above the critical point, a monotone increase in the intensity of LLS is observed. The region where the steady LLS intensity begins to increase is denoted as "transition

210

N. Dietz

point" and indicates the transition from a laminar to turbulent flow. A slight hysteresis in the LLS intensity is observed for increasing and decreasing flow rates in the reactor. However, within the error range the "transition point" remains the same. The "transition point" is denoted as "onset of turbulent flow". The onset of increased LLS for pure nitrogen flow is summarized in Fig.6, indicating the flow and pressure regime at which laminar flow can be maintained. The associated Reynolds number can be calculated via pul

tt)

Reynold's* Re-

where p=1.12 [kg-m"3] is the density of the gas, u is the flow velocity, / is a flow parameter - a characteristic length - associated with the reactor, and r|=1.12 [kg.m'V1] is the dynamic viscosity. For ideal gases, a direct proportionality between the density of the gas, p, and the pressure exists. In turn, this leads to an inverse proportionality between flow velocity and pressure, when the flow parameter "/" and the dynamic viscosity, rj, are constant. 5.1 -,

Gas: Nitrogen @ RT

3

sS

5.0

••a

4.9 -

GO

c

1)

X

4.8-

si 60

4.7-

increasing flow

0> 03

decreasing flow 4.6-'

Onset of turbulences

o 4.5 5

10 How (slm)

15

20

Fig. 5. Characterization of flow behavior in the HPCVD reactor. Laser light scattering (LLS) in forward geometry is used to analyze the onset of turbulence as function of pressure and flow.

InN Growth by HPCVD: Real-time and Ex-situ Characterization

211

The average of the calculated Reynolds number is around 1480 with no significant pressure dependency observed.23 As shown in section V, even more detail on the flow kinetic is revealed during pulsed precursor injection studies. Gas: N2 @ RT

\

16

:.•';•:•. A . 12

1

Turbulent gas flow regime

N.

- ' • • ' • •

V

••

^ Laminar gas flow regime I

\

^ ^

10 12 Pressure (bar)

^

— 14

Fig. 6. Transition from laminar to turbulent flow conditions as determined by LLS intensity measurements. The inset depicts the increase of the LLS with increasing pressure.

4. Precursor Characterization: Ammonia and Trimethlyindium For above atmospheric pressures, optical diagnostic techniques are uniquely suited to provide real time information pertaining to gas flow dynamics and the gas constituents. Real-time optical diagnostics are also crucial to obtain information on the precursor flow and their decomposition kinetics. Several optical techniques have been explored, but only a small group satisfies the requirements of being robust as well as sensitive. For example, the substrate temperature during typical InN growth lies between 800K and 1200K, emitting a significant radiation as shown in Fig.7. The heater radiation intensity limits the sensitivity of many optical probe techniques in the visible and infrared (IR) regime, even if modulation techniques are applied. As depict in the inset of Fig. 7, the radiation intensity for a 1000K back body emitter below 350nm vanishes very quickly with negligible contributions below 300nm. Utilizing UVAS techniques as well as UV induced fluorescence

N. Dietz

212

spectroscopy to identify the group-V and organometallic group-Ill precursors in the gas phase is well established in the literature.24"27

/

\

/

\

/

1200K

\

"3 Ofl —

\

J

1200K

/ /

\

/

/

s ^

\

\

/ /

/

/ J'y_^^

1000K

/

"^ ** "

N.

^800K ___

/

10OOK

/

/

800K

1 " \ \

/

300 /

/

/

Black Body emission

/

/

600K

500 700 Wavelength (nm)

>v

____^ 600K 4

6

10

Wavelength (nm)

Fig. 7. Intensities and spectral distribution of a black body emitter such as a hot substrate. The inset depicted on a logarithmic scale the onset the radiation from 1200K down to 600K. 4.1. Optical Characterization of Trimethylindium,

TMI,

[In(CH$)3]

A nitrogen carrier gas flow through the TMI bubbler is used to transport the TMI vapor from the bubbler to the gas control system. The molar flow of TMI directed from the bubbler in the reactor is expressed by n m = 8.3216 • 10'9 • x

[molV]

(3)

where x = (0... 100% full scale [%FS]) denotes the nitrogen flow through the bubbler with 100% = 0.5 slm maximum. The molar TMI flow ratio % through the reactor is given by n

x= — n

n „ +n

+n

2.237 • 10"5 • x z + 10"2 • x +2.237 • lfr;

(4)

InN Growth by HPCVD: Real-time and Ex-situ Characterization

213

where z = (0...100%FS) is the main nitrogen flow with 100% = 50 slm maximum flow, which dilutes the TMI stream through the reactor. 6.526

190

Energy (eV) 5.905 5.39

210 230 Wavelength (nm)

4.96

250

Fig. 8. Spectral resolved absorption on TMI diluted in N2-carrier gas as function of N2flow through TMI bubbler in %FS. The total flow through the reactor is maintained at 5 slm at 1630 mbar.

Utilizing UVAS, the TMI induced absorption was characterized through the broad absorption band observed in the wavelength range of 190nm 250nm with the absorption maximum centered at 210.7nm. Figure 8 shows the spectral resolved absorption structure as function of N2-carrier flow through the TMI bubbler. For higher TMI concentrations, two absorption centers, around 210.7nm and 221nm, can be distinguished. For higher TMI concentrations, two absorption centers, around 210.7nm and 221nm, can be distinguished. The strongest absorption maximum remains for all TMI concentrations investigated at 210.7nm, and the peak-maximum position does not change significantly with TMI concentration. The analysis of the absorption maximum at 210.7nm as a function of the molar TMI flow ratio % shows an exponential correlation in the form of

N. Dietz

214

a(x) = - 0.37367 + 0.37282 • epx

X 5.44 10

(5)

cm

which allows for the calculation of the number of TMI molecules per time unit as function of the observed absorption magnitude. Absorption maxima vs. TMI flow ratio % 100

x E c o

"-E o <

l° a(x) = -0.37367 + 0.37282 • exp( % I 5.44 • 10" ) [cm ]

8

10

12

14

16

TMI flow ratio % (105) Fig. 9. Absorption strength at X=210.7nm as function of the molar TMI flow ratio % under steady-state flow conditions.

4.2. Optical Characterization of Ammonia (NH3) Near atmospheric pressures, the ammonia vapor is transported from a gas cylinder, with a pressure set at 30 psi. The NH3 flow is controlled via a mass flow controller with a 1 slm maximum flow, expressed via y = 1 100%FS. The molar flow of ammonia and number of ammonia molecules per unit time are given by n_ii3 =7.4405-10"6-y [mols"'|.

(6)

The molar ammonia flow ratio X through the reactor, defined as the ratio of ammonia flow rate to total flow (precursor flow plus nitrogen main flow), can be expressed in term of the percentage of full scale flow, z and y as

InN Growth by HPCVD: Real-time and Ex-situ Characterization

n .

n . +n U*n_S2

215

50• z + y

^'>

where z = (0...100%FS) is, again, the main nitrogen flow with 100% = 50 slm maximum flow. The flow of ammonia through the reactor is analyzed by UVAS in the wavelength range of 180nm and 300nm as function of the molar ammonia flow through the reactor. Figure 10 shows the UV absorption spectra for ammonia flow ratios in the range of 10"1 to 10"3 at RT and a reactor pressure of 1.6 bar, which are typical molar ammonia flow ratios X required for the growth of InN. Illustrated in Fig. 10, even for the lowest flow setting, there are several of the absorption structures at higher energies that exhibit a saturation effect and are not suitable for ammonia characterization in the molar flow regime. For the ammonia flows used during the growth of InN, the UV absorption peaks centered at 217.lnm and 221.6nm are best suited. The correlation for these two absorption peak maxima with the flow ratio can be expressed as a^.2.7..„Cr) = 0.38 • ln(* +0.011) -2.0 • * +1.73 [cm1] 01

peak ,221,6mi

(8)

(%) =-45+45.01 ' e x p ^ / J • 10"2 [cm1]

For ammonia flow ratios in the range of % = 1.0 x 10~2 to 1.6 x 10"1, the absorption maxima at 221.6nm is used to provide the correlation between UV absorption and the molar flow ratio %. The number of NH3 molecules per unit time is computed as function of the observed UV absorption. For the UV absorption feature located at 221.6nm, we find the number of NH3 molecules per time unit as 7.17-10 21 .z- ln(a') N m , a=221.6nn) = mh

p

rll

.L ,

S ] With Of =

aa2216„-80 ® 22'6""

,„, .

(9)

1-32 • ln(a') 80.01 Figure 11 shows the correlation between ammonia molecules per time unit and the ammonia flow ratio in the range of % = 1.0 x 10"2 to 9.0 x 10'1, for a reactor pressure of 1.6 bar. Under those conditions, the ammonia flow can be varied between 1019 and 2.5 x 1020 NH3 molecules per sec.

216

N. Dietz

Energy (eV) 6.20

5.90

0

2

4

5.64

6

5.39

8

NH5 flow ratio x ( 10'! )

Fig. 11. Calculated concentration of ammonia molecules per sec using the absorption line at X = 221.6nm under continuous flow conditions.

5. Flow Kinetics: Analysis Utilizing Pulsed Gas Injection The flow of the precursors at higher pressures requires a compression and dilution step in order to allow the precursors to be injected in the

InN Growth by HPCVD: Real-time and Ex-situ Characterization

217

HPCVD reactor. To accomplish this, a reservoir is filled at slightly above atmospheric pressure. In the following steps the reservoir is compressed with nitrogen carrier gas and temporally controlled injected into the reactor. The cycle repetition rate, duration of injection, and position of injection can the adjusted within 10ms resolution. Figure 12 shows, as an example, typical absorption traces monitored at 210.7nm during pulsed TMI injection with a 6 sec repetition period for various reactor pressures, keeping the flow constant. The total number of TMI molecules flowing through the reactor can be calculated using the relationship between the UV absorption and TMI flow rate provided in equation 5, taking in to account the compression ratio and gas reservoir volume. ITMI pulse

|TMI pulse

Fig. 12. Absorption traces monitored at 210.7nm during TMI precursor pulse injection in the reactor at constant flow of 5slm. The reactor pressure was varied between 1 and 12 bar. The pulse cycle sequence is 6 s with 0.2s TMI injection time.

218

N. Dietz

The pulsed precursor injection has been analyzed as function of pulse width, precursor molecules per pulse, total reactor flow and reactor pressure. A carryover of the UV absorption trace from one sequence to the next is observed at reactor pressures above 10 bar. The result is an increase of the base line in the overall UV absorption. As the reactor pressure is increased for a given fix flow rate, the precursor pulses monitored at the substrate centerline show three distinct features: (i) a systematic shift in the pulse arrival time, (ii) a systematic TMI pulse broadening, and, (iii) a change in the TMI absorption for pressures larger 7 bar. 5.1. Flow Characterization During Pulsed Precursor Injection The time delay, At, between the start of the precursor pulse injection sequence and the arrival of the diluted TMI gas at the center of the substrate is determined by the pneumatic valve opening time tv, the reactor flow channel geometric factor rg, the reactor pressure pr (in bar), and the total gas flow through the reactor in terms of standard liters per min (slm). The analytic relationship is, At = t v

+

^=tv+rg.^g

(1Q)

slm

The reactor geometry factor rg is a constant. It is functional dependent on the system parameters such as the reactor cross section A and the distance ld between the injection valve and the substrate centerline. Its unit is [min"1 • bar"1 • s"1]. The analysis of At for precursor pulse injection rates of 6s is shown in Fig. 13 as function of total gas flow and reactor pressure. Under these conditions, an analysis of At reveals that the pneumatic valve opening time tv = 240ms and the reactor geometry factor rg = 0.70 can be treated as constant values. Based on the analysis the average gas velocity vg can be computed as v=

1 V

SSL =136.7

V

**- [cm« s ]

n n

InN Growth by HPCVD: Real-time and Ex-situ Characterization

219

Note, that this is not the gas velocity over the substrate but rather the average velocity for the reactor system. Based on the reactor cross section A, and the gas volume per time unit, the average flow velocity over the substrate is estimated a factor 2 smaller than indicated by the average velocity vg. This is due to the larger reactor flow channel cross section, compared to the gas lines. The systematic precursor pulse broadening shown in Fig. 12 is a direct result from the relationship between gas flow velocity and pressure given by equation 11. The reason for the pronounced increase in the TMI absorption observed for pressures larger than 6 bar is at present not fully understood and requires a more detailed study. A similar increase is observed for the ammonia precursor, for pressures above 8 bar. The UV absorption cross-section appears to have a functional dependence on pressure since at high pressures the UV absorption in noticeably increased. Such an increase would be beneficial for the decomposition kinetics of the precursors since it would result in a more efficient decomposition of the precursors. However, detailed theoretical calculations will be required and be validated by experimental real-time measurements. lOslm

2.0

600

TMI pulse travel monitored @ ^=210.7 run

W

•3

"o

1.5

400

5 slm

> 1.0 200

i

a

> <

0.5

2

4

6 8 Pressure (bar)

10

12

Fig. 13. Time shift between injection and onset of pulse arrival at the substrate centerline as function of the reactor pressure and for different flow rates. The right scale shows the computed average gas velocity vg between the TMI reservoir and the substrate center line.

220

N. Dietz

As shown above, the real-time optical analysis of pulsed precursor injection, provides crucial information pertaining to • the reactor flow characteristics and average gas flow velocity, • pulse broadening, and • pressure related changes in the optical properties of the precursors. What is most important, however, is that these features provide a pathway for the monitoring and engineering of the gas phase chemistry and surface chemistry by enabling precise engineering of precursor pulse separation and/or overlap. Such knowledge and capabilities are crucial for a more physically realistic and accurate modeling effort and for more precise control of the growth process at high-pressures. 6. Precursor Decomposition Dynamics at Higher Pressures The analysis of the decomposition dynamics of ammonia under continuous ammonia flow conditions and slightly above atmospheric pressures was done with the aid of UVAS.28 However, at higher reactor pressures, a continuous precursor decomposition analysis is no longer possible. A periodic pulse injection scheme is used in which the average absorption peak maxima are analyzed as a function of precursor concentration and temperature. Figure 14 shows the temperature dependency of the UV absorption of ammonia monitored at 210.7nm while maintaining a reactor pressure of 10 bar. For this elevated pressure, the onset of decomposition as indicated by a decrease in the UV absorption occurs at a temperature of about 850K. Compared to studies at atmospheric pressures, where the ammonia decomposition is observed at about 900K, this is a significant reduction in the decomposition temperature. As for ammonia, we also analyzed the TMI UV absorption peak maxima at 213nm during pulsed precursor injection as a function of pressure and temperature. At present, we a not aware of any experimental studies on the decomposition dynamics of TMI at higher pressures. Figure 15 shows the TMI peak absorption data for a reactor pressure of 10 bar as function of temperature. The analysis indicates that the onset of decomposition in the gas phase occurs around 800K.

InN Growth by HPCVD: Real-time and Ex-situ Characterization

221

This onset in decomposition is slightly higher than those reported under low-pressure OMCVD conditions.25'29 More detailed studies using UVAS and optical emission spectroscopy as function of pressure are required to correlate the experimental results to theoretical predications for the TMI decomposition at elevated pressures as formulated by 30 Cardelino et.al. NH absorption peak maxima *=210.7nm; p

a:

=10 bar; Y

1.9 • ](T

2.4-

800

1000

1200

Temperature ( K )

Fig. 14. Change of the ammonia absorption peak maximum as function of temperature.

TMI absorption peak maxima ^=210.7 nm; p ,.. = 10 bar; v

I

= 1.06 • 10

6-

I 700

900 Temperature ( K )

1100

Fig. 15. Decomposition of TMI at 10 bar reactor pressure, monitored during pulsed TMI injection as function of temperature.

222

N. Dietz

The observed decrease of the temperature at which the onset of ammonia decomposition occurs under elevated pressure conditions is crucial for the optimization of the growth of InN and the control of point defect chemistry in this material system. 7. Growth of InN: Real-time Optical Monitoring The decomposition studies for ammonia in the previous section suggests that for sufficient cracking of the ammonia precursor growth temperatures of above 1000K to be used. However, literature data for InN growth by OMCVD indicate a growth temperature of 675K to 750K,7 775K,31 810K - 840K.32 Under HPCVD conditions the growth temperatures can be increased significantly as shown below. Reported NH3:TMI flow ratios under low-pressure OMCVD conditions vary from 103 to 104 in order to counteract the low ammonia decomposition at the growth temperatures.7'31 The growth by HPCVD has to address the control of gas phase reactions and the effective diffusion of the nutrients to the growth surface. These considerations lead to the concept of a pulsed injection scheme as schematically shown in Fig. 16. The TMI and NH3 precursors are temporally controlled and embedded in a high pressure carrier stream, consisting of ultra-pure nitrogen, supplied from a boil-off liquid nitrogen tank. The total gas flow as well as the reactor pressure are kept constant at all times. The repetition rate "cycle sequence" as well as the precursor pulse length and position within the cycle sequence are crucial growth control parameter that allow the precise engineering of gas phase and surface chemistry processes during the nucleation phase and during steady-state growth conditions. For the InN layers discussed in more detail below, the cycle sequence time was varied from 4 sec to 10 sec, with TMI and ammonia pulse widths from 0.3-0.6sec and 0.8-1.5 sec, respectively. The pulse separation was varied from 1 to 5 sec. A typical growth procedure contains the following steps. The symmetrically embedded substrates in the upper and lower part of the reactor are heated to approximate 1150K - 1200K and exposed to ammonia for typically 30 min. Prior to the growth, the temperature is

InN Growth by HPCVD: Real-time and Ex-situ Characterization

223

lowered to the growth temperature, and the InN growth is initiated by supplying the precursors sequentially as shown in Fig. 16. The integrated optical access ports along the center axis of the substrates are used to monitor the gas constituents by UVAS, as described in section IV and V.

InN Precursor Injection Sequence

N2 - main flow |TMIi 0.0

Fig. 16. Schematic representation of a precursor cycle sequence used for the growth of InN via the precursors TMI and ammonia.

The growth surface conditions are monitored through the backside of the sapphire substrate using PAR and LLS.19'22'23 For the growth results presented here, the reactor pressures were kept around 10 bar and a total carrier flow rate in the range 2 slm to 12 slm. The precursor flow was evaluated for molar ratio ammonia to TMI, RNH3:TMI, from 100 to 5000. The growth temperature has been varied from 800K to 1150K. Note that all temperatures settings refer to a calibrated correlation curve between the measured black body radiation distribution as function of "heater" power setting. Keep in mind, that the actual gas phase temperatures and the growth surface temperatures are strongly influenced by the gas flow velocity (total main flow) and surface emissivity. Specifically, the surface emissivity changes during growth, which affects the actual growth temperature. The temperature referred to is not corrected for these effects. A typical set of real-time optical monitoring traces by PAR and LLS is illustrated in Fig. 17. The temporal evolution of the PAR trace contains crucial information related to the growth surface as well as to the growth

N. Dietz

224

history. The developing PAR interference fringe provides information on the overall layer growth (history), while the superimposed 'fine structure' on the PAR interference fringe (not resolved in Fig. 17) provides insights on the growth surface chemistry and kinetics (see below). The monitored LLS trace tracks the evolution of the surface morphology, providing details on the nucleation and overgrowth kinetics as well as the overall surface roughness. As shown in Fig. 17, the LLS signal increases during the film nucleation phase, but it decreases during the further steady-state growth. The ex-situ inspection revealed a mirrorsmooth surface.

0

2000

4000

6000

8000

10000

Time (s)

Fig. 17. Real-time optical monitoring of InN growth by PAR and LLS.

From the analysis of the PAR signal, the average growth rate and the difference between the dielectric functions of film and substrate can be estimated. The analysis of In#14u provided the average growth rate as 1.346A per cycle sequence. By varying the TMI and ammonia concentration per injection pulse, the growth rate per cycle sequence can

InN Growth by HPCVD: Real-time and Ex-situ Characterization

225

be adjusted from a fraction of a monolayer (ML) to IML and beyond per cycle sequence. This capability allows the precise engineered growth of nano-structured composites with sub-monolayer resolution. A more detailed insight in the growth process is gained by analyzing the PAR fine structure and by linking it to the UV absorption traces, which monitors the gas phase constituents. As an example, Fig. 18 shows the observed PAR and UV absorption traces during the nucleation phase and steady-state growth of InN. The lower half in Fig. 18a shows the UV absorption trace recorded for the wavelength >,=210.8nm, monitoring the un-decomposed ammonia and TMI species above the growth surface. The PAR trace in the upper half of the figure is recorded for the wavelength X=632.8nm, monitoring highly sensitive changes in the dielectric function at the substrate-ambient interface. Also indicated in the figure are the positions of the precursor pulse injections with a total cycle sequence repetition time of 6 sec. First, note that the precursor injection time and the response seen in UVAS and PAR are temporally shifted, which is due to the average travel time of the precursors between valve and substrate center line (see section V). Secondly note, it takes about two cycle sequences before the UV absorption feature for TMI clearly develop (see arrows). Looking at the PAR response, a large increase is observed after the first TMI ammonia combination is introduced, indicating the start of InN nucleation and the presence of TMI fragments in the vicinity of the growth surface. A steady state surface chemistry is typically reached after 5 to 20 cyclic precursor exposures, depending on substrate temperature, precursor flow ratio, gas phase velocity and reactor pressure. Figure 18b shows the PAR and UVAS responses during steady-state growth conditions. The periodic modulation of the PAR response can be directly correlated to the presence of ammonia and TMI fragments in a surface reaction layer and at the growth surface. The overall decrease in the PAR signal is correlated to the InN growth per cycle sequence as discussed in detail for p-polarized reflectance.22'33

226

N. Dietz

a) , 6.0

<

Nucleation of rnNonc-Al,O((0001)

5.0

p.

3 X)

12

18 Time (s)

Fig. 18. (a) Monitoring of InN nucleation by PAR and UV absorption traces. A precursor cycle sequence of 6 sec with 0.4 sec TMI and 1.4 sec ammonia pulses, separated by 1.4 sec were used, (b) PAR and UV absorption traces during steady-state InN growth at 990K. The rector pressure was 10 bar with a total flow of 5 slm. The overall decrease in the PAR signal corresponds to InN growth.

InN Growth by HPCVD: Real-time and Ex-situ Characterization

227

At present, only the UV absorption features of TMI are monitored and related to the PAR response. The optical access ports in the HPCVD reactor enable the use UV absorption spectroscopy,26 fluorescence,25 and Raman spectroscopy34 to monitor the decomposition process and the concentration of the TMI fragments. Such detailed knowledge will provide the base for a more comprehensive growth modeling that will link the growth process and flow kinetic models to surface chemistry models and the thin film growth process itself. 8. Ex-situ Characterization of InN Layers The structural properties of epitaxially grown InN films have been investigated using X-ray diffraction, Auger Electron Spectroscopy (AES), and Raman spectroscopy. Figure 19 depicts the XRD pattern recorded in the co-20 mode for sample #C, showing single phase InN diffraction peak at 31.265, due to the (0002) reflection from wurtzite-type InN with fullwidth at half-maximum (FWHM) on the order of 800 arcsec. The compositional analysis by AES (including depth profile analysis) shows indium and nitrogen as the main constituents. The AES analysis of earlier samples indicated oxygen and carbon contaminations of up to one percent. The installation of an in-line oxygen purifier reduced the oxygen concentrations below the detection limits, confirming that the oxygen contamination correlates to the residual water content in the ammonia source. No oxygen contamination was found in the main carrier gas, for which ultra-pure nitrogen (boil-off from liquid nitrogen tanks) is used. For the samples discuss below, no correlation of oxygen content with the absorption edge shift was found, confirming studies by Butcher et al.11 indicating that the incorporation of oxygen does not directly correlate to changes in the lattice constant. The observed carbon contamination was found to scale with the flow velocity and growth temperature. For a constant substrate temperature and estimated gas flow velocities below 40cm«s"1, visible carbon precipitations are observed towards the end substrate. Increasing the gas flow velocity to 45cm»s"1 and above resulted in uniform InN layers. Note, even though the substrate temperature was kept constant, an increase in the gas flow velocity will not only lead to an effective lower gas phase

N. Dietz

228

temperature, but will also effect the diffusion rate of the precursor constituents to the growth surface. More experimental studies are needed to link the observed data to a comprehensive model that combines gas flow kinetics, gas phase chemistry, and diffusion processes to the surface growth dynamics. The carbon incorporation in the layers and its effect on the material defect chemistry will be he most challenging part in the material optimization. Sapphire ((XX) 1)

3

i a OS

30

35 40 2 Theta / Omega [dgr]

Fig. 19. XRD spectra from InN layers grown on sapphire (0001) with an ammonia to TMI flow ratio of 1000.

Raman spectroscopy has been applied as a routine technique in order to evaluate the crystallinity of the layers. The Raman spectra were measured at room temperatures in a backscattering geometry with an excitation energy of 2.33eV. Here we focus on the analysis of five representative InN samples labeled sample #A through #E. The samples were grown at temperatures around 1100K, with molar ratio ammonia to TMI, RNH3:TMI, of 200 and 1000. All samples are typically 500nm to 800nm thick. Figure 19 shows the Raman spectra of these five samples, together with the Raman spectrum of the sapphire (0001) substrate. Sample #A and #B were grown with RNH3:TMI = 1000, while for the

InN Growth by HPCVD: Real-time and Ex-situ Characterization

229

samples #C , #D and #E, the ammonia to TMI ratio has been subsequent lowered to approximate 200. The Raman spectra of samples #A and #B show a broad asymmetric structure that can be fitted by three Lorentzian distribution functions located approximately at 580cm"1, 540cm"1, and 470cm"1. For samples #C through #E, the broad asymmetric structure breaks in at least 5 structures with significant variations in their peak intensities and FWHM. The polarization-dependent Raman studies on these samples remain to be performed. InN on Sapphire (0001) X. =532 nm @ RT f^j^f

"\

NH/TMI ratio 200

1

|^lNvAl|*f 300

400

500

700

600

800

1

Wavenumber (cm ) Fig. 20. Raman spectra for InN sample* A through #E. The layers were deposited around 1100K with varying the ammonia to TMI ratio.

Following the notation and peak assignment nomenclature provided by V. Davydov et.al.2 and considering the modes of symmetry that are Raman active, the high-resolution Raman spectra for samples #E and #D

N. Dietz

230

in the range of 420cm"1 and 620cm"1 were fitting by five Lorentzian distribution functions as depicted in Fig. 21. £

a) •£

2%a

InN on Sapphire (0001)

e

o

iff

^

,|;: 0 . 8 - Sample #C % =532 nm @ RT i """'

q 5"

/ 1.

"'

II

•*—•

_

f 1 t*/ •i 11 r. . \

•fr

• •m

3

' 'J'?1

;

'3=* •«

A :1 M

. ; f?'<|

''''fir .i* M *l ' Vi ^^ \Y<\• 'v*# 'WI ,*/j.i I •! >r'H ?v fY I •!

w

.'/::i :. .a" W-$J\ \ M V\! I;-:;, s?!!i(l/ ^' •'•fj *''Vi '''•"'''! ' >!.''' i'• '^Tj* I .'"*. * i *i "V

_S B 0.6 >S

man

(

£ .,.',§ %M00f\l \ / %

:;L« \fff,.f/ y * .

8

06 0.4

•; i ?l jt$$$f / -V l i i " * ^ ^ * ! ' / ..-••" \ ZwM -•

'

'..-&•••::' i

'

—

450

..•'

/•„ / \ "•*•.,•' \

':•-:•'' '

1 - •

0*1$ ••3;'

..-' T

—

.*-->' '

—

'

* -

600

500 550 Wavenumber (cm1)

InN on Sapphire (0001)

b) ''

S

I

Sample #E 0.8

•

' ' ,

1

^

l/S&'l' ' «"

f

S &

X^ =532 nm @ RT f :

-,

f<W$iA -M'

'•• '' <;':*

|^.;^^iy A. | V

.J^^Ui^fi''' '* / ^ \

I 2

;

1 m^Jr^ j Vm 0.4

iy$,J?/ \ S / | •'X\

'&&&•'•

450

%

500 550 Wavenumber (cm')

A; \ \

/ \ "• ? \

600

Fig. 21. Analysis of Raman modes for (a) sample* C and (b) sample #E (see text).

InN Growth by HPCVD: Real-time and Ex-situ Characterization

231

The analysis of the peak-positions and FWHM are summarized in Table 1, together with literature values. The most pronounced difference in the spectra of samples #C and #D is the significant increase of the Al(LO) and Al(TO) peak intensities in sample #D. Table 1: Frequencies and FWHM of Raman peaks in samples #C and #D Sample #

#C: Pos.

A,(TO)

449 cm"1

E^TO)

-

FWHM 40 cm-1 #D: Pos.

451cm-'

-

FWHM 8.9 cm'1 Ref. 2, 35 447

475

E2(high)

Bi(high)

490 cm"1

525 c m 1

567.4 cm"1

591 cm"1

17 cm"1

50 cm"1

13 cm"1

15 cm"1

488 m"1

528 m"1

565.3 m"1

588 m"1

35 cm'1

61 cm"1

40 cm"1

26.6 cm"1

488

565

586

593

Ai(LO)

E,(LO)

In order to correlate the structural properties, analyzed by Raman spectroscopy, to the optical properties, absorption and reflectance spectroscopy have been applied. The absorption spectra for the five samples were taken in the spectral range of 350nm to 2300nm, the results of which are summarized in Fig.22. All samples show a pronounced absorption peak, centered at 0.63eV. Starting with sample #A, three additional absorption peaks at 0.87eV, 1.35eV and 1.8eV (shoulder) are observed. In samples #B and #C, the absorption peaks at 0.87eV and 1.35eV are increased in intensity but still visible as absorption centers. The spectrum for #C suggest already an overall shifted absorption edge and it also indicate a new (or shifted) absorption structure at leV. The absorption peaks at 0.63eV and 0.87eV seem have merged for sample #C. The absorption spectrum for sample #D shows a further increase in the absorption intensity for the absorption center at 1.35eV, shifting the absorption edge further down below leV. The absorption spectrum for sample #E indicate an absorption edge below 0.63eV. The development of the absorption structures for these five samples suggest that the observed absorption edge shift from 1.85eV down to below 0.63eV is caused by a series of absorption structures, centered around 1.6eV, 1.35eV, leV, 0.87eV and 0.63eV. The appearance and

232

N. Dietz

strength of these absorption centers correlate / coincide with the strength and FWHM of the Raman modes shown in Fig.20 and 21. They also correlate with the reduction of the molar ammonia to TMI precursor flow ratio, suggesting a close correlation between precursor ratio and the stoichiometry of the deposited layers. These initial results demonstrate that the precise control of the indium to nitrogen ratio is of crucial importance for optimizing the InN layer quality and with it the optoelectronic properties in this material system.

I

3.0

'

'

'

'

i

'

2.5

'

'

'

i

2.0

•

'

'

'

i

1.5

'

'

'

'

i

'

'

•

'

1.0

Energy (eV)

Fig. 22. Absorption spectra for various InN layers grown at 12 bar and HOOK. The gas flow velocity over the growth substrates and the molar ratio of ammonia to TMI was varied (see text).

9. Summary and Outlook High-pressure chemical vapor deposition has been introduced for the growth of InN and related materials and shown to be a valuable method

InN Growth by HPCVD: Real-time and Ex-situ Characterization

233

for achieving this goal. The growth of InN has been accessed in the pressure range of 10 to 15 bars for various gas flow velocities and molar ammonia to TMI ratios. High quality InN layers were achieved for growth temperatures in the range of 1000K to 1150K with the potential of even higher substrate temperatures at higher reactor pressures. Realtime optical growth characterization by UVAS, PAR and LLS has been applied, providing critical insights in the gas phase decomposition kinetics, surface chemistry processes during the film growth process. The high sensitivity of PAR in combination with UVAS will allow to formulate a comprehensive gas flow and surface chemistry model which will enable growth control with sub-monolayer resolution. The optical InN layer characterization results show that the shift of the absorption edge form 1.85eV down to below 0.63eV is caused by the appearance of several absorption centers as the crystallinity of the InN lattice improves. At present, the origin of these absorption centers is not clear. However, they are clearly linked to the indium and nitrogen point defect chemistry. The initial results suggest that the molar ammonia to TMI flow ratio under HPCVD conditions is below 200 due to the efficient cracking of the nitrogen precursor at the high growth temperature. Further studies varying the ammonia to TMI flow ratio, the center flow velocity, and the growth temperatures will be needed to access the optimum growth window. A further increase in the reactor pressure may allow even higher growth temperatures, bringing it closer the optimum growth temperature of GaN. This will allow for the exploration of indium rich In!.xGaxN and In^AlxN alloys and hetero structures. Acknowledgments This work was supported by NASA grant NAG8-1686 and GSU-RPE. References 1. A. G. Bhuiyan, A. Hashimoto, and A. Yamamoto, J. Appl. Phys. 94, 2779 (2003). 2. V. Yu. Davydov and A. A. Klochikhin, Semiconductors 38, 861 (2004).

234

N. Dietz

3. T. Inushima, V. V. Mamutin, V. A. Vekshin, S. V. Ivanov, T. Sakon, M. Motokawa, and S. Ohoya, J. Crystal Growth 227-228, 481 (2001). 4. J. Wu, W. Walukiewicz, W. Shan, K. M. Yu, J. W. Ager III, E. E. Haller, Hai Lu, and William J. Schaff, Phys. Rev. B 66, 201403 (2002). 5. R. A. Oliver, C. Norenberg, M. G. Martin, M. R. Castell, L. Allers and G. A. D. Briggs, Surface Science 532-535, 806 (2003). 6. J. Aderhold, V. Yu. Davydov, F. Fedler, H. Klausing, D. Mistele, T. Rotter, O. Semchinova, J. Stemmer and J. Graul, J. Crystal Growth 222, 701(2001). 7. Bi Z.X. Bi, R. Zhang, Z.L. Xie, X.Q. Xiu, Y.D. Ye, B. Liu, S.L. Gu, B. Shen, Y. Shi and Y.D. Zheng, Mat. Lett. 58, 3641 (2004). 8. J. Wu, W. Walukiewicz, K. M. Yu, J.W. A.III, E. E. Haller, H. Lu, W.J. Schaff, Y. Saito, and Y. Nanishi", Appl. Phys. Lett. 80, 3967 (2002). 9. I. Vurgaftman and J.R. Meyer, J. Appl. Phys. 94, 3675 (2003). 10. E. Kurimoto, M. Hangyo, H. Harima, M. Yoshimoto, T. Yamaguchi, T. Araki, Y. Nanishi, K. Kisoda , Appl. Phys. Lett. 84, 212 (2004). 11. K. S. A. Butcher, M. Wintrebert-Fouquet, P. P.-T. Chen, T. L. Tansley, H. Dou, S. K. Shrestha, H. Timmers, M. Kuball, K. E. Prince, and J. E. Bradby, J. Appl. Phys. 95,6124(2004). 12. V. Yu. Davydov, A. A. Klochikhin, V. V. Emtsev, A. V. Sakharov, S. V. Ivanov, V. A. Vekshin, F. Bechstedt, J. Furthmuller, J. Aderhold, J Graul, A. V. Mudryi, H. Harima, A. Hashimoto, A. Yamamoto, J. Wu, H. Feick and E. E. Haller, Proc. SPIE 5023, 68 (2003). 13. V.Yu. Davydov, A.A. Klochikhin, V.V. Emtsev, D.A. Kurdyukov, S.V. Ivanov, V.A. Vekshin, F. Bechstedt, J. Furthmuller, J. Aderhold, J. Graul, A.V. Mudryi, H. Harima, A. Hashimoto, A. Yamamoto, E.E. Haller, Physica Status Solidi B, 234, 787 (2002). 14. F.-H. Yang, J.-S. Hwang, K.-H. Chen, Y.-J.Yang, T.-H. Lee, L.-G. Hwa and L.-C. Chen, Thin Solid Films 405(1-2), pp. 194-197 (2002). 15. V. Ya. Malakhov, Solar Energy Materials and Solar Cells, 76, 637 (2003). 16. B. Onderka, J. Unland, R. Schmid-Fetzer, J. Mater. Res. 17, 3065-3083 (2002). 17. J. MacChesney, P.M. Bridenbaugh, and P.B. O'Connor, Mater. Res. Bull. 5, 783 (1970). 18. B. H. Cardelino, C. E. Moore, C. A. Cardelino, D. O. Frazier, K. J. Bachmann, J. Phys. Chem. A 105, 849 (2001). 19. N. Dietz, S. McCall, K.J. Bachmann, Proc. Microgravity Conf. 2000, NASA/CP2001-210827, pp. 176-181 (2001). 20. N. Dietz, V. Woods, S. McCall and K.J. Bachmann, Proc. Microgravity Conf. 2002, NASA/CP-2003-212339,pp. 169-181 (2003). 21. N. Dietz, H. Born, M. Strassburg and V. Woods, Mat. Res. Soc. Symp. Proc. 798, Y10.45.1 (2004).

InN Growth by HPCVD: Real-time and Ex-situ Characterization

235

22. N. Dietz, "Real-time optical Characterization of thin film growth," Mat. Sci. & Eng. B 87(1), pp.1 -22 (2001). 23. V. Woods, J. Senawirante, and N. Dietz, J. Vac. Sci. Technol. B 23(4), pp. 17901794 (2005). 24. G. A. Hebner, K. P. Killeen and R. M. Biefeld, J. Crystal Growth, 98(3) pp. 293301 (1989). 25. G. A. Hebner and K. P. Killeen, J. Appl. Phys. 67(3) pp. 1598-1600 (1990). 26. H. Okabe, M. K. Emadi-Babaki and V. R. McCrary, J. Appl. Phys. 69(3), 17301735(1991).. 27. M. C. Johnson, K. Poochinda, N. L. Ricker, J. W. Rogers Jr. and T. P. Pearsall, J. Cryst. Growth 212, 11 (2000). 28. N. Dietz, M. Strassburg and V. Woods, J. Vac. Sci. Technol. A 23(4), pp. 12211227 (2005). 29. J. Haigh and S. O'Brien, J. Crystal Growth 68, 550 (1984). 30. B. H. Cardelino, C. E. Moore, C. A. Cardelino, S. D. McCall, D. O. Frazier, K. J. Bachmann; J. Physical Chemistry A, 107, 3708 (2003). 31. T. Schmidtling, M. Drago, U. W. Pohl and W. Richter, J. Crystal Growth 248, 523 (2003). 32. A. Jain, S. Raghavan and J. M. Redwing, J. Crystal Growth 269, 128 (2004). 33. N. Dietz and K. J. Bachmann, Vacuum 47, 133(1996). 34. C.Park,W.-S. Jung, Z. Huang and T. J. Anderson, J. Mater. Chem. 12, pp. 356-360 (2002). 35. V. Y. Davydov, V. V. Emtsev, I. N. Goncharuk, A. N. Smirnov, V. D. Petrikov, V. V. Mamutin, V. A. Vekshin, S. V. Ivanov, M. B. Smirnov and T. Inushima, Appl. Phys. Lett. 75, 3297 (1999).

This page is intentionally left blank

CHAPTER 7 A NEW LOOK ON InN

Li-Wei Tu*, Ching-Lien Hsiao and Min-Hsiung Tsai Department of Physics and Center for Nanoscience and Nanotechnology, National Sun Yat-Sen University, Kaohsiung, Taiwan 80424, Republic of China * Corresponding author: [email protected] Although the laser head of the next generation DVD and the blue/green and white light emitters are all made from nitrides, hot debate on the fundamental band gap energy of InN has just begun starting from successful growth of high quality materials. In year 2002, several reports showed experimental evidence of a band gap of smaller than 1 eV as compared with 1.9 eV in previous literatures. Values in the range of ~ 0.6 - 0.8 eV are most commonly seen in recent works. This discovery opens up a whole new territory in the potential applications of optoelectronics. Together with other nitrides semiconductors, InN, GaN, and AIN cover a complete energy spectrum from infrared to ultraviolet. But, up to now, there are still so many unresolved important issues from growth to device fabrications. No one can predict what the final scenario will be, yet, more challenges motivate more research activities. In this short article, a brief overview of the current status on InN is given. Critical issues are discussed while mysteries remain to be resolved. 1.

Introduction

Controversy around the recent discovery of a much smaller band gap of InN ignites massive research activities on this semiconductor although InN alloys are already very popular materials used in our daily life. This possible fact of knowing the truth only after extensive usage/application of a material certainly intensifies the surprises and the excitement of this drama. Instead of pure InN, mixed compounds with GaN and AIN are 237

238

Tu, Hsiao, Tsai

commonly employed in the devices to meet the required specifications. Although blue/green light emitting diodes (LEDs) made by InN alloyed with GaN are seen almost everywhere around us for a quite a long time, the true fundamental band gap energy of InN is still under hot debate at this moment. The story begins not long ago. Before year 2002, a value of 1.9 eV is the de facto standard given in a text book and literatures.1^1 In year 2002, Davydov et al., which is the first to report less than 1 eV luminescent emission from InN5' , and some other groups reported a very different band gap value of InN in a range of 0.7 to 1 eV.7"13 Before 2002, there are also reports of 1.1 eV band gap with transmission measurements and of anomalous features in the absorption and photoluminescence (PL) experiments of AlInN.14'15 Even earlier, indication from absorption data was reported.16 The large range of the reported values reflects the initial stage of this research subject and the complexity of the efforts to reach a coherent agreement. The impact of this much smaller energy value is profound. Not to mention that a totally new fact sheet of its fundamental physical properties is expected, a direct impact is to extend the available optical spectrum down to the near infrared (IR) range. Together with GaN (energy gap Eg = 3.4 eV) and A1N (Eg = 6.2 eV), these three nitrides cover a whole electromagnetic wave range from ultraviolet (UV) to IR. In addition, InN possesses the highest steady-state peak drift velocity of 4.2xl0 7 cm/s among common semiconductors17 and this enables applications in high speed devices. Field emission devices18 and 1.3 1.55 um optical communication using In-rich nitrides are explored also. On the other side of the story, there are articles claimed that 0.7 eV near IR emission is associated with surface states at the In/InN interfaces19-21 or with stoichiometry violation22 or with deep traps.23 Latest results reported a band transition value of 1.7 ± 0.2 eV by valence electron energy loss spectroscopy.24 Other recent reports of around 1.9 eV or higher are also seen. " There are several reviews " which cover quite substantially on this exciting field and this short article will give a brief overview of the latest status and examine the current evidences and arguments around this ambiguity following a route of InN growth by plasma-assisted molecular beam epitaxy (PAMBE) and characterizations.

A New Look on InN

239

2. Growth and Structures A major reason of this new discovery is due to the advance in modern technology of epitaxy and the ability of growing high quality samples. Before, most of the samples were from sputtering.1'30 They are basically polycrystalline with a high content of oxygen.27'28 A 12% oxygen containing polycrystalline film grown by the reactive radio frequency (RF) magnetron sputtering gives an optical band gap of 2.0 eV from transmittance spectra.35 With the state-of-the-art epitaxial techniques, like molecular beam epitaxy (MBE)33'36-45 and metal-organic chemical vapor deposition (MOCVD),34'46""50 single crystalline InN with high purity is obtained routinely. Other methods include MEE,51 FCVA,52 etc. This does not mean that the growing process is easy. First, a good matching substrate does not exist in terms of lattice constant and thermal expansion coefficient. Second, the decomposition temperature of InN is low around 630 °C in vacuum and around 500 °C in N2 atmosphere forming In-droplets quickly.53,54 In addition, the two elements of this binary compound sit far from each other in the periodic table with a large difference in their sizes and electronegativities. Sapphire in c-plane and Si in (111) plane are most common substrates used to grow InN for their availability. Sapphire is the primary substrate used in the nitride growth with extensive experience.39'47'50'55-63 Si is much cheaper and has the potential of integration with the existing IC industry.36'60'64"68 Other substrates used are Ge,69 GaAs,70"72 SiC,73 zirconia74 and others.75 Usual ex-situ cleaning processes use hot H2S04/H3P04 mixture for sapphire and diluted HF for Si wafer after standard degrease procedures. After loading the substrate into the MBE chamber, a thermal cleaning is a routine. In-situ monitoring of the epitaxial growth can be executed with the reflection high-energy electron diffraction (RHEED) in an MBE system. For Si substrate, (7x7) reconstructed surface RHEED patterns assure the readiness of the substrate. Figure 1 shows typical RHEED patterns obtained through a 15 kV electron beam with a grazing incidence angle and indicated the smoothness of the InN sample surface. The crystallographic alignments between the film and the substrates are

240

Tu, Hsiao, Tsai

InN//sapphire<10-10> and InN//Si. The InN inplane orientation has a 30° rotation with respect to the sapphire.

Fig. 1. RHEED patterns of epitaxial InN surface with electron beam direction along (a) (11-20) (b) (10-10).

To overcome the substrate mismatch issue, various schemes of buffer layers are studied including i) a single AIN or GaN or low-temperature (LT) InN buffer iayer36'39'56'59-60'65 or ii) double buffer layers like AIN/GaN or LT-InN/GaN or iii) nitrided sapphire.60 Figures 2(a)-2(g) are scanning electron microscopy (SEM) images of PAMBE-grown InN morphology under different growth conditions with a single or dual buffer layer. The general trend for a better epi-InN with a LT-InN buffer layer is to have a lower N/In ratio either in a single LT-InN buffer layer or dual LT-InN/GaN buffer layer structure. The best epitaxial InN film uses a single high-temperature (HT) AIN layer as the buffer and a flat surface is obtained. The vertical scale of the surface topography requires the technique of atomic force microscopy (AFM) which is shown in Fig. 3. Roughness analyses of better samples give a value less than 2 nm.

A New Look on InN

241

242

Tu, Hsiao, Tsai

Fig. 2. SEM surface images of a series InN grown by PAMBE with various buffers: (a) LT-InN at 270 °C, N/In = 100; epi-InN at 420 °C, N/In = 40 (b) LT-InN at 270 °C, N/In = 100; epi-InN at 350 °C, N/In = 40 (c) LT-InN at 270 °C, N/In = 50; epi-InN at 350 °C N/In = 20 (d) LT-InN(270 °C, N/In = 50)/GaN(480 °C, N/Ga = 20); epi-InN at 350 °C,' N/In = 20 (e) LT-InN(270 °C, N/In = 20)/GaN(480 °C, N/Ga = 100); epi-InN at 350 °c' N/In = 20 (f) GaN(670 °C, N/Ga = 20)/AlN(830 °C, N/Al = 20); epi-InN at 350 °C, N/In = 20 and (g) AIN at 830 °C, N/Al = 20; epi-InN at 350 °C, N/In = 20.

Fig. 3. AFM surface topography of InN film showing surface roughness.

Figure 4(a) is a cross-sectional transmission electron microscopy (TEM) micrograph of a PAMBE grown InN sample. A thin AIN buffer layer is deposited on the Si substrate first. The AIN layer is flat, uniform in its thickness. Threading dislocations can be seen and a density in ~ 1010 cm-2 range is common.76 Figure 4(b) is an electron diffraction pattern of an InN layer in transmission mode which shows a single crystalline hexagonal structure with crystallographic c-axis along the growth direction. Indium vacancies probed by low-energy positron beam was identified as having a density profile from ~ 5xl0 18 to below 1016 cm"3 with increasing layer thickness from 120 to 800 nm which is coincided with an increase of electron mobility.77

A New Look on InN

243

Fig. 4. TEM results, (a) Cross-sectional micrograph of an InN sample with a single buffer layer of A1N. (b) Transmission electron diffraction pattern of an InN layer.

For InN growth, low growth temperature in a range of ~ 350 - 550 °C is necessary to avoid decomposition. The downside of low substrate temperature is that it hampers the migration of In adatoms on the surface to facilitate two-dimensional growth. Higher growth temperature yields better crystal quality in terms of narrower full-width-at-half-maximum (FWHM) of x-ray diffraction (XRD) rocking curve while In metal drops may be seen on the sample surface if temperature is too high. MBE technique is more apt to grow crystal at lower temperatures as compared to MOCVD and this may be explored fully to its advantage in the InN growth. A polycrystalline InN grown at non-optimized condition as in Fig. 5(a) can be compared with a good single crystalline InN film on an A1N buffer as in Fig. 5(b) by PAMBE. Both are pure hexagonal structure without cubic signals. The peak around 33° in polycrystalline sample is from InN (10-11) instead of In (101) metal.36,78 A double-crystal rocking curve provides information on the sample structural perfection as in Fig. 6. The narrow FWHM of InN (0002) peak indicates the good perfection. Thicker InN tends to have a better crystal quality with narrower rocking width, lower Hall carrier density and higher mobility.77

244

Tu, Hsiao, Tsai

Si(111) lnN(10-10)

(a)

lnN(0002) lnN(10-11) lnN(10-12):

wm 30

35

40

30

45

35

40

45

2Theta (deg)

2Theta (deg)

Fig. 5. XRD 9-29 scan results: (a) a polycrystalline wurtzite InN and (b) a single crystalline wurtzite InN with A1N buffer layer.

-10000 -5000

0

5000 10000

Omega (arcsec) Fig. 6. XRD rocking curve measurement of InN (0002).

Polarity controlled growth is an interesting subject. With a large inherent spontaneous polarization, crystal polarity is a key factor in designing a properly functional device. Since the method to grow polar GaN films by PAMBE is known, that is, HT-A1N and LT-GaN buffer layers on c-sapphire yields Ga- and N-polar GaN epi-layer, respectively, polar GaN layers are used to control the polarity of the InN layers.79"81 Crystal polarity is decided by RHEED, wet etching,82 co-axial impact collision ion scattering spectroscopy (CAICISS)83 and convergent-beam electron diffraction (CBED).84'85 Different polar InN samples were grown on different polar SiC faces also.73 N-polar InN can be grown at higher temperature than In-polar InN with better quality.86

A New Look on InN

245

InN nanostructures including two-dimensional, one26,90 94 95 98 dimensional - and zero-dimensional structures are explored by several groups.99'100 A slight blue shift in PL peak with a change of InN well thickness from 5.7 nm to 2.3 nm is ascribed to a combination of the quantum size effect (QSE), quantum confined Stark effect (QCSE), and band filling effect.89 Diversity of the band gap values reported is quite dramatic also in nanowire structures.90 PL signals were reported with quite a large range from the 0.7 - 0.8 eV low value side91 to the 1.9 eV92 and to the very high value as 3 eV.93 From a collection of onedimensional nano-structural InN optical results, a 1.8 - 1.9 eV value was reported from a lateral size in a range of 10 - 80 nm, and this large size range can hardly back up the argument of a simple QSE raising a 0.7 eV band gap to a 1.9 eV in size larger than 10 nm.26'90'99 3. Compositions and Electrical Properties To probe the composition of InN, energy dispersive x-ray spectrometer (EDX or EDS), wavelength-dispersive x-ray spectrometer (WDS or sometimes called EPMA for electron probe micro-analysis), Rutherford backscattering spectroscopy (RBS), Auger electron spectroscopy (AES), secondary ion mass spectroscopy (SMS), and x-ray photoelectron spectroscopy (XPS) are common tools to give the In and N atomic content with various degrees of resolutions and limitations. Nonstoichiometry is a possible cause related to the explanation of higher or lower band gap values or the high un-intentional n-type carriers. A conclusion of a N/In > 1 for a higher band gap and N/In < 1 for a lower band gap in explaining 0.7 - 2 eV band gap range based on ~ 25 InN samples collected was drown.101 An ultra-high vacuum (UHV) environment in an MBE chamber is advantageous in its ability to eliminate the possible substantial oxygen content in the samples grown, which is one of the very popular argument of having a high band gap value in previous sputtered samples possibly forming ln203-like compound. A high unintentional, negative carrier density seen in as-grown undoped InN samples is one of the most challenging topics in InN research. A value in the mid 1018 cm"3 is mostly reported among better

246

Tu, Hsiao, Tsai

samples. A few lower values in the 1016 - 1017 cm"3 are reported also.102'103 Higher values in a range of 1019 - 1020 cm'3 causes substantially higher optical transition energy due to Burstein-Moss effect and this is suspected as the reason of having high energy gap value in previous sputtered samples.103 Similarity exists in nitride species as that both as-grown InN and GaN have n-type characteristics without doping them on purpose. The origin of these negatively charged carriers are investigated with great extent. An argument of nitrogen deficiency is frequently heard based on the small incorporation efficiency of nitrogen. Yet, a 2004 paper claimed that the InN samples are really nitrogen rich instead of deficient,27 although MBE grown samples are far more stoichiometric and have negligible oxygen as compared with sputtered samples.28 Another possible impurity discussed other than oxygen is hydrogen. With a transition level introduced by hydrogen lies well above the conduction band minimum, a behavior of shallow donor for hydrogen in InN is predicted and reported.104 Defects are another possible cause due to the fact that the current samples are still far from ideal and possess many imperfections including a high dislocation density. Minute impurities or impurity-defect complex can not be totally excluded to be the source. Although Hall measurements are usually employed due to its convenience to give an average value, more detail analyses are needed to distinguish various components contributing the apparent results. Multiple carrier effects in InN epilayers are investigated using variable magnetic field Hall measurements.105 A surface accumulation layer has been reported106 indicating a surface state density of 2.5xl013 cm"2 and compensation effects are discussed.107 Owing to the possible small band gap, surface excess charges and high carrier density, metals like Ti, Al, Ni, W, WSij, etc. can easily form ohmic contact to InN (n-type) without annealing.108'109 As in the GaN, Si doping is used to increase the electron concentration in the InN film and the results are successful. But, on the formation of p-type InN, results are vague and not successful at this stage. One most critical factor is to further lower down the unintentional n-type carrier concentration before p-type dopant is introduced into the material.

A New Look on InN

247

4. Optical Characterizations Although 1.9 eV was taken to be the energy gap of InN, light emission at 1.9 eV has never been reported before 2002 while other optical measurements, like absorption and Raman, can be seen in literatures with quite a quantity. In 2002, Davydov et al. reported the first InN PL at 0.8 ~ 0.85 eV for samples with n - 9 - 12xl018 cm"3. This near IR light emitted from InN creates great momentum towards achieving more solid facts around this new finding. A typical PL spectrum measured at room temperature is shown in Fig. 7. Proper selection of the detectors and gratings in the visible and the IR range has to be done to ensure the optimization of the signal. Usually, a photomultiplier tube (PMT) is used in the visible range and either an IR-enhanced InGaAs or a PbS detector is used in the IR range. All the samples in Fig. 2, poly- or single-crystals, displayed a unique peak in the IR and no 1.9 eV was observed. PL dependence on temperature110 and laser power66 was performed. It is worth a note that reports from samples grown by the technique of electron cyclotron resonance MBE (ECR-MBE) yield a luminescence around 1.9 eV which may result from possible system-related oxygen inclusion.25'67

n

c 3

.Q

i

i

'

8

i

'

i

lnNonSi(111)

3, 6 •5

4

c

B 2

J:

a! °

i

0.5

•

I

1.0

i

I

i

1.5

L

2.0

2.5

Energy (eV) Fig. 7. Room-temperature PL results of InN epifilm on Si. No prominent peak can be detected in the range of 1.0 - 2.3 eV.

248

Tu, Hsiao, Tsai

Clear blue shift of the PL peak energy with increasing electron concentration is accounted for by the band-filling effect.1U_1 Optical absorption curve shifts towards higher energy side due to the same reasoning.104'114 Hydrostatic pressure studies on PL reveal a much smaller pressure coefficient which is attributed to recombination through localized states.115 A much smaller effective mass is estimated for the magnitude of the blue shift and consistent with a smaller energy gap. Nonparabolicity of the energy band is discussed to account for the strong dependence of the electron effective mass on electron density.116

100 200 300 400 500 600 700

Wavenumber (cm1)

Fig. 8. Raman spectrum exhibits three wurtzite InN phonon modes in a back-scattering configuration.

Raman spectroscopy reveals the hexagonal phase of the InN crystal straightly.117 A typical Raman spectrum is in Fig. 8 which exhibits E2(low), E2(high) and Ai(LO) modes of a hexagonal InN phase in a back-scattering configuration with the laser along the c-axis. Strain analyses and carrier concentration correlation with Raman modes were reported.118 The possibility of a 1.9 eV band gap was claimed to be excluded by resonant Raman spectroscopy.119 An electron effective mass of 0.085mo for intrinsic InN was estimated from the plasma frequency derived from infrared reflectance measurements which showed a carrier dependent effective mass which means a nonparabolic band. InN annealed in oxygen environment showed a broad peak entered at 1.4 eV

A New Look on InN

249

and ln 2 0 3 phase was seen in Raman spectra. Ultrafast Raman spectroscopy measurements yield a transient velocity as high as 2xl0 8 cm/s.122 Time-resolved differential transmission measurements establish an inverse proportional relation between the carrier lifetime and the carrier density which suggests that the donor-like defects or impurities may stimulate a formation of non-radiative recombination centers.123 5. Four Possibilities To summarize to a minimum from massive, seemly contradictory information, four possibilities exist logistically for the true band gap energy of wurtzite single crystalline InN, that is, i) it is 0.7 eV, ii) it is 1.9 eV, iii) it is 0.7 and 1.9 eV, and iv) It is not 0.7 or 1.9 eV. A concise view of each possibility is as following. i) It is 0.7 eV and the reasons that a 1.9 eV value is obtained are a) Burstein-Moss effect,1 b) quantum size effect, and c) non-stoichiometry. For a), either a very large electron density of about > 5xl0 20 cm*3 or/and a fairly small electron effective mass will be required to possibly cause enough blue shift in energy.116 For b), a rather small InN nano-crystal uniformly in a size of ~ 5 nm is needed for a large size effect.92 A small effective mass will enlarge the effect and a large enough potential barrier will be necessary to confine the carriers. Polycrystalline structure does not constitute a strong enough statement for QSE of a 1.9 eV band gap value since polycrystalline InN does emit at 0.7 - 0.8 eV as seen in Fig. 5 and Fig. 7. For c), an N-rich InN sample with ~ 30% more N over In may cause a large optical band gap.22 ii) It is 1.9 eV and the reasons for a 0.7 eV measured are a) deep level, b) In nano-cluster, c) non-stoichiometry, and (d) band tailing. For a), deep traps with high radiative quantum efficiency will be needed created by either defects or impurities. For b), In metal nano-clusters uniform in a certain way will be needed to scatter the light through Mie resonance or to cause local potential dips. For c), an In-rich InN with a N/In ratio of ~ 0.8 may reduce the band gap to 0.7 eV.22 For (d), a highly degenerate sample may possibly shrink the gap to a certain extent through band tailing deep into the gap.

250

Tu, Hsiao, Tsai

iii) It is 0.7 and 1.9 eV. By definition, band gap is a single value, so that this is not physically valid. This is just to contemplate a scenario of a 0.7 eV gap at Y with a higher 1.9 eV conduction band minimum at T in k space or with another conduction band minimum at other k point with a value of 1.9 eV measured from the top of the valence band at T. Or, another scenario is to have a 1.9 eV gap at T with another conduction band minimum at other k point with a value of 0.7 eV measured from the top of the valence band at I\ The full band structure is yet to comply with this pure hypothesis. Base on the first-principles simulation, this scenario is not supported.116 iv) It is not 0.7 or 1.9 eV. The 1.9 eV may from other materials like InNxOy114 and 0.7 eV is from unknown origins like an impurity-defect complex. The true value may fall into a range of 1.1 - 1.5 eV for a real stoichiometric perfect sample.14'23'31 6. Conclusions and Outlook The controversy about the band gap energy of InN makes it inevitably attractive. At the current stage, more evidences will be needed before a conclusive remark can be made. We are witnessing the progressive history of an interesting scientific event and we are making the history ourselves. We do not want to miss the fun and excitement and hard works will pay in the end. The puzzle of either ~ 0.7 eV is the ultimate truth of the real band gap or a deep level or else is yet to be resolved. Samples with less unintentional carriers are desirable. A happy ending is foreseeable in the near future. Acknowledgments This work is supported by the National Science Council of Taiwan, Republic of China. References 1. T. L. Tansley and C. P. Foley, J. Appl. Phys. 59, 3241 (1986). 2. Q. Guo and A. Yoshida, Jpn. J. Appl. Phys. 33, 2453 (1994).

A New Look on InN

251

3. O. Madelung Ed., Semiconductors: Group IV Elements and III-V Compounds (Springer-Verlag, Berlin, 1991), p. 122. 4. V. V. Mamutin, T. V. Shubina, V. A. Vekshin, V. V. Ratnikov, A. A. Toropov, S.V. Ivanov, M. Karlsteen, U. Sodervall and M. Willander, Appl. Surf. Sci. 166, 87 (2000). 5. V. Y. Davydov, A. A. Klochikhin, R. P. Seisyan, V. V. Emtsev, S. V. Ivanov, F. Bechstedt, J. Furthmuller, H. Harima, A. V. Mudryi, J. Aderhold, O. Semchinova and J. Graul, Phys. Stat. Sol. B 229, Rl (2002). 6. V. Y. Davydov, A. A. Klochikhin, V. V. Emtsev, S. V. Ivanov, V. V. Vekshin, F. Bechstedt, J. Furthmuller, H. Harima, A. V. Mudryi, A. Hashimoto, A. Yamamoto, J. Aderhold, J. Graul and E. E. Haller, Phys. Stat. Sol. B 230, R4 (2002). 7. J. Wu, W. Walukiewicz, K. M. Yu, J. W. Ager III, E. E. Haller, H. Lu, W. J. Schaff, Y. Saito and Y. Nanishi, Appl. Phys. Lett. 80, 3967 (2002). 8. J. Wu, W. Walukiewicz, K. M. Yu, J. W. Ager III, E. E. Haller, H. Lu and W J. Schaff, Appl. Phys. Lett. 80,4741 (2002). 9. T. Matsuoka, H. Okamoto, M. Nakao, H. Harima and E. Kurimoto, Appl. Phys. Lett. 81,1246 (2002). 10. M. Hori, K. Kano, T. Yamaguchi, Y. Saito, T. Araki, Y. Nanishi, N. Teraguchi and A. Suzuki, Phys. Stat. Sol. B 234, 750 (2002). 11. V. Y. Davydov, A. A. Klochikhin, V. V. Emtsev, D. A. Kurdyukov, S. V. Ivanov, V. A. Vekshin, F. Bechstedt, J. Furthmuller, J. Aderhold, J. Graul, A. V. Mudryi, H. Harima, A. Hashimoto, A. Yamamoto and E. E. Haller, Phys. Stat. Sol. B 234, 787 (2002). 12. Y. Saito, H. Harima, E. Kurimoto, T. Yamaguchi, N. Teraguchi, A. Suzuki, T. Araki and Y. Nanishi, Phys. Stat. Sol. B 234, 796 (2002). 13. J. Wu, W. Walukiewicz, W Shan, K. M. Yu, J. W. Ager III, E. E. Haller, H. Lu and W. J. Schaff, Phys. Rev. B 66, 201403 (2002). 14. T. Inushima, V. V. Mamutin, V. A. Vekshin, S. V. Ivanov, T. Sakon, M. Motokawa and S. Ohoya, J. Cryst. Growth 227-228,481 (2001). 15. S. Yamaguchi, M. Kariya, S. Nitta, T. Takeuchi, C. Wetzel, H. Amano and I. Akasaki, Appl. Phys. Lett. 76, 876 (2000). 16. T. Matsuoka, H. Tanaka, T. Sasaki and A. Katsui, Proceedings of the Sixteenth International Symposium on GaAs and Related Compounds, Karuizawa, Japan, 2529 September 1989 (Institute of Physics, Bristol, 1990), p. 141. 17. B. E. Foutz, Stephen K. O'Leary, M. S. Shur and L. F. Eastman, J. Appl. Phys. 85, 7727 (1999). 18. C. F. Shih, N. C. Chen, P. H. Chang and K. S. Liu, J. Cryst. Growth 281, 328 (2005) 19. T. V. Shubina, S. V Ivanov, V. N. Jmerik, D. D. Solnyshkov, V. A. Vekshin, P. S. Kop'ev, A. Vasson, J. Leymarie, A. Kavokin, H. Amano, K. Shimono, A. Kasic and B. Monemar, Phys. Rev. Lett. 92,117407 (2004). 20. F. Bechstedt, J. Furthmuller, O. Ambacher and R. Goldhahn, Phys. Rev. Lett. 93, 269701 (2004).

252

Tu, Hsiao, Tsai

21. T. V. Shubina, S. V. Ivanov, V. N. Jmerik, P. S. Kop'ev, A. Vasson, J. Leymarie, A. Kavokin, H. Amano, B. Gil, O. Briot and B. Monemar, Phys. Rev. Lett. 93 269702 (2004). 22. T. V. Shubina, S. V. Ivanov, V. N. Jmerik, M. M. Glazov, A. P. Kalvarskii, M. G Tkachman, A. Vasson, J. Leymarie, A. Kavokin, H. Amano, I. Akasaki, K. S. A. Butcher, Q. Guo, B. Monemar and P. S. Kop'ev, Phys. Stat. Sol. A 202, 377 (2005). 23. O. Briot, B. Maleyre, S. Ruffenach, B. Gil, C. Pinquier, F. Demangeot and J. Frandon, J. Cryst. Growth 269 (2004) 22. 24. P. Specht, J. C. Ho, X. Xu, R. Armitage, E. R. Weber, R. Erni and C. Kisielowski, Solid State Commun. 135, 340 (2005). 25. T. Yodo, Y. Kitayama, K. Miyaki, H. Yona, Y. Harada, K. E. Prince, K. Scott and A. Butcher, J. Cryst. Growth 269, 145 (2004). 26. S. Luo, W. Zhou, Z. Zhang, X. Dou, L. Liu, X. Zhao, D. Liu, L. Song, Y. Xiang, J. Zhou and S. Xie, Chem. Phys. Lett. 411, 361 (2005). 27. K. S. A. Butcher, M. Wintrebert-Fouquet, P. P.-T. Chen, T. L. Tansley, H. Dou, S. K. Shrestha, H. Timmers, M. Kuball, K. E. Prince and J. E. Bradby, J. Appl. Phys. 95, 6124 (2004). 28. S. K. Shrestha, H. Timmers, K. S. A. Butcher, M. Wintrebert-Fouquet and P. P.-T. Chen, Nucl. Instr. and Meth. in Phys. Res. B 234, 291 (2005). 29. A. G Bhuiyan, A. Hashimoto and A. Yamamoto, J. Appl. Phys. 94, 2779 (2003). 30. J. Wu and W. Walukiewicz, Superlatt. Microstruct. 34, 63 (2003). 31. K. S. A. Butcher and T. L. Tansley, Superlatt. Microstruct. 38, 1 (2005). 32. B. Monemar, P. P. Paskova and A. Kasic, Superlatt. Microstruct. 38, 38 (2005). 33. X. Wang and A. Yoshikawa, Prog. Cryst. Growth Character. Mater. 48-49 42 (2004). 34. T. Matsuoka, Superlatt. Microstruct. 37, 19 (2005). 35. Y. Yi, S. Cho, Y. Roh, M. Noh, C-N Whang, K. Jeong and H-J Shin, Jpn. J. Appl. Phys. 44, 17 (2005). 36. C. L. Hsiao, L. W. Tu, M. Chen, Z. W. Jiang, N. W. Fan, Y. J. Tu and K. R. Wang, Jpn. J. Appl. Phys. 44, L1076 (2005). 37. Y. Nanishi, Y. Saito, T. Tamaaguchi, M. Hori, F. Matsuda, T. Araki, A. Suzuki and T. Miyajima, Phys. Stat. Sol. A 200, 202 (2003). 38. Y. Nanishi, Y. Saito and T. Yamaguchi, Jpn. J. Appl. Phys. 42, 2549 (2003). 39. Th. Kehagias, E. Iliopoulos, A. Delimitis, G. Nouet, E. Dimakis, A. Georgakilas and Ph. Komninou, Phys. Stat. Sol. A 202, 777 (2005). 40. Y. F. Ng, Y. G. Cao, M. H. Xie, X. L. Wang and S. Y. Tong, Appl. Phys. Lett. 81, 3960 (2002). 41. Y. Saito, N. Teraguchi, A. Suzuki, T. Araki and Y. Nanishi, Jpn. J. Appl. Phys 40, L91 (2001). 42. R. J. Kinsey, P. A. Anderson, C. E. Kendrick, R. J. Reeves and S. M. Durbin, J. Cryst. Growth 269, 167 (2004). 43. T. Araki, Y. Saito, T. Yamaguchi, M. Kurouchi, Y. Nanishi and H. Naoi, J. Vac. Sci. Technol. B 22, 2139 (2004).

A New Look on InN

253

44. P. Specht, R. Armitage, J. Ho, E. Gunawan, Q. Yang, X. Xu, C. Kisielowski and E. R. Weber, J. Cryst. Growth 269, 111 (2004). 45. K. M. Yu, Z. Liliental-Weber, W. Walukiewicz, W. Shan, J. W Ager III, S. X. Li, R. E. Jones, E. E. Haller, H. Lu and W. J. Schaff, Appl. Phys. Lett. 86, 071910 (2005). 46. Y. Huang, H. Wang, Q. Sun, J. Chen, D. Y. Li, Y. T. Wang and H. Yang, J. Cryst. Growth 276,13(2005). 47. Z. X. Bi, R. Zhang, Z. L. Xie, X. Q. Xiu, Y. D. Ye, B. Liu, S. L. Gu, B. Shen, Y. Shi and Y. D. Zheng, Mater. Lett. 58,3641 (2004). 48. B. Maleyre, S. Ruffenach, O. Briot, B. Gil and A. Van der Lee, Superlatt. Microstruct. 36, 517 (2004). 49. P. Singh, P. Ruterana, M. Morales, F. Goubilleau, M. Wojdak, J. F. Carlin, M. Ilegems and D. Chateigner, Superlatt. Microstruct. 36, 537 (2004). 50. A. G. Bhuiyan, T. Tanaka, K. Kasashima, A. Hashimoto and A. Yamamoto, Jpn. J. Appl. Phys. 42, 7284 (2003). 51. H. Lu, W. J. Schaff, J. Hwang, H. Wu, W. Yeo, A. Pharkya and L. F. Eastman, Appl. Phys. Lett. 77, 2548 (2000). 52. X. H. Ji and S. P. Lau, J. Cryst. Growth 282, 271 (2005). 53. O. Ambacher, M. S. Brandt, R. Dimitrov, T. Metzger, M. Stutzmann, R. A. Fischer, A. Miehr, A. Bergmaier and G Dollinger, J. Vac. Sci. Technol. B 14, 3532 (1996). 54. Y. Huang, H. Wang, Q. Sun, J. Chen, J. F. Wang, Y. T. Wang and H. Yang, J. Cryst. Growth 281, 310(2005). 55. M. Kurouchi, T. Araki, H. Naoi, T. Yamaguchi, A. Suzuki and Y. Nanishi, Phys. Stat. Sol. B 241, 2843 (2004). 56. C. J. Lu, L. A. Bendersky, H. Lu and W. J. Schaff, Appl. Phys. Lett. 83, 2817 (2003). 57. J. Jasinski, Z. Liliental-Weber, H. Lu and W. J. Schaff, Appl. Phys. Lett. 85, 233 (2004). 58. H. Xiao, X. Wang, J. Wang, N. Zhang, H. Liu, Y. Zeng, J. Li and Z. Wang, J. Cryst. Growth 276, 401 (2005). 59. B. Liu, T. Kitajima, D. Chen and S. R. Leone, J. Vac. Sci. Technol. A 23, 304 (2005). 60. B. Maleyre, S. Ruffenach, O. Briot, B. Gil and A. V. D. Lee, Superlatt. Microstruct. 36,517(2004). 61. K. Fujiwara, A. Ishiia, J. Ohta, H. Fujioka and M. Oshima, Thin Solid Films 464465, 112(2004). 62. S. V. Ivanov, T. V. Shubina, V. N. Jmerik, V. A. Vekshin, P. S. Kopev and B. Monemar, J. Cryst. Growth 269,1 (2004). 63. T. Yodo, H. Yona, Y. Harada, A. Sasaki and M. Yoshimoto, Phys. Stat. Sol. C 0, 2802 (2003). 64. T. Yamaguchi, Y. Saito, C. Morioka, K. Yorozu, T. Araki, A. Suzuki and Y. Nanishi, Phys. Stat. Sol. B 240,429 (2003). 65. J. Grandal and M. A. Sanchez-Garcia, J. Cryst. Growth 278, 373 (2005). 66. S. Gwo, C. L. Wu, C. H. Shen, W. H. Chang, T. M. Hsu , J. S. Wang and J. T. Hsu, Appl. Phys. Lett. 84, 3765 (2004).

254

Tu, Hsiao, Tsai

67. T. Yodo, H. Yona, H. Ando, D. Nosei and Y. Harada, Appl. Phys. Lett. 80, 968 (2002). 68. T. Yodo, Y. Kitayama, K. Miyaki, H. Yona and Y. Harada, Jpn. J. Appl. Phys. 43, LI39 (2004). 69. E. Trybusa, G Namkoonga, W. Hendersona, W. A. Doolittlea, R. Liub, J. Meib, F. Ponceb, M. Cheungc, F. Chenc, M. Furisc and A. Cartwright, J. Cryst. Growth 279, 311 (2005). 70. G Kaczmarczyk, A. Kaschner, S. Reich, A. Hoffmann, C. Thomsen, D. J. As, A. P. Lima, D. Schikora and K. Lischka, Appl. Phys. Lett. 76, 2122 (2000). 71. Q. Guo, M. Nishio, H. Ogawa and A. Yoshida, Jpn. J. Appl. Phys. 38, L490 (1999). 72. Z. G Qian, W. Z. Shen, H. Ogawa and Q. X. Guo, J. Appl. Phys. 93, 2643 (2003). 73. F. Matsuda, Y. Saito, T. Muramatsu, T, Yamaguchi, Y Matsuo, A. Koukitu, T. Araki and Y Nanishi, Phys. Stat. Sol. C 0, 2810 (2003). 74. T. Honke, H. Fujioka, J. Ohta and M. Oshima, J. Vac. Sci. Technol. A 22, 2487 (2004). 75. H. Lu, W. J. Schaff, L. F. Eastman, J. Wu, W. Walukiewicz, V. Cimalla and O. Ambacher, Appl. Phys. Lett. 83, 1136 (2003). 76. C. J. Lu, L. A. Bendersky, H. Lu and W. J. Schaff, Appl. Phys. Lett. 83, 2817 (2003). 77. J. Oila, A. Kemppinen, A. Laakso, K. Saarinen, W. Egger, L. Liszkay, P. Sperr, H. Lu and W. J. Schaff, Appl. Phys. Lett. 84,1486 (2004). 78. K. M. Yu, Z. Liliental-Weber, W. Walukiewicz, W. Shan, J. W. Ager III, S. X. Li, R. E. Jones, E. E. Haller, H. Lu and W. J. Schaff, Appl. Phys. Lett. 86, 071910 (2005). 79. E. Dimakis, K. Tsagaraki, E. Iliopoulos, P. Komninou, T. Kehagias, A. Delimitis and A. Georgakilas, J. Cryst. Growth 278, 367 (2005). 80. E. Dimakis, E. Iliopoulos, K. Tsagaraki, T. Kehagias, P. Komninou and A. Georgakilas, J. Appl. Phys. 97, 113520 (2005). 81. E. Dimakis, G Konstantinidis, K. Tsagaraki, A. Adikimenakis, E. Iliopoulos and A. Georgakilas, Superlatt. Microstruct. 36, 497 (2004). 82. D. Muto, T. Araki, H. Naoi, F. Matsuda and Y. Nanishi, Phys. Stat. Sol. A 202, 773 (2005). 83. Y. Saito, Y Tanabe, T. Yamaguchi, N. Teraguchi, A. Suzuki, T. Araki and Y Nanishi, Phys. Stat. Sol. B 228, 13 (2001). 84. T. Mitate, S. Mizuno, H. Takahata, R. Kakegawa, T. Matsuoka and N. Kuwano, Appl. Phys. Lett. 86, 134103 (2005). 85. H. Naoia, F. Matsudab, T. Arakib, A. Suzukic and Y. Nanishi, J. Cryst. Growth 269, 155 (2004). 86. K. Xu and A. Yoshikawa, Appl. Phys. Lett. 83, 251 (2003). 87. H. K. Choa, D. C. Kima, B. H. Konga, K. H. Leeb, J. T. Ohb, S. Kimb, D. J. Kimb, J. W. Kimb and S. K. Hong, J. Cryst. Growth 281, 349 (2005). 88. S. B. Che, W. Terashima, Y Ishitani, A. Yoshikawa, T. Matsuda, H. Ishii and S. Yoshida, Appl. Phys. Lett. 86, 261903 (2005).

A New Look on InN

255

89. M. Kurouchi, H. Naoi, T. Araki, T. Miyajima and Y. Nanishi, Jpn. J. Appl. Phys. 44, L230 (2005). 90. Z. H. Lan, W. M. Wang, C. L. Sun, S. C. Shi, C. W. Hsu, T. T. Chen, K. H. Chen, C. C. Chen, Y. F. Chen and L. C. Chen, J. Cryst. Growth 269, 87 (2004). 91. M. C. Johnson, C. J. Lee, E. D. Bourret-Courchesne, S. L. Konsek, S. Aloni, W. Q. Han and A. Zettl, Appl. Phys. Lett. 85, 5670 (2004). 92. C. H. Liang, L. C. Chen, J. S. Hwang, K. H. Chen, Y. T. Hung and Y. F. Chen, Appl. Phys. Lett. 81, 22 (2002). 93. J. Zhang, B. Xu, F. Jiang, Y. Yang and J. Li, Phys. Lett. A 337, 121 (2005). 94. K. Takemoto, N. Takahashi and T. Nakamura, Sol. Stat. Commun. 134, 617 (2005). 95. A. Yoshikawa, N. Hashimoto, N. Kikukawa, S. B. Che and Y. Ishitani, Appl. Phys. Lett. 86, 153115(2005). 96. B. Maleyre, O. Briot and S. Ruffenach, J. Cryst. Growth 269, 15 (2004). 97. O. Briot, B. Maleyre, S. Ruffenach, B. Gil, C. Pinquier, F. Demangeot and J. Frandon, J. Cryst. Growth 269, 22 (2004). 98. C. F. Shih, N. C. Chen, P. H. Chang and K. S. Liu, J. Cryst. Growth 281, 328 (2005). 99. K. Sardar, F. L. Deepak, A. Govindaraj, M. M. Seikh and C. N. R. Rao, Small 1, 91 (2005). 100. K. Sardar, M. Dan, B. Schwenzer and C. N. R. Rao, J. Mater. Chem. 15, 2175 (2005). 101. T. V. Shubina, S. V. Ivanov, V. N. Jmerik, M. M. Glazov, A. P. Kalvarskii, M. G Tkachman, A. Vasson, J. Leymarie, A. Kavokin, H. Amano, I. Akasaki, K. S. A. Butcher, Q. Guo, B. Monemar and P. S. Kop'ev, Phys. Stat. Sol. A 202, 377 (2005). 102. T. L. Tansley and C. P. Foley, Electron. Lett. 20, 1066 (1984). 103. J. Wu, W. Walukiewicz, S. X. Li, R. Armitage, J. C. Ho, E. R. Weber, E. E. Haller, H. Lu, W. J. Schaff, A. Barcz and R. Jakiela, Appl. Phys. Lett. 84, 2805 (2004). 104. E. A. Davis, S. F. J. Cox, R. L. Lichti and C. G Van de Walle, Appl. Phys. Lett. 82, 592 (2003). 105. C. H. Swartz, R. P. Tompkins, N. C. Giles, T. H. Myers, H. Lu, W. J. Schaff and L. F. Eastman, J. Cryst. Growth 269, 29 (2004). 106. I. Mahboob, T. D. Veal, C. F. McConville, H. Lu and W. J. Schaff, Phys. Rev. Lett. 92, 036804 (2004). 107. D. C. Look, H. Lu, W. J. Schaff, J. Jasinski and Z. Liliental-Weber, Appl. Phys. Lett. 80, 258 (2002). 108. H. Lu, W. J. Schaff, L. F. Eastmand and C. E. Stutz, Appl. Phys. Lett. 82, 1736 (2003). 109. C. B. Vartuli, S. J. Pearton, C. R. Abemathy, J. D. MacKenzie, R. J. Shul, J. C. Zolper, M. L. Lovejoy, A. G Baca and M. Hagerott-Crawford, J. Vac. Sci. Technol. B 14, 3520 (1996). 110. F. Chena, A. N. Cartwrighta, H. Lu and W. J. Schaff, Physica E 20, 308 (2004). 111. M. Higashiwaki, T. Inushima and T. Matsui, Phys. Stat. Sol. B 240, 417 (2003). 112. M. Higashiwaki and T. Matsui, J. Cryst. Growth 269, 162 (2004).

256

Tu, Hsiao, Tsai

113. V. M. Naik, R. Naik, D. B. Haddad, J. S. Thakur, G W. Auner, H. Lu and W. J. Schaff, Appl. Phys. Lett. 86, 201913 (2005). 114. A. G Bhuiyan, K. Sugita, K. Kasashima, A. Hashimoto, A. Yamamoto and V. Y Davydov, Appl. Phys. Lett. 83, 4788 (2003). 115. S. X. Li, J. Wu, W. Walukiewicz, W. Shan, E. E. Haller, H. Lu, W. J. Schaff, Y. Saito and Y. Nanishi, Phys. Stat. Sol. B 241, 3107 (2004). 116. First-principles calculations by Min-Hsiung Tsai, unpublished. 117. G Kaczmarczyk, A. Kaschner, S. Reich, A. Hoffmann, C. Thomsen, D. J. As, A. P. Lima, D. Schikora, K. Lischka, R. Averbeck and H. Riechert, Appl. Phys. Lett. 76, 2122 (2000). 118. V. Darakchieva, P. P. Paskov, E. Valcheva, T. Paskova, M. Schubert, C. Bundesmann, H. Lu, W. J. Schaff and B. Monemara, Superlatt. Microstruct. 36, 573 (2004). 119. M. Kuball, J. W. Pomeroy, M. Wintrebert-Fouquet, K. S. A. Butcher, H. Lu, W. J. Schaff, T. V. Shubina, S. V. Ivanov, A. Vasson and J. Leymarie, Phys. Stat. Sol. A 202, 763 (2005). 120. T. Inushima, M. Higashiwaki and T. Matsui, Phys. Rev. B 68, 235204 (2003). 121. V. Y Davydov, A. A. Klochikhin, V. V. Emtsev, A. N. Smirnov, I. N. Goncharuk, A. V. Sakharov, D. A. Kurdyukov, M. V. Baidakova, V. A. Vekshin, S. V. Ivanov, J. Aderhold, J. Graul, A. Hashimoto and A. Yamamoto, Phys. Stat. Sol. B 240, 425 (2003). 122. K. T. Tsen, C. Poweleit, D. K. Ferry, H. Lu and W. J. Schaff, Appl. Phys. Lett. 86, 222103 (2005). 123. F. Chen, A. N. Cartwright, H. Lu and W. J. Schaff, J. Cryst. Growth 269, 10 (2004).

CHAPTER 8 GROWTH AND ELECTRICAL/OPTICAL PROPERTIES OF AlxGai.xN IN THE FULL COMPOSITION RANGE

Feng Yun Department of Electrical and Computer Engineering Virginia Commonwealth University Richmond, Virginia 23284, USA E-mail: [email protected] The material development of ternary alloy, AlGaN, is key to GaNbased device applications. This chapter includes a technical review of the key issues such as growth techniques suitable for AlGaN, microstructural analyses for extended defects and cracks accompanied with growth on foreign substrates, and the electrical and optical properties of AlGaN epitaxial films. Growth methods for the full range control of AI mole fraction in AlGaN have been introduced, together with a detailed discussion of bowing parameters in association with the determination of alloy band gap as a function of Al mole fraction.

1. Introduction The wide band gap material GaN and its ternary alloys AlGaN and InGaN are direct band gap semiconductors with wurtzite structure. Driven by the paramount interest in device applications, the research of GaN and its ternary alloys has been very successful as advanced materials for modern technology, owing to their unique material characteristics that surpass Si and GaAs-based materials. Table I listed the major physical parameters for GaN, A1N, and InN. These materials and their ternary alloys are characterized by high breakdown voltage, high temperature operation, and mechanical and chemical robustness.

257

F. Yun

258

The potential of Ill-nitride semiconductors (Ga, Al, In)N for use in high power, high frequency transistors has been well documented.1' AlGaN is an important electronic material key to the fabrication of AlGaN/GaN heterostructures and two dimensional electron gas (2DEG) which form the basis of modulation doping with enhanced mobility and thus high frequency operation. The AlGaN/GaN modulation-doped field effect transistor (MODFET) exhibits superior performance featuring high-power density, raised efficiency and bandwidth, increased lifetime of modules, and reduced cost of systems with reduced circuit complexity. For example, power densities ranging up to 10.7 W/mm at 10 GHz have been achieved with a cutoff frequency of fT= 121 GHz and a max frequency of fmax=162 GHz. Optimizization of the electronic properites of the Table I. Physical parameters of GaN and A1N (after B. E. Foutz, from website http://iiiv.tn.cornell.edu/www/foutz/nitride.html). Crystal structure Density (g/cm ) Transverse constant (Q) (dyn/cm2) Longitudinal constant (Q) (dyn/cm2) Transverse sound velocity (cm/s) Longitudinal sound velocity (cm/s) Static dielectric constant High frequency dielectric constant Energy gap (T valley) (eV) Electron effective mass (T valley) (trie) Deformation potential (eV) Polar optical phonon energy (meV) Intervalley coupling coefficient (eV) Intervalley deformation potential (eV/cm) Lattice constant, a (A) Lattice constant, c (A) Electron mobility (cm2/Vs) Hall Mobility (cmVVs) Saturation velocity (cm/s) Peak velocity (cm/s) Peak velocity field (kV/cm) Thermal conductivity (W/cmK) Melting temp (°C)

GaN (Wurtzite) 6.15 4.42x10' 2.65x10' 2.68xl05 6.56X103 8.9 5.35 3.39 0.20 8.3 91.2 91.2 lxlO 9 3.189 5.185 >1100 (bulk) 30 2.5xl0 7 3.1xl07 150 1.5 >1700

A1N (Wurtzite) 3.23 4.42x10* 2.65x10' 3.70xl05 9.06x10' 8.5 4.77 6.2 0.48 9.5 99.2 99.2 1X109 3.11 4.98 135 14 1.4xl07 1.7xl07 450 2 3000

Growth and Electrical/Optical Properties ofAlJlGa1.Jlf

259

AlGaN/GaN hetrojunction are pivatol to the performance of such devices, the key to which turns out to be a precise control of the Al content in the AlGaN layer, and doping in the AlGaN layer. AlGaN is also an important photonic material. In modern materials sciences, the production of light emitting devices in a wide portion of the optical spectrum is expected to have numerous practical applications. GaN and its In and Al alloys span most of the visible region and extend well into the ultraviolet wavelengths. This opens up a whole new region of the optical spectrum to semiconductor technology in much the same way that infrared and longer visible wavelengths are presently exploited by GaAs technology. The AlGaN alloy in particular is uniquely suited for fabricating optoelectronic devices in the ultraviolet and visible bands of the spectrum, e.g. solar-blind photodetectors, light emitting diodes (LED) and laser diodes (LD), operative in the green, blue, and ultraviolet range. With the advent in material growth of high-quality AlGaN ternary alloys, AlGaN-based solar-blind photodetectors emerged as a potential alternative for the photomultiplier tube and Si-based solar-blind detector technology. They have the advantage of intrinsic solar-blindness, and as such, do not need complex and costly filters. In addition, AlGaN-based solar-blind detectors can operate under harsh conditions due to their wide band gaps and robust material properties. AlGaN forms the barrier for all AlGaN/GaN heterojunctions and quantum-well (QW) devices and determines such properties as carrier and light confinement and sheet carrier density. With the recent advent of short wavelength light sources and photodetectors, high Al composition AlGaN is increasingly taking the role of active emission and absorption medium.3 Though a number of these devices have been reported, there are still serious issues to be addressed. High Al mole fraction is difficulty to grow, mostly due to the increased alloy scattering, increased resistivity, and phase segregation. This is on top of the other challenges common to the production of device quality nitride materials, namely, large lattice mismatch between nitride films and substrates, high n-type background concentration, and difficulty in p-type doping. These growth issues result in the deterioration of radiative recombination efficiency with increasing Al content.4 Another critical issue is the control of Al mole fraction. The influence of growth conditions on the Al mole fraction and the quality of

260

F. Yun

AljGa/.jN film have been studied in some detail for organometallic vapor phase epitaxy (OMVPE)5'6'7 In the case of molecular beam epitaxy (MBE), however, the process for AlxGa.].xN is not well understood, though some regularities of the growth was established using electron cyclotron resonance-microwave plasma as a nitrogen source.8 Nowadays, low dimensional and nanometer-sized semiconductors have emerged as important class of material due to the possibility of developing novel devices utilizing quantum confinement effects, which will likely open opportunities for quantum lasers, single electron transistors, and a host of other applications.9 Strong exciton confinement effects are expected when the crystal size value is lower than the exciton effective Bohr radius (generally around 10 nm). l o n The growth and fabrication of AlGaN nanoparticles12 has just started and if successful, will warrant novel device structures for the nitride family. Given the paramount interest in the ternary alloy, the properties of AlGaN have been mostly reviewed within the context of GaN and A1N, or more closely with AlGaN/GaN heterostructures. In this chapter, the growth techniques commonly used for AlGaN will be critically reviewed, with both n-type and p-type conductivity, expanding upon the control of Al content in the full range covering the two binary extremes. We will review in detail the structural properties, defect control, and strain and morphology analyses of the AlGaN films prepared by different techniques. The electrical properties will be described in terms of electrical transport for both the AlGaN material itself and with the formation of the ubiquitous AlGaN/GaN 2DEG. Optical properties of AlGaN will be reviewed to understand the roles of impurities and defects. Another important issue addressed here is the energy band gap bowing in the ternary alloy. An equally important but more complicated quaternary alloy, InAlGaN,13,14 will not be covered within the scope of this chapter. 2. Material Growth of AlGaN 2.1. Growth Techniques and Full Range Control of Al Hydride Vapor Phase Epitaxy (HVPE) is basically a carbon-free technology compared to other CVD processes. The gaseous hydrogen

Growth and Electrical/Optical Properties ofAlxGai.xN

261

chloride (HC1) used in the HVPE process provides an impurity-cleaning effect during growth. In the early days, Si and O contamination was a concern for HVPE-grown nitride layers, but later studies clarified that they are not responsible for n-type doping, rather the nitrogen vacancies are more likely associated with the donor-like conductivity15 Although HVPE was traditionally deemed not suitable for thin layer growth (less than 50 nm), it has several advantages over other techniques such as OMVPE and MBE. They are, to name a few, higher growth rates, lower growth costs, and the demonstrated capability to grow thick films which consequently reduce defect density. Historically, HVPE was the earliest technique to produce highquality GaN epitaxial layers,16 and later produced freestanding GaN wafers with superior quality17'18 AlGaN alloy was grown by Baranov et al.19 in the 1970s, on sapphire substrates by HVPE using a mixture of GaCl, A1C1, and ammonia (NH3). The Al composition of x=0.45 was achieved with an n-type carrier concentration in the low 1019 cm"3. Later on, Nicolaev et al.20 achieved high quality AlGaN alloy using an enhanced HVPE technique. The growth of both n-type21 and p-type20 AlGaN has been made available utilizing HVPE technique. Recently, the same group demonstrated an all-HVPE LED22 structure with emission at 341 nm, following the report of a submicron-thick multilayer growth.23 More recently, the complete structure of a high electron mobility transistor (HEMT) has also been successfully grown by HVPE24 with a 30 nm Alo.22Gao.78N unintentionally doped layer. The above activities demonstrate the potential of HVPE, if properly designed, to be a competing growth technique in nitride based device structures. OMVPE has been so far the workhorse in the growth of high quality GaN, A1N and AlGaN films, with reportedly the best performances for all devices such as FETs, photodetectors, and light sources. The source materials generally used are trimethylgallium (TMGa) for Ga, trimethylaluminum (TMA1) for Al, and NH3 as nitrogen source. Mg is widely used as a p-type dopant and Si as an n-type dopant. Biscyclopentadienyl (Cp2Mg) is used as a source of Mg, and methyl silane (MeSiH3) as a source of Si. The most commonly used substrates include sapphire and SiC, due to the lack of commercially available GaN or A1N substrates. Other substrates such as Si, GaAs, ZnO, LiGa0 4 , have

262

F. Yun

also been investigated. In the case of sapphire, the lattice mismatch is -16% to GaN and -13% to A1N. SiC has smaller lattice mismatch to both GaN (-3.2%) and A1N (<1%). The generation of extended defects is mainly due to the lattice mismatch, and in the case of sapphire, stacking order mismatch as well. In addition, thermal coefficient mismatch due to the adoption of foreign substrates presents another serious problem for AlGaN growth in that high-density cracks occur especially for thick layers. Many approaches have been taken to eliminate the extended defects such as dislocations, while minimizing the cracks induced by temperature ramping during and after the growth. Some recent results will be discussed below exemplifying efforts in defect and crack control in AlGaN thin films grown by OMVPE. Maskless lateral epitaxial growth of high Al content Alo.96Gao.04N on sapphire substrates has been investigated by Katona et al.25 in a vertical close-space showerhead OMVPE chamber. First, 0.6um A1N was grown on sapphire at 1160 °C serving as a template. Parallel, periodic trenches of 5um, with 5um stripes, were dry-etched in the A1N along the (1010) direction. Lateral growth of Alo.96Gao.04N was along the (1120) direction for 1.25um, accompanied by a vertical growth of 2.54 um. From the unetched mesas, an ultraviolet multiple quantum well (MQW) structure of Alo.42Gao.58N/Alo.36Gao.64N was grown, with cap layers on both sides. Cathodoluminescence (CL) performed on both the laterally overgrown "wings" and un-etched "seed" regions shown by the panchromatic images indicate that the emission from quantum wells located above the wing region is more intense than emission above the seed region. The authors suggest that the AlGaN grown by such a lateral growth method reduces the nonradiative recombination in laterally overgrown high Al content AlGaN which is critical for high power deep UV light emitting devices. Buffer layer selection has been demonstrated to affect critically the AlGaN quality. AlGaN has been grown on various buffer layers such as A1N,26'27 AlGaN,28 AlGaN/AIN and AlGaN/GaN short period superlattices.29 Besides, for the growth of intrinsic AlGaN on Si substrates, Si diffusion can introduce serious contamination. To circumvent this problem, a reflective buffer ZrB2 was proposed due to the close lattice-match to GaN and AlGaN. According to Vegard's law,

Growth and Electrical/Optical Properties ofAlxGa,.xN

263

the in-plane lattice constant of ZrB2(0001), a=3.169 A, perfectly matches Alo.26Gao.74N. The thermal expansion coefficients along [1010] on the basal plane are also well matched between ZrB2 (5.9xl0~6 K"1) and GaN (5.6xl0"6 K"1). For LED structure, it is believed that the 100% reflective ZrB2 conductive buffer layer ensures no loss in the emitted light due to absorption of the Si substrate. Tolle et al.30 prepared a 25 nm ZrB2 (0001) buffer on Si (111) by gas-source MBE (GSMBE) using the unimolecular precursor zirconium tetrahydroborate, Zr(BH4)4. Then the buffer was used to eptiaxially grow a 1.4um-fhick Al0.2Gao.8N(0001) film in an OMVPE chamber at 1050 °C. High resolution TEM images were taken along the [110] projection of Si, as shown in Figure 1.

(SI

Fig. 1. A high-resolution XTEM image showing the epitaxial relationship between the Si(lll) substrate, the ZrB2(0001) buffer layer, and the Al02Gao.8N film grown by OMVPE. [After ref. 30]

The ZrB2(0001) buffer layer shows no degradation subject to high temperature growth, and the interface between ZrB2 and AlGaN is very sharp. From selected-area electron diffraction patterns taken for the entire heterostructure, perfect alignment of [1120]AlGaN //[1120]ZrB2

264

F. Yun

//[l_10]Si is shown without rotation. Both Auger Electron Spectroscopy (AES) and Secondary Ion Mass Spectroscopy (SIMS) measurements confirm that no unintentional Si doping occurs in the Alo.2Gao.8N film even at the growth temperature of 1050 °C. Optical properties of the AIGaN films grown on ZrB2 buffer are comparable to a high-quality undoped Alo.2Gao.8N commercial sample grown on sapphire.

£ig|

Accumulation of In

I

Accumulation of In J 1.7 ML of In

1

;&

E 0.1 -

U-

•Eo.2

jfesF^^i:^'"-'"*' 640

650

660

670

680

690

700

Substrate Temperature ( S C)

Fig. 2. Windows of 1 and 1.7 ML of In coverage on GaN, as a function of substrate temperature and impinging in flux. [After ref. 32]

<1 ML of In 0.0

0.2

0.4

0.6

0.8

1.(

Al Mole Fraction

Fig. 3. Windows of 1 ML of In dynamically stable on AIGaN, as a function of Al mole fraction and impinging In flux. [After ref. 32]

AIGaN films have also been epitaxially grown by MBE. Although the incorporation of Al is not as well controlled as OMVPE, MBE offers unique advantages over HVPE and OMVPE. For instance, MBE is a low temperature process which minimizes impurity contamination. The growth kinetics allows atomic level control of film thickness thus making the growth of MQW very convenient. As we know, surfactants such as indium have been used as a means to influence the surface morphology and growth mode of GaN during growth, either by decreasing the surface free energy or by altering the surface kinetics.31 Monroy et al?1 assessed comprehensively the surfactant capability of In for AIGaN growth by plasma-assisted MBE (PAMBE). Those authors determined the windows of 1 and 1.7 ML of In coverage on GaN, as a function of substrate temperature and impinging In flux, as shown in Figure 2. They also determined the range for ALGa,.xN with varying Al mole fraction, in Figure 3, and found that the window becomes narrower with increasing

Growth and Electrical/Optical Properties ofAlxGa,.xN

265

Al mole fraction. The formation of a dynamically stable, self-regulated l x l In adlayer on AlGaN(OOOl) modifies the surface energy and growth kinetics in that the two-dimensional growth of AlGaN is favored under stoichiometric conditions, while at the same time the formation of metal droplets on the surface is inhibited. The structural quality of AUGa^N films 0=0.08 to 0.81) thus grown was analyzed by high-resolution XRD. It has been previously established33 that symmetric (0002) rocking curve is related to the column tilt (rotation of columns out of the growth axis), while the asymmetric diffraction is only accessible to the in-plane twist. In the symmetric (0002) diffraction, the authors observed a monotonous increase of the FWHM in the 0-2 d scan, which points out to an enhancement of alloy disorder. The FWHM in the asymmetric reflection of AlGaN are always comparable to the low values of the template, indicative of a low density of edge dislocations similar to the high quality template. These results demonstrate the capability of In as a surfactant for AlGaN growth by PAMBE.

Fig. 4. (a) SEM cross section image showing compact and columnar regions of an AlGaN nanocolumnar sample, and (b) (1120) cross sectional HRTEM image, revealing the absence of extended defects. [After ref. 37]

In the growth of GaN, the use of porous templates or substrates has been demonstrated34'35 to be an effective method in reducing the strain between the foreign substrates and epitaxial layers. This was found to be the same for AlGaN as reported by R. S. Qhalid Fareed et al.,36 when AlGaN was grew on porous GaN (P-GaN) templates. The compressive strain generated in the P-GaN during AlGaN growth helps share the

266

F. Yun

strain generated by lattice mismatch. This allowed those authors to grow crack-free layers with high optical quality, with thickness exceeding the critical limits for AlGaN deposition on the conventional OMVPE GaN or HVPE-GaN. These high quality thick AlGaN layers over P-GaN templates can serve as excellent buffer layers for deep UV optoelectronic device structures.

Fig. 5. (a) (1120) cross-sectional TEM image showing GaN quantum discs and (b) larger magnification of one column. [After ref. 37] Changing from AlGaN material growth to utilization of AlGaN for growth of other nitrides, AlGaN nanocolumns are shown to serve as an excellent template for defect-free GaN quantum discs. Ristic et al?1 reported the growth of AlGaN nanocolumns, with Al content in the range of 0 to 60%. The high crystalline quality of the AlGaN nanocolumns (30150nm) was demonstrated by PL and HRTEM measurements (shown in Figure 4), which show no traces of extended defects. GaN quantum discs embedded in AlGaN nanocolumns on Si(lll) substrates were grown by PAMBE under highly N-rich conditions. GaN quantum discs of 2 and 4 nm thick embedded in AlGaN nanocolumns were obtained by switching on and off the Al flux during variable time spans. Figure 5 shows the high-resolution TEM images of the GaN discs. Strong optical emissions from GaN quantum discs, observed by PL and CL measurements, reveal quantum confinement effects. Raman data indicate that the nanocolumns are fully relaxed, while the quantum discs appear to be fully strained.

Growth and Electrical/Optical Properties ofAlxGa/.xN

267

These nanostructures have a high potential for application in efficient vertical cavity emitters.

Fig. 6. TEM micrographs of A10.6Ga0.4N buffer layer grown by MBE, viewed alone the [0110] direction, showing open-end nanopipe structures near the surface. [After ref.38]

Similarly, Yun et a/.38 reported the growth and defect reduction of GaN layers by MBE utilizing AlGaN nanopipes. AlGaN films were first grown by PAMBE under metal-rich condition. Within the Al composition range of 0.5-0.6, open-end nanopipes were formed at the surface of AlGaN films with a density of ~6xl 09 cm"2 and a size ranging from 10 to 20 nm, as shown in Figure 6 viewed along the [OHO] direction of a Al0.6oGao.4oN template. These nanopipes, observed within -300 nm of the surface, served as a nanoporous AlGaN template for regrowth of GaN epilayers. GaN epilayers were grown to different thickness by MBE to study the microstructural and optical properties. For an AlGaN buffer layer with dislocation density of 3xl0 10 cm"2 near its surface, the overlaying GaN layers with thickness ranging from 0.1 um to ~2um were grown and analyzed by TEM for dislocation density. Figure 7 shows cross-sectional TEM images of the overgrown GaN. The GaN layers started with hexagonal islands on the AlGaN nanopipes and began to coalesce at a thickness of -0.1pm. At a thickness of 2.0 um, the dislocation density reduced to -lxlO 9 cm"2. Low temperature PL data show that the relative intensity between the excitonic peak and the DAP band increases drastically with the increase of GaN thickness. The same is true regarding the relative contribution between the excitonic peak and

268

F. Yun

YL. Besides, the shape of DAP bands is better resolved with increasing GaN thickness. All these factors are a strong signature of improved GaN quality due to the reduction of nonradiative recombination centers, which is in accord with the TEM analysis of defect reduction effect of nanoporous AlGaN buffer layer. Typically, AlGaN films are grown under N-rich conditions. Growth kinetics studies reveal that due to the stronger Al-N bond than Ga-N bond, it is more difficult to grow AlGaN especially for high Al mole fraction. In the OMVPE growth kinetics,39 Al and N precursors would easily allow pre-reactions to form A1N before they reach the substrate, while in MBE, Al species suffer a reduction of surface mobility40 as compared to Ga species, resulting in both cases the insufficient incorporation of Al in to GaN. Therefore, the growth of AlGaN with Al mole fraction covering the full range from x=0 to x=l is quite challenging.

Fig. 7. TEM micrographs of the 2.0 um thick GaN on nanoporous A10.6Ga0.4N buffer layer showing the total dislocation density (a), edge and mixed components (b), and screw and mixed components (c) of dislocations. [After ref.38]

Choi et al.5 studied systematically the Al composition dependence on growth conditions of AljGa^N alloys in OMVPE. The authors reported that the Al solid solution and the growth rate of AlGaN films grown on

Growth and Electrical/Optical Properties ofAlfia^N

269

c-plane sapphire substrates are strongly affected by gas-phase parasitic reaction between NH3 and group-Ill sources such as TMA1. Thus, the parasitic reaction must be considered to control Al concentration in the solid AlGaN films. The relationship between Al composition and various growth parameters are plotted in Figure 8. Decreasing the TMGa flow rate in particular is more effective than increasing the TMA1 flow rate to obtain high Al concentration in the solid in AlGaN films (Fig. 8(a)). This is due to the fact that the degree of parasitic reaction is nearly independent of variation in the TMGa flow rate. It was also found that the Al concentration in the solid saturated as increasing the Al gas composition increased by increasing the flow rate of the Al source precursor (Fig. 8(b)). By increasing the total flow rate of TMA1 and TMGa sources (while keeping the Al gas mole composition the same), both Al concentration in the solid and the growth efficiency is increased (Fig. 8(c)). As the NH3 flow rate increases with fixed flow rate of groupIll sources, both the Al concentration in the solid and the growth rate of AlGaN are decreased. A,

;(a) 80 g 60

„Q".«N 1

-

20 01 ( 7 l 0 20

.

1

40

1

.

TO

x;r<%)

1.-1—

80

100

10

20

30

SO 100 160 200 tTMGa]*[TMAI] (pmol/mln)

Fig. 8. Relationship between Al concentration in the solid and Al gas composition in AlxGai_xN films growth on GaN/sapphire. Al gas composition was controlled by varying (a) TMGa and (b) TMA1. (c) Variations of the Al concentration in the solid and the growth rate of AlGaN films with varying group-Ill flow rate. [After ref.5]

In MBE growth, Al mole fraction can be controlled by varying parameters such as Al cell temperature,5'41 Ga cell temperature,5 both Ga and Al fluxes with the fixed Al/Ga ratio,5'8 and Al/Ga ratio with the fixed

F. Yun

270

total flux. Nikishin et al. reported the growth of Al^Ga/^N, over the full composition range (0<x
c o

'S

I 0.6 Si

a £ | 0.4 c

e 3

3 0.2 2.0x10'"

'

L

4.0x10*

6.0x10*

8.0x10"

1.0x10'*

Pressure (Torr)

Fig. 9. Al mole fraction x in AljGa/.jN alloys as a function of nitrogen plasma pressure under Ga-rich conditions. [After ref.44]

He et al.44 reported the growth of AlGaN films on c-plane sapphire by PAMBE under Ga-rich condition. First, a nominal 50-nm thick AIN buffer layer was deposited at a substrate temperature of 850 °C. Then the

Growth and Electrical/Optical Properties ofAlxGaj_xN

271

substrate temperature was lowered to 750 °C for AljGa^N growth. Two Ga cells were used to ensure Ga-rich condition. AlxGa/.^N films covering the full range of Al mole fractions have been realized by varying the nitrogen flow rate, while keeping the arrival rates of Al and Ga species onto the substrates constant. The nitrogen partial pressure in the MBE chamber was varied in a moderate range of2.7xl0- 6 -9xl0" 6 torrto ensure N-limited conditions. It is found that Al mole fraction increases with decreasing N flow due to preferential bonding of Al and N over Ga and N, as indicated in Figure 9. This is because the available nitrogen species will always consume all available Al species first before they react with Ga species. With decreasing nitrogen flow, the available nitrogen species reacting with Al species will gradually change from surplus to deficient. Consequently, the growth rate decreases as the Al mole fraction increases. As a comparison, the same authors also grew some A l / j a ^ N epilayers under N-rich conditions. The Al composition in this regime was controlled by varying the Ga source temperature with a constant Al flux for practical reasons. In this regime the Al mole fraction in A\xG&i. XN remained the same with increasing N pressure, since it is determined by Al and Ga fluxes at a given substrate temperature. 2.2. Conduction Type Control of AlGaN p-type AlGaN layer is an indispensable part of all p-n junction devices such as LEDs and lasers. The control of p-type doping with high conductivity is very critical in making good ohmic contacts to devices.45 In general, p-type doping by epitaxial techniques requires post growth annealing to activate the dopants. Experimental results point out that in AlGaN, Mg activation energy increases with Al mole fraction,46 making it especially challenging for applications requiring both p-type conduction and high Al content, such as solar-blind photodetectors and UV lasers. Recently, in order to improve p-type conduction in AlGaN, Nakarmi et al.41 proposed Mg 8-doping in AlGaN epilayers grown by OMVPE. As shown in Figure 10, a 5-junction-like doping profile is implemented by interrupting the usual crystal-growth mode via the switching of Ga

272

F. Yun

(Al) flow and leaving the NH3 flow constant. This is a so-called impurity-growth mode where Mg impurities remain on when the host crystal does not continue to grow. Those authors hoped that by using this technique, a small fraction of available Ga sites in the 5-doped plane would be occupied by Mg impurities, thus enhancing the Mg incorporation. A twofold reduction of resistivity was achieved in Alo.07Gao.93N epilayer. By this method, those authors also observed a lOx dislocation density reduction possibly due to the growth interruption in the Mg 5-doping duration that partially terminates the dislocation propagation in the growth direction.

GaNorAlG;iN:Ms?

d

PM£ < "i..\ i-in'.tfi

V, 7

Al,0, substrate Fig. 10. Schematic diagram of Mg 5-doped GaN or AlGaN, where d (=15 nm) and PMg denote the distance between two 5-planes and the two-dimensional Mg doping concentration, respectively. [After ref. 47]

2.3. Dislocation and Morphology Analysis High-density cracks48 usually generate in GaN and AlGaN films with thickness exceeding their respective critical thickness.49 This is especially true when grown on substrate materials having smaller thermal expansion coefficients than GaN and AlGaN, such as SiC or Si. The cracking represents a serious problem for device applications. Epitaxial lateral overgrowth (ELO) technique by itself is hard to eliminate this problem. Several other techniques have been introduced to

Growth and Electrical/Optical Properties ofAlxGa,.xN mitigate cracks and reduce threading dislocation (TD) density for AlGaN, including antisurfactant,50 low-temperature A1N interlayers,51 Si doping,52 and so on. T. Akasaka et al.50 reported the formation of Si^Al^N at the initial stage on SiC substrates to obtain crack-free AlGaN thin films by OMVPE. By growing very thin (l-2nm) heavily Si-doped A1N multiple interlayers prior to AlGaN, the TD density of the latter was reduced by one order of magnitude. They believed that the interlayers form Si^Al^N ternary alloys with Si molar fraction ranging from 0.07 to 0.17. These SijAli.jN alloys were in fact nanosized facets and therefore facilitate lateral growth. Due to the large biaxial strain between Si^Al^N and AlGaN layers, some TDs change their direction to horizontal propagation, and finally loop with other TDs having opposite Burgers vectors. With this technique, those authors were able to grow crack-free AlGaN films at 1 um, much smaller than that required in conventional ELO techniques.

0.4

0.3

I

00.2 AlGaN (tilt) I 00.2 GaN (tilt) i 11.0 AlGaN (twist) I 0.2

0.1 ka 80

120

160

200

AIN interlayer thickness (A)

Fig. 11. The Rocking curve FWHM versus interlayer thickness. Lines are linear least squaresfits.[After ref. 54] The interlayer thickness and Al composition both affect the particulars of the defect structure in AlGaN epilayers, such as the twist and tilt mosaic structure. Lafford et al. investigated the effect of an AIN interlayer between micron thick GaN and AUGa/.^N layers grown by OMVPE on basal plane sapphire.

273

F. Yun

274

u

?

5.140

3.175

5.135 a o

3.170

g

Out-ot-plane lattice param In-plane lattice parameter

r

«

1

a 5.130 5.125

3.165

g 3

»c

Rf

f

^ \

_ _ _

3.160

5.120

B.

5 115

? 8

-

80

120

160

200

240

AIN interlayer thickness (A)

Fig. 12. a and c lattice parameters of the AlGaN layer vs. interlayer thickness ;. Solid lines are fits to the strain relief equation y=A-(B/t), A and B being constants. [After ref. 54]

The rocking curve FWHM was plotted against AIN interlayer thickness in Figure 11 for both surface symmetric reflections and grazing incidence in-plane diffraction. No change is found in the tilt mosaic (threading screw dislocation density) with thickness or Al fraction x of the upper layer. A linear increase in the twist mosaic (threading edge dislocation density) was observed as a function of interlayer thickness and x. For all samples the twist mosaic of the AlGaN was significantly greater, by at least a factor of two, than that of the GaN layer. The a and c lattice parameters was determined as a function of the AlGaN layer thickness independently of the Al fraction, and are plotted in Figure 12. With increasing interlayer thickness the in-plane lattice parameter of the AlGaN decreased. It therefore becomes more strained with respect to the underlying AIN interlayer (assumed to be fully relaxed with respect to the GaN buffer). The variation in strain does not fit well the standard strain relief models, shown by the solid lines in Figure 12, but the trend is similar. As observed by Biasing et al.,55 those authors explained this based on the relaxation at the GaN/AIN interface that generates extra threading edge dislocations. The twist mosaic in the GaN buffer layer arises from the threading components of the dislocations associated with the lattice mismatch between the nucleation layer and the sapphire substrate. After the initial island growth, the subsequent growth favors lateral overgrowth. The twist mosaic arises from the coalescence of the

Growth and Electrical/Optical Properties ofAlxGa,.xN

275

islands, giving blocks of high perfection separated by high dislocation density cell walls and rotated with respect to each other.

Fig. 13. (a) and (b) SEM images of cracks in an MOVPE sample with Al mole fraction of 0.24. (c) Cross-section profile of a crack of type A (dashed curve) and type B (solid curve), respectively, in an Al02Gao.8N layer as determined by FE simulation for a crack spacing of 10 (J m. [After ref. 57]

Comparing the Ga- and N-rich AlGaN films, Jasinski et al.56 reported on the TEM analyses of AlGaN layers grown on sapphire substrates by MBE, with Al composition in the range of x=0.10-0.25. For a Ga-rich AlGaN sample, the density of TDs near the top surface is ~5xl0 9 cm"2, while for a N-rich AlGaN sample, the TD density is ~2 x 10!Ocm"2. However, the interesting point is that the distribution of TDs among screw, edge and mixed types (with different burgers vectors) is quite different for Ga and N-rich samples. By using the two beam conditions in TEM, edge type dislocations are out of contrast with gvectors along [0002] direction, and screw dislocations out of contrast with g-vectors along [1120] directions. Those authors showed that for Ga-rich AlGaN sample, edge and mixed type dislocations have a proportion of -70%, while in N-rich sample, the dominant dislocation type is edge dislocation (-95%). Einfeldt et al.51 studied the strain relaxation due to cracks of different depths in AlGaN layers grown on GaN template layers. They observed

276

F. Yun

two types of cracks: type A terminates at the AlGaN/GaN interface, and type B penetrates through the AlGaN/GaN interface and terminates at the GaN/sapphire interface, as shown in Figure 13. The experimental data consistently show that the relief of tensile stress increases with decreasing crack spacing. The measured strain profiles between the cracks are well described by the theoretical calculations for small crack spacing (Fig. 13c). For larger crack spacing, the mechanisms are related to inelastic strain relaxation, the reliability of the deformation potential for AlGaN, and the spatial variations of Al composition in AlGaN. Their CL data indicate a parabolic strain profile between the cracks; the strain in the center between two cracks depends linearly on the crack spacing. At the large crack spacing, the averaged strain in AlGaN levels off and approaches its maximum value, corresponding to a coherent AlGaN layer on relaxed GaN, only in the XRD but not in the CL measurements. They argued that the picture of a purely elastic strain relaxation in a homogeneous AlGaN layer is not enough to explain the observed strain patterns. Inelastic lattice deformation due to dislocations, in the manner suggested by Hearne et al.58 is most likely at play. Continuing on strain, Wu et al.59 studied an AlGaN layer grown by OMVPE on sapphire (0001) using a GaN intermediate layer. The Al composition was measured by a combination of XRD and RBS/channeling, from which the perpendicular and parallel elastic strain of the Alo.28Gao.72N layer can be separately drawn as e~L=-0A6% and e7/=+0.39%, respectively. The small ratio |e"7en 1=0.41 indicates that the Alo.28Gao.72N lattice is much stiffer in the c-axis direction than in the a-axis direction. Misfit dislocation arrays were also observed by Bell et al.60 in AljGay.jN alloys grown by facet controlled ELO (FACELO) on serrated GaN templates. The serrated GaN templates were prepared by first growing a 4-um GaN followed by Si02 stripe patterning (4 um period, 2um width) along <11_00> direction. Then a second layer of GaN was grown on the stripes at elevated temperature of 900 °C and a reduced NH3 flow rate. These conditions are conducive for faceted growth to form a serrated surface as shown in the schematic diagram in Figure 14. AlGaN was deposited at 1100 °C on the serrated GaN surface and the growth was continued until the sample surface flattened, ~5um above the

Growth and Electrical/Optical Properties ofAlxGa,.xN

277

apex of the GaN serrated template. The cross-sectional TEM image was also shown in Fig. 14. The clear dark contrast line (marked as "1") corresponds to the large TD density with Burgers vectors on the basal plane, which is associated with lattice misfit relaxation.61 A second set of dislocations of the same nature was observed on inclined boundaries within the AlGaN layer (marked as "2"). These misfit dislocations are closely related to regions with significantly large variations in Al composition. The dislocations are on the inclined planar boundaries and result from basal-plane slip, which is allowed in this inclined facet geometry. The spatial variation of the Al composition in the overgrowth region is determined by CL and ranges from x =0.06 to 0.27, for constant growth conditions that after planarization result in a uniform composition at x=0.16. The above results suggest that growth direction significantly affects the Al incorporation which is preferred for facets parallel to the basal plane. Therefore, use of faceted lateral growth has the potential to obtain crack-free thick AlGaN films.

Fig. 14. Cross-section schematic diagram of the specimen (left) is shown with a crosssection TEM image (right) at the same scale. Two sets of dislocation lines are labeled 1 is between the GaN and the AlGaN and 2 is located with in the AlGaN. [After ref. 60] '

278

F. Yun

wavenumber [cm 1 ]

Fig. 15. Raman spectra of an Al072Gao28N after 15 min anneals at various temperatures, recorded at room temperature at back scattering Z(X,.)Z geometry using an excitation wavelength of 224nm. [After ref. 62]

The thermal properties of AlGaN have been studied by Kuball et al. employing ultraviolet Raman scattering spectroscopy. The degradation pathway of Alo.72Gao.2sN was monitored for hightemperature treatments. Figure 15 plots the Raman spectra after various 15 min. anneals at temperatures up to 1200 °C. Before annealing, the Ai(LO) phonon appears at 850 cm"1, which is resonantly enhanced due to the UV excitation of 244 nm, together with the second- and third-order Raman scattering at 1700 and 2550 cm"1, respectively. The E2 phonon of Alo.72Gao.28N is much weaker than the A^LO) phonons under those resonant conditions and thus not visible in Fig. 15. For annealing temperatures higher than 1150 °C, the Alo.72Gao.2sN film decomposes into low- and high-Al composition AlxG&i_xN phases, which includes the buildup of strain, the introduction of microscopic defects, and the transformation of Alo.72Gao.28N into Alo.46Gao.54N and Al0.9oGao 10N. At 1100 °C, prior to the Alo.72Gao.2sN decomposition, the Raman spectra show the buildup of a large strain in the Alo.72Gao.2sN film. The crystalline quality of Alo.72Gao.2sN is unaffected up to 1000 °C. The 62

Growth and Electrical/Optical Properties ofAlxGaj.xN

279

results illustrate the enhanced thermal stability of AlGaN compared to GaN. a)

b) 582 -Sample A (x-0.3*) - Sample apea.31) - Sample C(x«0.08) / * - GaN /•* a

-Sample A (x«0.34) -Samples (x*3.3ij KTISJ - o - S a m p l e C(x=0.08)

580 "• " i S i ! \„

x>0.31

7

P

576 I§ 574 | 572 I 570 HI

)

800 1000 1200 Annealing temperature CO

568

0

800 1000 1200 Annealing temperature fC)

Fig. 16. E2 phonon frequency as a function of annealing temperature for annealing in (a) air and (b) nitrogen. Also included is the datafor GaN. [After ref. 63]

The annealing effects of AlGaN in air, oxygen and nitrogen were comprehensively studied on stress and compositional variation by Rajasingam et al.6i using UV micro-Raman spectroscopy. Low (x=0.08) and high (x=0.31 and 0.34) composition AlGaN, grown by OMVPE and MBE, were compared. Composition and morphology were monitored using AES and AFM, respectively. Shown in Figure 16 are the Raman spectra of different AlGaN samples annealed in various ambient. All samples up to x=0.34 exhibit maximum stress changes in the compressive direction when annealed in air. This appears as an increase in compressive stress for the low composition sample (x=0.08), and a reduction in tensile stress for higher composition samples (x=0.31 and 0.34). AES confirms this to be due to higher oxygen incorporation after annealing in the air ambient, and shows higher oxygen incorporation in the vicinity of cracks and defects. OMVPE and MBE samples of a similar composition were found to reach the same biaxial stress after anneal, despite differences in initial stress and growth temperature. Relaxation of a parabolic inter-crack stress profile to homogeneous stress was observed with annealing in all ambient for cracked samples. Sample deterioration occurs at 1100 °C for anneal-ing in air ambient, and causes a relaxation of stress in all the AlGaN layers. Sample deterioration

280

F. Yun

occurs at 1200 °C for annealing in nitrogen ambient. Annealing in nitrogen ambient does not cause any significant change in the stress state of un-cracked AlGaN. Stress or strain information can also be obtained in Raman measurements via the so-called phonon deformation potentials (DPs), which are linear coefficients linking the change in the phonon frequency and the strain or stress in the material, and are required to convert phonon frequency shifts into stress values. Although elastic properties64-66 and phonon deformation potentials65'67"71 of GaN and A1N materials have been reported in the literature, little information is available regarding the values of the phonon DPs of the Al^Ga^N ternary alloys. Sarua et al.n studied, by micro-Raman spectroscopy, the high frequency phonon deformation potentials (E2) in AijGa/JNF material grown by OMVPE directly on Si(lll) substrates. In their study, mechanical bending was applied to introduce biaxial stress in the AlxGa.jjN layers and Raman shifts were measured as a function of the applied deformation. The Si phonon mode provides a reference for the applied stress and allowed determination of phonon deformation potential values for AljGa/.jN with varying Al mole fractions. The effective elastic properties of AlGaN alloys are found to be similar to GaN and A1N as they themselves possess very close figures (see Table I), although the theoretical values seem to underestimate the experimental deformation potentials. 2.4. Chemical Ordering in AlGaN Korakakis et al.13 first observed the long-range order of AlGaN films grown by MBE, on both sapphire and 6H-SiC substrates. As we know, AljGa/.jN alloys crystallize in the wurtzitic, hexagonal close packed structure. The observation of chemical ordering in AlGaN is governed by the allowed Bragg reflection (hkil), where the geometrical structure factor is given by the expression:

Growth and Electrical/Optical Properties of AlxGa\_xN

281

where /i and f2 are the average scattering factors of the (Al, Ga) atoms occupying the (000) and ( y , y , y ) sublattice sites of the hexagonal cell, respectively. In the cases of pure GaN, A1N, or homogeneous AlGaN, the two sites are occupied by the same atomic species or by a random mixture of the two species, Bragg reflections with /=odd and h+2k=3n such as (0001) and (0003) are forbidden according to Eq.(l).

17

18

35

36

53 ,

54 74

75

98

29 (degrees)

Fig. 17. 9-26 XRD scan from an Al05Gao.sN sample with particularly strong ordering grown on c-plane sapphire. For this sample the higher order superlattice peaks, (0003) and (0005) are measurable. [After ref 73]

If, one of the sublattice sites is preferentially occupied by Al or Ga in an AljGa/.jN alloy, the two terms no longer cancel and superlattice peaks can be observed in x-ray diffraction, as demonstrated in Figure 17. Study of the existence of chemical ordering, its domain size, its dependence on Al mole fraction, etc., is very important for understanding and harnessing the material properties and device performance. Indeed, Iliopoulos et al.s later pointed out that the chemical ordering of AlGaN was close related to the ratio of group-Ill to group-V fluxes during MBE growth. They also demonstrated by TEM and electron diffraction that three types of spontaneously formed superlattice structures were present with periodicities of 2, 7, and 12 ML in their AlGaN films with Al mole fraction ranging from 0.45 to 0.87. The 2 ML ordering was preferred under N-rich conditions, while the 7 and 12 ML orderings were more likely observable under metal-rich conditions.

282

F. Yun

Fig. 18. The diffraction patterns containing superlattice spot, (a) 1:1 ordered phase alone [1010] zone axis, gl: 0001, g2: 1120, (b) 3:1 with the three spots between 0000 and 0002, alone [1010] zone axis, and (c) superlattice spot, d~3.1nm, diffraction pattern alone [1120], gl: 0001, g2:1010. [After ref. 74]

Ruterana et al.14 made a crystallographic analysis of a series of AlGaN layers grown by OMVPE and found by electron diffraction that the layers were inhomogeneous, meaning they have multiple chemical ordering. Figure 18 shows three diffraction patterns related to superlattice spots in Alo.14Gao.86N. It can be seen that even in the low Al composition range (x=10%-15%), many types of chemical ordering take place: AIN/GaN (1:1), Alo.25Gao.75N (3:1) which was also reported in InGaN.75 Moreover, another type of ordering is shown to take place in which a series of 5 GaN cells and one A1N form the new supercell (i.e., 10 GaN monolayers and 2 A1N monolayers). This third type of ordering was systematically detected by XRD in all the AlGaN layers by those authors, giving an average composition corresponding to Alo.i66Gao.834N. The three types were found to coexist in the same layers, leading to inhomogeneity in AlGaN growth. Another important crystallographic property accompanied by the chemical ordering is the evolution to lower symmetry than wurtzite structures, which have the (P3ml) space group.

3. Electrical Properties 3.1. Electrical Transport Variable temperature Hall effect measurements are commonly employed to study the electrical transport properties of AlGaN. For example, Zhu

Growth and Electrical/Optical Properties ofAlxGaj_xN 0.050

1

*»

0.8-

N„ - 6.0x10™ em-* _

0.045

~7. (©"an")

0.040-

• •• •

4.0-

25-

0.025

0.010 '

I

•

i

•

1

«

1

•

l

•

1

'

100 200 300 400 500 600 700 T(K)

0.005

' I '"I ' P* 2 3 4 S 8 7 N.(1
' + <.

0.015

2015-

'V

0.020-

|T~,

(c)

0

0.030-

• • ••

I' I • I • I • T-300K.

Q.

0.035 (b)

§ 3 2

283

&

(d)

b**™®® 100 200 300 400 500 600 700

T(K)

Fig. 19. Variable temperature Hall effect measurement results for n-Aln.7Gao.3N (N si =6.0xl0 19 cm"3), (a) resistivity, (b) electron concentration n, and (c) electron mobility |i. (d) resistivity of AI07Gao3N with varying Si doping levels. Inset shows roomtemperature resistivity vs. N si . [After ref.76]

et al.16 prepared heavily Si-doped n-type Alo.7Gao.3N films by OMVPE on sapphire substrates. SIMS profiling indicates the nominal carrier concentration (NSi) ranging from 2.6 to 6.8x10 cm"3. The electrical transport measurement results are plotted in Figure 19. T ^ in Fig. 19(c) indicates the characteristic temperature at which electron mobility reaches a maximum value. Strong thermal activation process is evident for the sample with the lowest doping level (see Fig. 19(d)). At high doping levels, samples exhibit degenerated behavior. There is a clear trend of decreasing resistivity with increasing Si doping level. For the sample with NSi=6.0xl019 cm"3, n-type resistivity of 0.0075 Q. cm was achieved with an electron concentration of 3.3xl019 cm"3 and a mobility of 25 cm2/Vs at room temperature. For the same sample, the effective donor activation energy E0 was determined to be as low as 10 meV. E0 increases to 25 meV as iVSi is reduced to 2.6xl019 cm"3, which can be explained by the band gap renormalization effect. This implies that heavy doping is necessary in high-Al-content AlGaN alloys to bring down the donor activation energy in order to obtain higher conductivity.

F. Yun

284

1000°C, 1.0|im/h (series A )

o

.o <3

c

o «J

o

0.4

Al composition

Fig. 20. Composition dependence of the room temperature electron concentration in AlGaN layers grown at 1000°C and 1050°C. [After ref. 77]

Like GaN, undoped AlGaN films show n-type conductivity due to electrically active defects. Polyakov et al?1 investigated the origin of these defects in AlGaN by measuring the shallow and deep centers using temperature-dependent Hall effect and PL measurements. Two sets of undoped n-type AlGaN layers grown by OMVPE were studied with different Al mole fraction in each set, and with orders of magnitude difference in electron concentrations between the two sets. As shown in Figure 20, the concentration of donors in AlGaN layers is greatly reduced by increasing the growth temperature from 1000° to 1050 °C and decreasing the growth rate from 1 to 0.7 um/h. The type of donors in the samples grown at 1000 °C is believed to be different from donors observed in samples grown at 1050 °C, although they are both related to native defects. The composition dependence of ionization energies of dominant donors in these two sets of samples is very different indicating that different types of defect centers are involved. Those authors suggest that the defects behave as hydrogen-like donors for low Al compositions and become increasingly deeper with increasing Al content. The shallowdeep transition occurs at about x=0.2 in the low conductivity Al^Ga/.^N series and at about x=0.5 for the high conductivity series.

Growth and Electrical/Optical Properties ojAlfia^N

285

Later on, Polyakov et al.ls performed a more systematic Si doping study on ALxGai.xN and found an abrupt decrease in the free carrier concentration at compositions around x=0.4. This dependency has been explained by the sudden deepening of the Si donor level,79 as well as oxygen-DX formation at this composition.80,81'82 10"

4

2 *

10"

0.20

0.25

0,30

0.35

0.40

0.45

0.50

0.55

Alloy composition Fig. 21. Composition dependence of the free-carrier and ND-NA concentrations, as determined by 300 K Hall and capacitance-voltage measurements, respectively. [After ref. 83]

To further investigate the abrupt decrease in resistance previously observed for rc-type Al^Ga/.^N at around x=0.4, a series of AlxGa;.xN layers were grown over the range of x=0.2- 0.5 by Wagener et al.83 In addition to Hall effect measurement and the regular capacitance-voltage measurement, a so-called optical isothermal capacitance transient spectroscopy (O-ICTS) was employed to experimentally determine the origin of the midgap states. The composition dependence of the freecarrier and ND-NA concentrations, as determined by 300 K Hall and capacitance-voltage measurements, was plotted in Figure 21. The intrinsic compensation of AlGaN increases systematically with increasing Al compositions. O-ICTS spectra reveal the presence of two midgap states with concentrations in the low 1017 cm"3 range. These two levels, which are thought to be responsible for the observed compensation, have been assigned to the third and second ionization states of the Al vacancy.

286

F. Yun

3.2. Deep Level Defects It is known that Si donor binding energy increases with increasing Al mole fraction, mainly due to the smaller dielectric constant of A1N than GaN and the larger effective mass of the electron in A1N.84 However, it is not exactly known how Si donor energy changes with increasing Al mole fraction. Si has been reported to form localized deep level or DX state for higher mole fractions of AlGaN. Si has also been reported to remain as shallow donor86 up to A1N. The conductivity of intrinsic AlGaN is known to drop significantly with increasing Al mole fraction, particularly for x>0.4,87'78 indicating a large increase in the carrier localization energy.88 Beside, oxygen has been known to pose a threat to doping in high mole fraction AlGaN due to its large affinity to Al during growth. For low mole fraction AlGaN, oxygen acts as a shallow donor, while for high mole fraction AlGaN, has been predicted to be a deep level by a DX transition for x>0.4.86 Such studies provide the physical basis for understanding the dopant and deep level electronic properties especially for high Al-content AlGaN. Along this vein, Bradley et al.&9 explored the nature of deep level defects and their effects on Si-doped AlGaN with high Al content (25%100%), using the depth-dependent CL spectroscopy and SIMS profiling. SIMS results provide correlations between AlGaN deep level emissions from CL and elemental impurities distributed throughout the epitaxial bulk films. The highest Al mole fraction samples exhibit deep level optical emissions that correlate with O and C impurities measured by SMS. The O impurities contribute as donors at low and intermediate Al content, while form deep levels in high Al-content AlGaN. Interestingly, bowing parameter for AlGaN is determined from the CL energy onset of near-band-edge peak emissions, and b=l is obtained for 0<x<0.98. Figure 22 shows that the activation energy is significantly less for high Al (x>0.8) AlGaN (-36 meV) than the lower Al samples (-54 meV). Correspondingly, SIMS measurements demonstrate enhanced O concentrations in the high Al-content AlGaN films than lower Al mole AlGaN. This absence of free carriers for x>0.80 is consistent with Si donor compensation due to deep levels associated with oxygen.

Growth and Electrical/Optical Properties ofAlxGa/.xN

287

E

5 keV excitation 0.2

03

04

OS

0.6

07

08

0-9

1.0

At mote fraction

Fig. 22. Activation energy of the total integrated near band edge emission with 5 keV excitation for AlGaN samples. [After ref. 89]

High Al mole fraction intrinsic AlGaN has been of interest since it can reduce the dark current in solar blind photodetectors. For such application, both the electrical and optical properties under radiation are very important. Polyakov et al.90 studied the proton implantation effects on the electrical and optical properties of undoped n-type Al0.4oGao.6oN films. The samples were studied before and after implantation of various doses of 100 keV protons. In the virgin samples, the electrical properties were determined by deep donor defects with an energy level near 0.25 eV from the conduction band edge and a concentration of ~1018 cm"3. Other deep centers present had energy levels of 0.12, 0.3, and 0.45 eV. The luminescence spectra were dominated by two defect bands near 2.3 and 3.6 eV. Figure 23 shows how the measured sheet resistivities changed with the proton dose. It clearly shows that proton implantation significantly decreased the concentration of major donors even at the lowest doses of 1012 cm"2. For higher doses the Fermi level became progressively deeper and the data indicated complexion of defects present in the sample with either primary radiation defects or hydrogen introduced by implantation. The overall effect of the proton implantation on the intensity of luminescence bands was the increase of intensity of defect bands with implantation. The threshold dose at which detectable changes in the carrier concentration and luminescence efficiency are observed is, for undoped «-AlGaN with x=0.4, quite low, about 1012 cm"2

288

F. Yun

(for comparison, in n-GaN with electron concentration of some 1016 cm"3 comparable changes start at doses of about 1014 cm"2)91 which may have implications for radiation hardness of piezoelectrically doped HEMTs and solar-blind photodetectors. Proton implantation is deemed capable of making AlGaN films with very high sheet resistivity which could be of use in device isolation.

1 a

hee res tivity

JS

10121011^ 1010.

109!

;

108!

w 107i

'

10s 0.0

5.0x10 1 3

1.0x1014 2

Protons d o s e (cm" )

Fig. 23. Dependence of the room temperature sheet resistivity on the 100 keV proton does for studied undoped n-AlGaN samples. [After ref. 90]

la

s a>

QPH

4.25 4 .24 4.23 4.22 4.2-1 4 .20 3.7 6 75 74

A l„ , . G a „ , , N

.73 .7 2 .7 1 .7 0 .83 62 .6 1 00 3.59

~~v 0

100

200

300

Temperature (K)

Fig. 24. T-dependent PL spectra of AlxGa!_xN epilayers with various Al compositions. [After ref. 97]

Growth and Electrical/Optical Properties ofAlxGaj.xN

289

4. Optical Properties Potential fluctuation plays an important role in determining the optical properties of alloy semiconductors. Although theoretical calculations do not show an unstable phase segregation in Al^Gay.^N alloys,92 optical studies of AlxGa/.xN alloys with high Al contents have observed the Sshaped PL shift and Stokes shift,93'94 which can be readily explained by alloy potential fluctuation in a fashion similar to InGaN/GaN MQWs.95,96 Chung et al.97 presented an optical study of alloy potential fluctuation in Al/ja^N using a combination of optical characterization tools en-compassing PL, optical absorption (OA), photo-current )PC), and persistent photoconductivity (PCC). The band edge peak positions of the AljGa/.jN epilayers measured by PL for x=0.08, 0.15, and 0.33 are plotted in Figure 24 as a function of temperature. Alo.0sGao.92N follows the typical temperature behavior of the energy band-gap shrinkage described by Varshni's equation:

E(T)=E(0)+aT2/(T-j3) However, Alo.33Gao.67N shows the "5-shaped" emission peak shift, i.e., decrease-increase-decrease, behavior with increasing temperature. It clearly shows that the increase of Al content is accompanied by the aggravation of the 5-shaped behavior of PL peak energy. The depth of localized states obtained from PPC decay kinetics for Alo.33Gao.67N is 152 meV that is somehow related to the value of 121 meV determined from PL, OA, and PC spectra. This increase in the degree of localized states with large Al compositions, together with the 5-shaped behavior and Stokes shift can be described in terms of localized states formed by alloy potential fluctuations in AlxGa;.xN epilayers. Collins et a/.98 reported enhanced room-temperature luminescence efficiency through carrier localization in Al^Ga^^N alloys. The AlGaN samples were grown by PAMBE on sapphire (0001) substrates, with Al content of 20%-50%. All samples show intense room-temperature PL that is significantly redshifted by 200^400 meV from band edge, as seen in Figure 25.

290

F. Yun

red-shiltud -10K -26K 43 K -73K 103 \ 143 i 193 I»JK K £i ?4?K -•V~X2K 1'

. *

K f\\; i"

'#

\ \

\\ |\\ I \\ band edge

3.8

3.9 4.0 4.1 Photon Energy (eV)

4.2

4,3

Fig. 25. T-dependent PL spectra of the Alo.33Gao.67N film comparing redshifted, 3.82 eV, and band edge, 4.11 eV, emissions. Inset shows a room-temperature monochromatic CL image, 2.56 x 2.56 |im, for the redshifted emission. [After ref. 98] 6.5

—1

¥

i

A

A l , ,<5»„N

c S.5 O 43

4J « ^ V ~4jS"~y Pt Energy (eV)

JC4.S

^

CS 0>

a

/ *

_i

d) (0 (b)

' t i s

"a

! 1 1 , 1

L..^ , l _ i _ L j _ t _

0.10.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Al Composition Fig. 26. PL peak position as a function of composition. The inset presents characteristic spectra acquired from Alo.5Gao.5N at two adjacent spots. The spectral variation is 39 meV. [After ref. 99]

r ' 1 Alo.057Gao.943N

K ,

A/^lffA. JI In A. _J AIV All yl \V \V

"(e) '•

1

A A Ms—

: ;

4

1 XX

N

:

._

O Q.

r~—'

* 3.58 3.60 3.62 3.64 PHOTON ENERGY (eV)

Fig. 27. Excitonic PL at 8 K taken from Alo.057Gao.943N under excitation-power densities of (a) 0.011, (b) 0.099, (c) 0.28, (d) 0.90, (e) 2.5, (f) 4.8, and (g) 11 kW/cm2. [After ref. 100]

Growth and Electrical/Optical Properties ofAlxGa,.xN

291

The intense emission is characterized by a long room-temperature lifetime (375 ps) comparable to that seen in low defect density (108 cm"2) GaN. Room temperature monoch-romatic CL images at the red-shifted peak reveal spatially nonuniform emi-ssion similar to that observed in In(Al)GaN alloys, which was attri-buted to compositional inhomogeneity. These observations suggest that spatial localization en-hances the luminescence efficiency despite the high defect density (1010 cm-2) of the films by inhibiting movement of carriers to nonradiative sites. Bergman et al." exploited deep UV photoluminescence to study the alloy spatial compositional distribution at the submicron scale. The spatial dependence of the band gap light-emission energy of Al^Ga;.vN alloys at full Al composition range 0<x
292

F. Yun

The energy separation AE between exciton luminescence and biexciton luminescence increased with increasing Al composition, suggesting the increase in the binding energy of biexcitons in ternary alloys. The PL decay-time constants of both excitons and biexcitons increased with increasing Al composition in the ternary alloys, as shown in Figure 28, which resulted from the effect of localization due to alloy disorder. T

"

<

r—

AI,Ga,.EN

h

s-

zo <*•

U300

P >200

<

om a

100 J

0

•

1

1

I

i

I

i_t

•

L

0.05 0.10 Al COMPOSITION, i

Fig. 28. PL decay-time constants of both the line X (closed circles) and the line XX (closed squares) as a function of Al composition in ternary alloys. The solid lines are guides for the eye. [After ref. 100]

A comparison between the PL intensity of Ga-rich and N-rich AlGaN grown by MBE was made with regard to the temperature behavior by He et al.44 Figure 29 plots the temperature dependence of the PL intensity in terms of quantum efficiency. It shows that Alo.13Gao.87N grown under Ga-rich condition quenched at higher temperatures, as compared to the N-rich Alo.13Gao.87N sample. The quantum efficiency of PL at 15 K was markedly higher for the Al^Ga^N layers grown under Ga-rich conditions (3%^48%) compared to the layers grown under Nrich conditions (1%-10%), and is even higher than the radiative efficiencies obtained for a large set of Ga-polarity GaN layers grown under similar conditions. Such improvement of radiative efficiency in AljGa^N, especially under the Ga-rich conditions, points to a reduced density of dislocations for Ga-rich AlGaN, although carrier confinement cannot be excluded.

Growth and Electrical/Optical Properties ofAlxGa,.xN

10"

• • • i . . . . i

10

20

. . . .

r.

. . .

I .

30 40 lO^K"')

293

. .

50

60

70

Fig. 29. T-dependence of PL quantum efficiency for the AlxGai.xN layers grown under Ga-rich and N-rich conditions. [After ref. 44]

300

350 400 450 500 550 Wavelength (ran)

600 650

Fig. 30. Evolution of the refractive index (n) as a function of wavelength for several Al compositions of AlxGa!.xN alloy (x=0;0.19;0.47;0.675;0.785;l). Inset displays the extinction coefficient (k) for x=0 and x=0.19. [After ref. 101]

To permit the growth and design of vertical cavity structures and distributed Bragg reflectors on Si substrates, the refractive indices of AljGa/.jN must be known. Antoine-Vincent et al.m measured the refractive indices of several Al^Ga^N alloys deposited on silicon by ellipsometry and reflectivity measurements. The AlGaN layers were grown on (11 l)Si substrate by MBE on top of an AIN/GaN/AIN buffer in order to reduce the strain of the alloy. The Al composition is deduced

294

F. Yun

from energy dispersive x-ray spectroscopy and PL experiments. The refractive index n and the extinction coefficient k are determined in the 300-600 nm range, and are plotted in Figure 30. The extinction coefficient is also determined above the band-gap energy for GaN and Alo.19Gao.8iN, in a wavelength range where few values of the alloy composition have already been examined. For the transparent region of Al/ja/.jN, the refractive index is given in the form of a Sellmeier law.

Fig. 31. PL band peak position (a) and intensity (b) as a function of the excitation power density at temperatures of UK (circles) and 300 K (squares), (c) T-dependence of PL spectra of an AlGaN MQW. [After ref. 102]

PL study carried out by Tamulaitis et al. focuses on the AlGaN QWs deposited on A1N single crystal substrates. Al^Ga;.^N /ALGa^N MQW structures were deposited by a so-called migration-enhanced metalorganic chemical vapor deposition (MEMOCVD).103104 This technique allows the AlGaN layers to be deposited by repeats of a unit cell grown by sequential metalorganic precursor pulses of Al, Ga, and NH3. A better mobility of precursor species on the surface and, thus, better atomic incorporation and improved surface coverage can be achieved for AlGaN. PL dynamics have been investigated with increasing temperature and power density of photoexcitation, under pulsed band-to-band excitation of the well material. The results of excitation power density dependent and the temperature dependent PL

Growth and Electrical/Optical Properties ofAlxGa,.xN

295

data are plotted in Figure 31. The abnormal temperature dependence of the PL peak position and differences in the character of the peak shift with increasing excitation power density observed at low and elevated temperatures are interpreted in terms of carrier/exciton localization and screening of the built-in electric field. The formation of these localized states with narrow energy distribution and high density is favorable for efficient light emission. 5. Band Gap Bowing of AlGaN Knowledge of the band gap of AlGaN in relationship to Al composition is essential. For example, AlGaN can be used as a barrier in AlGaN/GAN QWs for carrier and optical confinements for emitters. It can also be used to define the cutoff wavelength for detectors. Composition dependent band gap is a prerequisite to design such devices. In MODFETs, accurate knowledge of the band gap discontinuity and band-offset ratio are pivotal in predicting the behavior of 2DEG, which of course is also affected by piezoelectric effects. Precise determination of the band gap as a function of x is needed for band gap engineering in order to fulfill these device applications. Optical measurements of the band gap of AlGaN began in 1970s, however, there is a wide scatter in the data up to date in the literature. (See ref. 105 and references therein.) The dispersion of bowing parameters reported by various researchers extends from -0.8 eV (upward bowing) to +2.6 eV (downward bowing), most likely emanating from AljGa/.jN alloys prepared by different techniques with various quality and, in some instances, the range of alloy compositions explored being narrow. With the band gaps of GaN and A1N known, the band gap of AlGaN can be expressed as: EAlfia{.xN

=x£MN

+^x)EOaS

_bx^_x)

(2)

where the coefficient of the parabolic term, b, is defined as the bowing parameter. The band gap of Al^Ga/.xN is denoted as E " ai~" , whereas the band gap of GaN and A1N are represented by E" and E , respectively.

296

F. Yun

The first reliable optical measurements of the AlGaN epilayer measurements were made by Koide et al.106 The layers were grown by OMVPE on (0001) sapphire substrates with a 50 nm thick A1N buffer layer. The buffer layer was grown at 800 °C and the alloy layer at 1000 °C. The absorption coefficients were measured in the photon energy range 3.2-4.4 eV and for 0<x<0.4. The plots of (ad) 2 versus the photon energy h V were straight lines near the absorption edge for Al^Ga/.^N as is true for other direct band gap materials. From the plot, the band gap can be extrapolated. The bowing parameter thus determined is b=l.0+0.3 eV.

Photon Energy [eV]

Fig. 32. Absorption coefficient at room temperature as a function of energy for various alloys with 0<x
Angerer et al.42 prepared AlxGa;.^N alloys on c-plane sapphire by PAMBE. The Al content x was varied over the whole composition range (0<x
Growth and Electrical/Optical Properties ofAlxGaj.xN

297

biaxial strain of the alloys can be calculated by an additional determination of a, using asymmetric reflections. The results obtained by x-ray diffraction and elastic recoil detection provide evidence for the validity of Vegard's law in the AlGaN system. The absorption coefficient oc as a function of photon energy h V is shown in Figure 32. The band gap values were calculated using these plots as mentioned above. In alloys with high Al concentration the assumption of a parabolic density of states near the band edge does not remain valid because absorption due to impurity levels and free excitons also contributes to the absorption coefficient. Therefore, the band gap values were also determined from the photon energy at which the absorption coefficient has a value 7.4 xlO4 cm"1. At this value the alloys even with high Al mole fraction have a clear direct band-edge absorption. A value of b=1.3 eV of the bowing parameter was determined. Akasaki and Amano107 have determined the band gaps of the strained AkGa^N layers using the PL method. The bowing parameter was b=0.25 for A l / j a ^ N layers for values of x in the range 0-0.25. Jiang et al.m used reflectance measurements to determine the optical band gap of AlGaN from am AlGaN/GaN heterostructure The normalincidence reflectance measurement was employed to obtain the free exciton transition energy (£FX) of AlGaN alloys in ALGa^N /GaN/sapphire heterostructure grown by OMVPE. It was found that the thickness variation of the AlGaN layer may cause a noticeable change in the line shape of reflectance spectrum and impede the identifying cation of the desired excitonic position. By using a reflection model of two absorbing layers with a transparent substrate, the experimental reflectance spectra were theoretically simulated and utilized to explain the reflection mechanism in Al^Ga/.^N/GaN heterostructures. Using the Al mole fraction derived from x-ray diffraction measurement, the bowing parameter of the epitaxial AlGaN layers in the range of 0<x<0.3 shows a downward value of b=0.53 eV. Lee et al.109 measured the band gap of Al/Ja^N for the composition range 0«cc<0.45 using PL measure-ments. The samples were excited with the 260 nm output of a frequency-tripled Ti:sapphire laser. Biaxial strains and alloy composi-tions were determined by XRD. Two different types of structures were grown on the LT buffers after ramping to 1050

298

F. Yun

°C: type-I structures are AljGa^N single or multiple heterolayers of - 1 um-total thickness grown on ~l-um-thick GaN; type-II structures are ~1um-thick Al^Ga/.^N single hetero-layers grown directly on the buffers. The resulting bowing parameter, b=+0.69 eV, is compared to 20 previous works. A correlation is found between the measured band gaps and the methods used for epitaxial growth of the Al^Ga/.^N: directly nucleated or buffered growths of Al^Ga/.^N initiated on sapphire at temperatures T=800 °C usually lead to stronger apparent bowing (b> + \3 eV); while growths initiated using low-temperature buffers on sapphire, followed by high-temperature growth, lead to weaker bowing (Z?<+1.3 eV). Based on the then extant data they suggest that the intrinsic band-gap bowing parameter for AlGaN alloys is &=+0.62(+0.45) eV. 6.5 I

0

i

i

0.2

0.4

0.6

0.8

1

Al composition (x)

Fig. 33. Energy band gap of AlGaN (0<x
Yun et al. revisited the bowing parameter issue using a set of MBE grown Al^Gay.jN thin films covering the entire range of alloy compositions, 0<x
Growth and Electrical/Optical Properties ofAlxGa].xN

299

resulted in a bowing parameter of &=1.0 eV over the entire composition range. Because of the improved accuracy of the composition and band gap determination, and the largest range of the Al composition over which the study has been conducted, this result is deemed to be more realistic for the complete range of Al mole fraction. 6. Summary There has been significant progress in the field of Ill-nitride material growth driven by the impetus of device application. Over the last decade, AlGaN has been grown with much improved quality and reasonably good control of conductivity, with Al mole fraction covering the whole range between the two binary extremes, i.e. GaN and A1N. Yet, due to the lack of native substrates and the associated high defect density inherent to foreign substrates, the growth technology of nitride semiconductors at large, and of AlGaN in particular, is far from being fully mastered. It can be expected, however, with the surging of research activities in this field, high quality AlGaN material will be available for application in the near future. References 1. S. J. Pearton, J. C. Zolper, R. J. Shul, and F. Ren, J. Appl. Phys. 86, 1 (1999). 2. M. N. Yoder, IEEE Trans. Electron Devices 43, 1633 (1996). 3. F. Omnes, N. Marenco, B. Beaumont, and Ph. De Mierry, J. Appl. Phys. 86, 5286 (1999). 4. G. Coli, K. K. Bajaj, J. Li, J. Y. Lin, and H. X. Jiang, Appl. Phys. Lett. 78, 1829 (2001). 5. S. C. Choi, J.-H. Kim, J. Y. Choi, K. J. Lee, K. Y. Lim and G. M. Yang, J. Appl. Phys. 87, 172 (2000). 6. J. R. Jenny and J. E. Van Nostrand, R. Kaspi, Appl. Phys. Lett. 72, 85 (1998). 7. M. Asif Khan, Q. Chen, C. J. Sun, M. Shur, and B. Gelmont, Appl. Phys. Lett. 67, 1429 (1995). 8. E. Iliopoulos, K. F. Ludwig, Jr., T. D. Moustakas, and S. N. G. Chu, Appl. Phys. Lett. 78, 463 (2001). 9. R. F. Service, Science 271, 920 (1996). 10. L. E. Brus, J. Chem. Phys. 90, 2555 (1986). 11. Y. Kayanuma, Phys. Rev. B 38, 9797 (1988).

300

F. Yun

12. M. Benaissa, K. E. Gonsalves, and S. P. Rangarajan, Appl. Phys. Lett. 71, 3695 (1997). 13. J. B. A. Mitchell and S. L. Guberman, Eds., Dissociative Recombination: Theory, Experiment and Applications (World Scientific, Singapore, 1999). 14. S. Ito, Y. Yamasaki, S. Omi, K. Takatani, T. Kawakami, T. Ohno, M. Ishida, Y. Ueta, T. Yuasa, and M. Taneya, Jpn. J Appl. Phys. 43, 96 (2004). 15. D. Doppalapudi and T. D. Moustakas, in "Handbook of thin films", (H. S. Nalwa, editor), vol. 4, New York, Academic Press, 2002. 16. H. P. Maruska and J. J. Tietjen, Appl. Phys. Lett. 15, 327 (1969). 17. M. K. Kelly, R. P. Vaudo, V. M. Phanse, L. Gorgens, O. Ambacher, and M. Stutzmann, Jpn. J. Appl. Phys. 38, L217 (1999). 18. F. Yun, M. A. Reshchikov, K. Jones, P. Visconti, H. Morkoc, S. S. Park and K. Y. Lee, Solid State Electron. 44, 2225 (2000). 19. B. Baranov, L. Daweritz, V. B. Gutan, G. Jungk, H. Neumann and H. Raidt, Phys. Status Solidi a, Vol. 49 629 (1978). 20. A. E. Nikolaev, Yu. V. Melnik, N. I. Kuznetsov, A. M. Strelchuk, A. P. Kovarsky, K. V. Vassilevski, and V. A. Dmitriev, Mater. Res. Soc. Symp. Proc. 482, 251 (1998). 21. Yu. V. Melnik, A. E. Nikolaev, S. Stepanov, I. P. Nikitina, K. Vassilevski, A. Ankudinov, Yu. Musikhin, and V. A. Dmitriev, Mater. Sci. Forum 264-268, 1121 (1998). 22. G. A. Smith, T. N. Dang, and T. R. Nelson, J. L. Brown, D. Tsvetkov, A. Usikov, and V. Dmitriev, J. Appl. Phys. 95, 8247 (2004). 23. D. Tsvetkov, Yu. Melnik, A. Davydov, A. Shapiro, O. Kovalenkov, J. B. Lam, J. J. Song, and V. Dmitriev, Phys. Status Solidi A 188, 429 (2001). 24. M. A. Mastro, D. Tsvetkov, V. Soukhoveev, A. Usikov, V. Dmitriev, B. Luo, F. Ren, K. H. Baik, and S. J. Pearton, Solid-State Electron. 47, 1075 (2003). 25. T. M. Katona, M. D. Craven, J. S. Speck, and S. P. DenBaars, Appl. Phys. Lett. 85, 1350(2004). 26. F. A. Ponce, J. S. Major, Jr., W. E. Piano, and D. F. Welch, Appl. Phys. Lett. 65, 2302 (1994). 27. C. Wetzel, E. E. Haller, H. Amano, and I. Akasaki, Appl. Phys. Lett. 68, 2547 (1996). 28. S. Takigawa, K. Furuta, S. Shimizu, X.-Q. Shen, T. Kitamura, and H. Okumura, Jpn. J. Appl. Phys. 43, 952 (2004). 29. S. A. Nikishin, N. N. Faleev, A. S. Zubrilov, V. G. Antipov, and H. Temkin, Appl. Phys. Lett. 76, 3028 (2000). 30. J.Tolle, J.Kouvetakis, D.W.Kim,S. Mahajan, A. Bell, Fernando A. Ponce, I. S. T. Tsong, M. L. Kottke, and Z. D. Chen, Appl. Phys. Lett. 84, 3510 (2004). 31. N. Grandjean and J. Massies, Phys. Rev. B 53, R13231 (1996). 32. E. Monroy, B. Daudin, E. Bellet-Amairic, N. Gogneau, D. Jalabert, F. Enjalbert, J. Brault, J. Barjon, and L. S. Dang, J. Appl. Phys. 93, 1550 (2003).

Growth and Electrical/Optical Properties ofAlxGaj.xN

301

33. B. Heying, X. H. Wu, S. Keller, Y. Li, D. Kapolnek, B. P. Keller, S. P. DenBaars, and J. S. Speck, Appl. Phys. Lett. 68, 643 (1996). 34. F. Yun, M. A. Reshchikov, L. He, H. Morkoc, C. K. Inoki, and T. S. Kuan, Appl. Phys. Lett. 81, 4142 (2002). 35. A. Sagar, C. D. Lee, R. M. Feenstra, C. K. Inoki, and T. S. Kuan, J. Vac. Sci. Technol. B 21,1812 (2003). 36. R. S. Qhalid Fareed, V. Adivarahan, C. Q. Chen, S. Rai, E. Kuokstis, J. W. Yang, M. Asif Khan, J. Caissie, and R. J. Molnar, Appl. Phys. Lett. 84, 696 (2004). 37. J. Ristic, E. Calleja, M. A. Sanchez-Garcia, J. M. Ulloa, J. Sanchez-Paramo, J. M. Calleja, U. Jahn, A. Trampert, and K. H. Ploog, Phys. Rev. B 68, 125305 (2003). 38. F. Yun, L. He, M. A. Reshchikov, H.Morkoc, J. Jasinski, and Z. Liliental-Weber, Mater. Res. Soc. Symp. Proc. Vol. 764, C7.4 (2003). 39. A. Thon and T. F. Kuech, Appl. Phys. Lett. 69, 55 (1996). 40. M. D. Bremer, W. G. Perry, T. Zheleva, N. V. Edwards, O. H. Nam, N. Parikh, D. E. Aspnes, and R. F. Davis, MRS Internet J. Nitride Semicond. Res. 1, 8 (1996). 41. S. Yoshida, S. Misawa, and S. Gonda, J. Appl. Phys. 53, 6844 (1982). 42. H. Angerer, D. Brunner, F. Freudenberg, O. Ambacher, M. Stutzmann, R. Hopler, T. Metzger, E. Born, G. Dollinger, A. Bergmaier, S. Karsch, and H.-J. Korner, Appl. Phys. Lett. 71,1504 (1997). 43. S. Nikishin, G. Kipshidze, V. Kuryatkov, K. Choi, lu. Gherasoiu, L. Grave, de Peralta, Z. Zubrilov, V. Tretyakov, K. Copeland, T. Prokofyeva, M. Holtz, R. Asomoza, Yu. Kudryavtsev, and H. Temkin, J. Vac. Sci. Technol. B 19, 1409 (2001). 44. L. He, M. A. Reshchikov, F. Yun, D. Huang, T. King, and H. Morkoc, Appl. Phys. Letts, 81, 2178 (2002). 45. M. Suzuki, J. Nishio, M. Onomura, and C. Hongo, J. Cryst. Growth 189/190, 511(1998). 46. J. Li, T. N. Oder, M. L. Nakarmi, J. Y. Lin, and H. X. Jiang, Appl. Phys. Lett. 80, 1210 (2002). 47. M. L. Nakarmi, K. H. Kim, J. Li, J. Y. Lin, and H. X. Jiang, Appl. Phys. Lett. 82, 3041 (2003). 48. T. Akasaka, S. Ando, T. Nishida, H. Saito, and N. Kobayashi, Appl. Phys. Lett. 79, 1261 (2001). 49. K. Itoh, T. Kawamoto, H. Amano, K. Hiramatsu and I. Akasaki, Jpn. J. Appl. Phys., 30,1924(1991). 50. S. Tanaka, M. Takeuchi, and Y. Aoyagi, Jpn. J. Appl. Phys. 39, L831 (2000). 51. M. Iwaya, S. Terano, N. Hayashi, T. Kashima, T. Detchprohm, H. Amano, I. Akasaki, A. Hirano, and C. Pemot, Mater. Res. Soc. Symp. Proc. 595, Wl.10.1 (2000). 52. H. Hirayama, M. Ainoya, A. Kinoshita, A. Hirata, and Y. Aoyagi, Appl. Phys. Lett. 80, 2057 (2002).

302

F. Yun

53. T. Akasaka, T. Nishida, Y. Taniyasu, M. Kasu, T. Makimoto, and N. Kobayashi, Appl. Phys. Lett. 83, 4140 (2003). 54. T. A. Lafford, P. J. Parbrook, and B. K. Tanner, Appl. Phys. Lett. 83, 5434 (2003). 55. J. Biasing, A. Reiher, A. Dadgar, A. Diez, and A. Krost, Appl. Phys. Lett. 81, 2722 (2002). 56. J. Jasinski, Z. Liliental-Weber, L. He, M. A. Reshchikov, H. Morkoc, "Comparison of AlxGai.xN films grown on sapphire by MBE under N-rich and Ga-rich conditions", Presented at the 4th Symposium on Non-Stoichiometric III-V Compounds, 2-4 October 2002, Asilomar Conference Grounds, Pacific Grove, California. 57. S. Einfeldt, M. DieBelberg, H. Heinke, D. Hommel, D. Rudloff, J. Christen, R. F. Davis, J. Appl. Phys. 92, 118 (2002). 58. S. J. Hearne, J. Han, S. R. Lee, J. A. Floro, D. M. Follstaedt, E. Chason, and I. S. T. Tsong, Appl. Phys. Lett. 76, 1534 (2000). 59. M. F. Wu,S. Yao, A. Vantomme, S. M. Hogg, G. Langouche, J. Li, and G. Y. Zhang, J. Vac. Sci. Technol. B 17, 1502 (1999). 60. A. Bell, R. Liu, U. K. Parasuraman, F. A. Ponce, S. Kamiyama, H. Amano, and I. Akasaki, Appl. Phys. Lett. 85, 3417 (2004). 61. R. Liu, A. Bell, F. A. Ponce, H. Amano, I. Akasaki, and D. Cherns, Phys. Status SolidiC 0,2136 (2003). 62. M. Kuball, F. Demangeot, J. Frandon, M. A. Renucci, H. Sands, D. N. Batchelder, S. Clur, and O. Briot, Appl. Phys. Lett. 74, 549 (1999). 63. S. Rajasingam, A. Sarua, M. Kuball, A. Cherodian, M. J. Miles, C. M. Younes, B. Yavich, W. N. Wang, and N. Grandjean, J. Appl. Phys. 94, 6366 (2003). 64. A. F. Wright, J. Appl. Phys. 82, 2833 (1997). 65. J.-M. Wagner and F. Bechstedt, Phys. Rev. B 66, 115202 (2002). 66. K. Shimada, T. Sota, and K. Suzuki, J. Appl. Phys. 84, 4951 (1998). 67. F. Demangeot, J. Frandon, M. A. Renucci, O. Briot, B. Gil, and R. L. Aulombard, Solid State Commun. 100, 207 (1996). 68. V. Yu. Davydov, N. S. Averkiev, I. N. Goncharuk, D. K. Nelson, I. P. Nikitina, A. S. Polkovnikov, A. N. Smirnov, M. A. Jackobson, and O. K. Smirnova, J. Appl. Phys. 82, 5097 (1996). 69. T. Prokofyeva, M. Seon, J. Vanburskirk, M. Holtz, S. A. Nikishin, N. N. Faleev, H. Temkin, and S. Zollner, Phys. Rev. B 63,125313 (2001). 70. A. Sarua, M. Kuball, and J. E. van Nostrand, Appl. Phys. Lett. 81,1426 (2002). 71. F. Demangeot, J. Frandon, P. Baules, F. Natali, F. Semond, and J. Massies, Phys. Rev. B 69, 155215(2004). 72. A. Sarua, M. Kuball, and J. E. Van Nostrand, Appl. Phys. Lett. 85, 2217 (2004). 73. D. Korakakis, K. F. Ludwig, and T. D. Moutakas, Appl. Phys. Lett. 71, 72 (1997). 74. P. Ruterana, G. De Saint Jores, M. Laugt, F. Omnes, and E. Bellet-Amalric, Appl. Phys. Lett. 78, 344 (2001).

Growth and Electrical/Optical Properties ofAlfiaj^N

303

75. M. K. Behbehani, E. L. Piner, S. X. Liu, N. A. El-Masry, and S. M. Bedair, Appl. Phys. Lett. 75, 2202 (1999). 76. K. Zhu, M. L. Nakarmi, K. H. Kim, J. Y. Lin, and H. X. Jiang, Appl. Phys. Lett. 85, 4669 (2004). 77. A. Y. Polyakov, M. Shin, J. A. Freitas, M. Skowronski, D. W. Greve, and R. G. Wilson, J. Appl. Phys. 80, 6349 (1996). 78. A. Y. Polyakov, N. B. Smirnov, A. V. Govorkov, M. G. Mil'vidskii, J. M. Redwing, M. Shin, M. Skowronski, D. W. Greve, and R. G. Wilson, Solid-State Electron. 42, 627 (1998). 79. J. Li, K. B. Nam, J. Y. Lin, and H. X. Jiang, Appl. Phys. Lett. 79, 3245 (2001). 80. M. D. McCluskey, N. M. Johnson, C. G. Van de Walle, D. P. Bour, and M. Kneissl, Phys. Rev. B 80,4008 (1998). 81. C. G. van de Walle, Phys. Rev. B 57, 2033 (1998). 82. C. Stampfl, J. Neugebauer, and C. G. Van de Walle, Mater. Sci. Eng., B 59, 253 (1999). 83. M. C. Wagener, G. R. James, and F. Omnes, Appl. Phys. Lett. 83, 4193 (2003). 84. J. Hwang, W. Schaff, L. Eastman, S. Bradley, L. Brillson, D. Look, J. Wu, W. Walukiewicz, M. Furis, and A. Cartwright, Appl. Phys. Lett. 81, 5192 (2002). 85. R. Zeisel, M. W. Bayerl, S. T. B. Goennenwein, R. Dimitrov, O. Ambacher, M. S. Brandt, and M. Stutzmann, Phys. Rev. B 61, R16283 (2000). 86. C. Van de Walle, Phys. Rev. B 57, R2033 (1998). 87. H. Lee, M. Gershenzon, and B. Goldenberg, J. Electron. Mater. 20, 621 (1991). 88. C. Skierbiszeski, T. Suski, M. Leszczynski, M. Shin, M. Skowronski, M. Bremser, and R. Davis, Appl. Phys. Lett. 74, 3833 (1999). 89. S. T. Bradley, S. H. Goss, L. J. Brillson, J. Hwang, and W. J. Schaff, J. Vac. Sci. Technol. B 21, 2558 (2003). 90. A. Y. Polyakov, N. B. Smirnov, A. V. Govorkov, N. V. Pashkova, S. J. Pearton, J. M. Zavada, and R. G. Wilson, J. Vac. Sci. Technol. B 21, 2500 (2003). 91. A. Y. Polyakov, A. S. Usikov, B. Theys, N. B. Smirnov, A. V. Govorkov, F. Jomard, N. M. Shmidt, and W. V. Lundin, Solid-State Electron. 44, 1971 (2000). 92. T. Matsuoka, MRS Internet J. Nitride Semicond. Res. 1, 11 (1996). 93. K. C. Zeng, K. C. Zeng, J. Y. Lin, and H. X. Jiang, Appl. Phys. Lett. 76, 1728 (2000). 94. A. Cremades, M. Albrecht, J. Krinke, R. Dimitrov, M. Stutzman, and H. P. Strunk, J. Appl. Phys. 87, 2357 (2000). 95. P. Perlin, V. Iota, B. A. Weinstein, P. Wisniewski, T. Suki, P. G. Eliseev, and M. Osinski, Appl. Phys. Lett. 70, 2993 (1997). 96. Y. Narukawa, Y. Kawakami, M. Funato, S. Fujita, S. Fujita, and S. Nakamura, Appl. Phys. Lett. 70, 981 (1997). 97. S. J. Chung, M. S. Kumar, H. J. Lee, and E.-K. Suh, J. Appl. Phys. 95, 3565 (2004).

304

F. Yun

98. C. J. Collins, A. V. Sampath, G. A. Garrett, W. L. Sarney, H. Shen, M. Wraback, A. Yu. Nikiforov, G. S. Cargill III, and V. Dierolf, Appl. Phys. Lett. 86, 031916 (2005). 99. L. Bergman, X.-B. Chen, D. Mcllroy, and R. F. Davis, Appl. Phys. Lett. 81, 4186 (2002). 100. Y. Yamada,a) Y. Ueki, K. Nakamura, T. Taguchi, Y. Kawaguchi, A. Ishibashi, and T. Yokogawa, Appl. Phys. Lett. 84, 2082 (2004). 101. N. Antoine-Vincent, F. Natali, M. Mihailovic, A. Vasson, J. Leymarie, P. Disseix, D. Byrne, F. Semond, and J. Massies, J. Appl. Phys. 93, 5222 (2003). 102. G. Tamulaitis, I. Yilmaz, M. S. Shur, Q. Fareed, R. Gaska, and M. A. Khan, Appl. Phys. Lett. 85, 206 (2004). 103. Q. Fareed, R. Gaska, M. S. Shur, J. Wu, W. Walukiewicz, and M. A. Khan, Appl. Phys. Lett. 84, 1892 (2004). 104. J. Zhang, E. Kuokstis, Q. Fareed, H. Wang, J. Yang, G. Simin, M. Asif Khan, R. Gaska, and M. Shur, Appl. Phys. Lett. 79, 925 (2001). 105. F. Yun, M. A. Reshchikov, L. He, T. King, H. Morkoc, S. W. Novak, and L. Wei, J. Appl. Phys. 92, 4837 (2002). 106. Y. Koide, H. Itoh, M. R. H. Khan, K. Hiramatsu, N. Sawaki, and I. Akasaki, J. Appl. Phys. 61,4540(1987). 107. I. Akasaki and H. Amano, in GaN, edited by J. I. Pankove and T. D. Moustakas (Academic, New York, 1998), Vol. 1, pp 459-72. 108. H. Jiang, G. Y. Zhao, H. Ishikawa, T. Egawa, T. Jimbo, and M. Umeno, J. Appl. Phys. Lett. 89, 1046 (2001). 109. S. R. Lee, A. F. Wright, M. H. Crawford, G. A. Petersen, J. Han, and R. M. Biefeld, Appl. Phys. Lett. 74, 3344 (1999).

CHAPTER 9 OPTICAL INVESTIGATION OF InGaN/GaN QUANTUM WELL STRUCTURES GROWN BY MOCVD

Tao Wang Department of Electronic and Electrical Engineering University of Sheffield, Mappin Street, Sheffield, SI 3JD, United Kingdom E-mail: [email protected] The emission mechanisms of InGaN/GaN quantum well structures have been investigated under conditions of a low excitation and a high excitation, corresponding to spontaneous and stimulated emission processes, respectively. The spontaneous emission is dominated by quantum confined Stark effect and exciton-localization effect, which are strongly affected by indium composition and well thickness as well as strain relaxation, whereas, in contrast, InGaN/GaN quantum wells in the process of stimulated emission behaves in the same manner as the classical AlGaAs/GaAs system. Subpicosecond time-resolved differential transmission spectroscopy has been used to investigate the carrier density and temperature dependence of the quantum well electron capture time of InGaN/GaN MQW structures. The capture time varies significantly with both temperature and carrier density, the latter effect being consistent with carrier-induced band bending or increased carriercarrier scattering. At room temperature, the electron capture time is in the range 0.4-0.8 ps for carrier densities < 5xl0 18 cm'3. 1.

Introduction

InGaN-based light emitting diodes (LEDs) with performances superior to the conventional AlGaAs or AlGalnP-based LEDs have been commercialized in spite of high dislocation density up to 101 /cm due to

305

306

T. Wang

an extremely large lattice-mismatch (-16%) between sapphire and GaN. However, InGaN-based laser diodes (LDs) on sapphire substrates without any additional technology to decrease this intrinsic highdislocation-density can be demonstrated only under a pulsed mode or under a continuous wave (cw) mode but with a short life-time due to extremely high threshold current. Only grown on the low-dislocationdensity substrate obtained by additional technology can InGaN-based LDs operate for more than 10000 h. ' Therefore, it is necessary to investigate the emission mechanisms for the spontaneous and stimulated processes occurring in InGaN alloy. A series of work 2"12 for investigation of this issue from the optical view have resulted in successful development of both a 410 nm InGaN-based laser grown by an atmospheric horizontal metalorganic chemical vapour deposition (MOCVD) at the University of Tokushima (Japan) in 2000,13 and a 425 nm InGaN-based laser grown by a low-pressure vertical multiple-wafer MOCVD at the University of Sheffield (UK) in early 2004.14 Both nitride lasers are working under electrical injection current with a good threshold current density at room temperature. Normally, a InGaN-based near-UV/violet/blue/green LED chip with a typical size of 350x350 |J.m2 generally operates under 20-50 mA injection current, giving a density of the order of a few A/cm2 injection current density, whereas, in contrast the threshold current density for lasing is generally in the order of a few to tens kA/cm2, around three orders of magnitude higher than that for LEDs. Similarly, a high power pulse laser with a femtosecond scale pulse width is necessary to generate stimulated emission, while a cw He-Cd in the order of W/cm2 power density is used to study the spontaneous emission. In this chapter, we study fundamental optical processes of InGaN/GaN quantum well structures under the conditions of a low and a high excitation power, corresponding to the spontaneous emission and stimulated emission processes. Generally, InN and GaN are far from being lattice-matched. Furthermore, the piezoelectric constants of Ill-Nitride alloys are intrinsically large, compared to other III-V semiconductors. In a strained case, a piezoelectric field exerted across InGaN quantum well is self-

InGaN/GaN MQW MOCVD-Optical Investigation

307

produced, which results in quantum confined Stark effect (QCSE).2"12'15" However, under a high optical excitation that is generally required to achieve a stimulated emission, the piezoelectric field can be strongly screened. Therefore, the emission mechanism that is dominated by QCSE becomes weak or disappears. Secondly, a low miscibility between InN and GaN leads to a large fluctuation of indium or so-called phase separation in InGaN alloys, depending on indium composition and well thickness, namely, exciton localization effect. " • Currently, exciton localization effect is accepted to play an important role in high quantum efficiency of blue/green LEDs. One of the evidences is that the quantum efficiency decreases as a result of reducing the indium concentration.39 It is expected that exciton localization effect can affect optical properties of InGaN alloy in a different manner under a high excitation. This chapter treats these fundamental issues. First section studies the strain-relaxation in InGaN/GaN multiple quantum wells in order to establish the relationship between the stain-relaxation and optical properties. This investigation is carried out under a low excitation. Section II investigates the influence of QCSE and exciton localization effect on the performance of InGaN-based LED, aiming at understanding the mechanism that dominates spontaneous emission. The exciton behavior of InGaN/GaN multiple quantum wells under condition of a high excitation is studied in Section III and IV, and the relationship between threshold for stimulated emission and well thickness is established. Femtosecond study of electron capture times in InGaN/GaN multiple quantum wells is briefly introduced in Section V. Final section summaries the chapter. 25

2. Strain-relaxation in InGaN/GaN MQW The investigated samples in this section are InGaN/GaN multiple quantum well (MQW) structures with 2, 3, 5 or 10 periods, denoted as 2QW, 3QW, 5QW and 10QW, respectively. All samples were grown under identical conditions in rapid succession to ensure their uniformity as far as possible. The samples were grown using an atmospheric horizontal metalorganic chemical vapour deposition (MOCVD). In each case, 4 nm InxGai_xN well was separated by 9 nm GaN barriers with different periods mentioned above. Indium concentration is measured to

308

T. Wang

be 13% based on a bowing factor of 3.2eV. The thickness of well and barrier was measured by transmission electron microscopy (TEM). Because there are no any satellite peaks in the x-ray diffraction (XRD) pattern of a single-quantum-well structure, it is difficult to compare a SQW structure with a MQW structure. Therefore, we start the investigation from a MQW sample with two-period, namely, 2QW. Figure 1 shows the photoluminescence (PL) spectra of these samples, measured at 10K under an excitation of a 0.5 mW He-Cd laser. In each sample, there appears a strong and narrow emission peak from InGaN well region, which is accompanied by a weak low-energy peak due to the phonon replica. With increasing quantum-well number, the main peak shows a clear blue-shift. In particularly, when the quantumwell number rises from 2 to 3, the emission energy shows a large shift. With further increasing quantum-well number, the increase of the emission energy becomes slow. For examples, the blue shift of emission energy between 2QW and 10QW is around 89 meV, while that between 3QW and 10QW is about 16 meV. I-"

1

»T—1

•

T=10K, P=0.5mW 10QW 5QW

3QW 2QW 2.4

2.5

^

A\ \ i

-—-9 /

2.6 2.7 2.8 Photo energy(eV)

2.9

3.0

Fig. 1. Photoluminescence spectra, measured at 10K under an excitation of a 0.5 mW. The quantum well number in the four samples is 2, 3, 5 and 10, respectively. (From Ref.8 with permission of reprinting from American Institute of Physics).

Figure 2 (a) and (b) shows the excitation-power dependent PL spectra of 2QW and 10QW sample as two extreme cases, respectively. Both are measured at 10K under different excitation powers. The emission energies in both samples show clear blue-shift when the

InGaN/GaN MQW MOCVD-Optical Investigation

309

excitation power increases from 0.5 mW to 50 mW; however, by comparing the blue-shift between 0.5 mW and 50 mW excitation power, the blue-shift for 2QW sample is found to be around 154 meV, while that for 10QW is only 19 meV. In order to show the relationship between number of quantum well and blueshift clearly, the energy shift for all the samples is plotted as a function of quantum-well number under a low excitation-power (0.5 mW) and a high excitation-power (50 mW), as shown in Figure 3, in which the blue-shift becomes small with increasing quantum-well number, in particular, the blue-shift for 2QW is much larger than those for other three samples.

rmalized Intent

•£

£

2QW T=10K

A

(a)

50rrt/V/

20rrt/V___^W/

ZSnW^yj 1rrW___y/ 0.EhW/ ^

\ \

2.5 2.6 2.7 2.8 2.9 3.0 3.1 2.5 2.6

Energy(eV) Fig. 2. Excitation-power dependent photoluminescence of InGaN/GaN MQW with 2 (a) and 10 (b) periods at T=10K. (From Ref.8 with permission of reprinting from American Institute of Physics).

The blue-shift in InGaN/GaN quantum well structure generally results from QCSE due to the large lattice-mismatch between InGaN and GaN, mentioned above. As we knew, the strain-induced polarization field tilts the potential profile, and thus the energy of optical transition is reduced by the amount eEp2d (Epz is the piezoelectric field, e the electron

T. Wang

310

charge, and d the QW thickness), which results in a red-shift, 41'42 that is, QCSE. When the samples are irradiated with an excitation source, the piezoelectric fields in the strained InGaN well layer can be screened due to photo-pumped carriers, weakening the QCSE. Increasing the excitation intensity further weakens the QCSE, and then increases the transition energy. That is, the blue-shift occurs. However, strain-relaxation can modify the QCSE, which results in the shift of the emission energy. Based on the model of piezoelectric field induced QCSE, the strain-relaxation can be quantitatively estimated. According to a previous calculation, 15 a 4 nm Ino.13Gao.g7N/GaN, under assumption that there is no any strain-relaxation, should show about a 150 meV blueshift under a high excitation power, compared with the case under a low excitation power (i.e., QCSE is screened out). By carefully examining the data in Fig 2, it can be found that the blue-shift of emission energy for 2QW sample is about 154meV, which is almost the same as the above calculated value. Therefore, it is very reasonable to assume that the 2QW sample should be fully strained.

2.90 T=10K > 2.85 >.

^••->^P=50mW • •

•

s

•

c 2.80

T

0) c 0

P=0.5mW

**

c 2.75

2 •

2.70

2

4 6 8 Quantum well number

10

Fig. 3. Emission energy of InGaN/GaN MQW as a function of quantum-well number under a lower excitation power (0.5mW) and a high excitation power (50mW), measured at 10K. (From Ref.8 with permission of reprinting from American Institute of Physics).

InGaN/GaN MQW MOCVD-Optical Investigation

311

In the next step, we would like to estimate the strength of the straininduced piezoelectric field Epz in 3QW, 5QW and 10QW InGaN/GaN in terms of percentage, compared with 2QW sample which is assumed as fully strained case discussed above. The estimation is also based on the piezoelectric field induced QCSE. Under the first order approximation, the emission energy shift is proportional to the piezoelectric field strength. If 2QW is assumed fully strained, the percentage of the residual piezoelectric field strength in 3QW and 5QW as well as 10QW can be estimated, which is based on the shift of emission energy shown Fig.2. Consequently, the piezoelectric field strength of 3QW, 5QW and 10 QW is calculated to reduce to about 51%, 46% and 40%, respectively. Therefore, the residual strain occurring in 3QW, 5QW and 10QW can be estimated, based on the relationship between strain-induced piezoelectric field Epz and in-plane strain exx given by 15 *,Z=-^-[—*33-*3l]-*„ £r • £0

CD

C 33

where £r, £0 and ey are the dielectric constant of the material, the permittivity of free space and the piezoelectric constant. Based on the equation (1), the residual strain £xx is calculated to be 51%, 46%, 40% in 3QW, 5QW and 10QW, respectively compared with 2QW that is assumed fully strained. These data will be used for the X-ray kinetic simulation to compare with our measured XRD data. Figure 4 shows the measured XRD spectra of all the samples in (0006) 20-co mode, and also gives the simulated XRD patterns (dashed lines) based on X-ray kinetic theory, which will be explained in the following. In all cases, there appear clear satellite peaks of InGaN/GaN MQWs in addition to an identical GaN diffraction peak. Even in the case of 2QW, the satellite peaks can be clearly observed, indicating high quality of these InGaN/GaN MQWs. Comparing the positions of satellite peaks of 2QW and other samples, a clear shift can be found. However, the shifts in 3QW, 5QW and 10QW are very small, while in contrast, the shift shown in their PL spectra in Figure 2 can be observed clearly. Therefore, it indicates that PL measurement is more sensitive method to examine the strain-relaxation than XRD.

312

T. Wang

I

-i^--^

|

""I—T

I

—I""1"™"

I

•

I

•

Fig. 4. Measured and simulated XRD patterns of InGaN/GaN MQW structures based on (0006) 26-co mode. The quantum-well number of InGaN/GaN MQW is 2, 3, 5 and 10, respectively. The dash lines correspond to the simulated data. (From Ref.8 with permission of reprinting from American Institute of Physics).

The XRD simulation method is based on the XRD kinetic theory,40 which has been well used in semiconductor superlattices and quantum well structures. The scattered amplitude at angle 0 can be written as 43 M

F(0) = XfjXeiQd'

(2)

and

fj=flJ+f2jeiQdi where Q-^E^^1,

f •, flj,

fl.

(3) and dj are the complex scattering

A

factor of the jth monolayer, the scattering factor of N atom, the scattering factor of Ga (or In) atom and the spacing between N atomic layer and Ga (or In) atomic layer. The thickness of well and barrier is 4 nm and 9 nm from TEM measurement. The indium concentration is measured to be 13% based on the bowing factor of 3.2 eV mentioned as above. In our simulation, a full strain for 2QW, and 51%, 46%, 40% residual strain for 3QW, 5QW and

InGaN/GaN MQW MOCVD-Optical Investigation

313

10QW are used, which corresponds to in-plane strain £** as discussed above. Consequently, the strain element ^ along z-direction (0001) can be obtained thorough following equation: 19 2cn £

=

ZZ

—• £

C

XX

(4)

V

/

33

where exx is the in-plane strain, ezz is the perpendicular strain which is directly related to our XRD data. C33 and C i3 are elastic constants.15 Furthermore, based on the following equation, the lattice-constant along z-direction can be obtained:

where C and co are the z-direction lattice constant of the epitaxial layer and free-standing layer, respectively. The simulated XRD curves plotted in Figure 4, denoted as dashed lines, are in good agreement with the experimental data, which in turn supports the discussion on the PL results. 3. Quantum-confined Stark Effect and Exciton-localization Effect The pioneering work of QCSE started from AlGaAs/GaAs quantum well structures by Miller et al, 41'42 who found this distinct physical effect when the electric field was exerted across the AlGaAs/GaAs quantum well structure. One of important consequences of QCSE is peak linewidth broadening. Generally, the linewidth of PL emission from InGaN/GaN quantum well structure is quite broad with a full width at half maximum (FWHM), typically in the range of 50-100 meV due to poor crystal quality resulting from the currently limited technology and indium fluctuation. Therefore, the peak linewidth broadening due to QCSE is often ignored. Secondly, it is also difficult to distinguish the different origins for peak linewidth broadening since high quality InGaN/GaN samples are required to minimize the possibilities for this broadening. With high quality InGaN/GaN single quantum well structure, the peak linewidth broadening can be observed. Figure 5(a) show the PL spectra of InGaN/GaN single quantum well (SQW) structure under different excitation powers, measured at 2IK.

T. Wang

314

Figure 5(b) shows the excitation power dependence of FWHM of emission peak, in which the FWHM decreases from 39 meV to about 33.5 meV when the excitation power is raised to 37 mW from 1.6 mW. This is the narrowest report until now in the InGaN/GaN material system. This sample was grown by an atmospheric horizontal MOCVD, and the well thickness is 2.5 nm, typical value for InGaN-based optical devices. In terms of FWHM, the crystal quality of SQW should be higher than that of MQW, which can be well understood from the view of growth. Generally, the growth temperature of InGaN is much lower than that of GaN. The first InGaN well is deposited on the surface of high quality GaN layer, which is grown at high temperature over than 1000°C. But the subsequent InGaN wells are deposited on the GaN layer grown at the low temperature, which is the same as or little higher than the growth temperature of InGaN (i.e. 700-800°C). In addition, during MQW growth, N2 as a carrier gas was used instead of H2 in order to enhance indium incorporation. All of these unavoidably decrease the quality of GaN on which subsequent InGaN wells were deposited. All these result in the degradation of the quality of subsequent InGaN wells.

Energy (eV)

Exciation Power (mW)

Fig. 5. (a) PL spectra of InGaN/GaN single quantum well structure under different excitation powers, measured at 21K (b) FWHM of the SQW as a function of excitation power, indicating FWHM decreases with increasing excitation power. (From Ref.2 with permission of reprinting from American Institute of Physics).

InGaN/GaN MQW MOCVD-Optical Investigation

315

Since the crystal quality of SQW is higher than that of MQW, SQW as an active region was used for LEDs. In this section, the influence of QCSE and exciton-localization effect on the performance of LEDs is investigated, which is based on the study of the LEDs using SQW with different well thickness as an active region. Two series of InGaN/GaN-based LED structures are used, one consisting of four samples, in which the indium-mole-fraction of InGaN well layer is around 23% (based on 1 eV of bowing parameter, 11.5% if a bowing parameter of 3.2 eV is used) and the well thickness is 1.5 nm, 2.5 nm, 4 nm and 5 nm, respectively, (denoted by LED A, LED B, LED C and LED D), and another LED structure (denoted by LED E), which contains about 10% indium (based on 1 eV of bowing parameter, 5% if a bowing parameter of 3.2 eV is used) in the InGaN well layer with 2.5nm. Except InGaN/GaN SQW, a 3 |J.m thick layer of heavily doped n-GaN were grown before InGaN/GaN SQW and a 0.3 |J,m Mg doped GaN is finally grown. In order to avoid the interference effect of top p-type GaN layer on the PL measurement, the other 5 samples are also grown, in which a 100 nm undoped GaN layer is used instead of 0.3 urn Mg dopedGaN. For simplicity, these 5 samples are labeled A, B, C, D, E, corresponding to LED A, LED B, LED C, LED D and LED E, respectively. All the investigated samples in this section were grown by an atmospheric horizontal MOCVD. Firstly, the temperature-dependent PL measurements are carried out on the sample A, B, C, D, E to investigate so-called exiton localization effect as a function of the indium concentration or well-thickness. Based on the band-tail model, the temperature-dependent emission energy could be described by the following expression.29

E(T) = E(0)—^~—^— T + fi KBT

(6)

The first term describes the energy gap at zero temperature; a and J3 are known as Varshini's fitting parameters. The third term comes from the localization effect, in which c indicates the degree of localization effect, i.e., the large value of a means a strong localization effect. KB is Bolzmann's constant. In addition, this model is based on the assumption of non-degenerate occupation.

316

T. Wang

2.9

3.0

Energy (eV)

3.1

3.2

2.8

3.0

3.1

3.2

Energy (eV)

Fig. 6. Temperature-dependent of PL spectra of Ino.23Gao.77N/GaN SQW structure (sample A) with 1.5nm well and the In0.i0Gao.9oN/GaN SQW with 2.5nm well (sample E). The emission energy of sample A firstly decreases with increasing temperature up to about 50K, and then increases with further increasing temperatures up to around 130K. After that, it decreases monotonically with increasing temperature. The emission energy of sample E monotonically decreases from 10k to RT. The solid circles are guide to eyes. (From Ref.7 with permission of reprinting from American Institute of Physics).

As represented cases, Figure 6 shows the temperature-dependent PL spectra of Ino.23Gao.77N/GaN SQW with 1.5 nm well (i.e., Sample A) and Ino.1Gao.90N/GaN with 2.5 nm well (i.e., Sample E). The emission energy of sample A decreases with increasing temperature up to 50 K, and then increases with further increasing temperature up to 130 k. After that, the emission energy decreases with increasing temperature. In contrast with it, the emission energy of sample E decreases monotonically with increasing temperature. Figure 7 shows the emission energy as a function of temperature for samples A, B, C, D, and E, respectively. Except for samples D and E, there appears a temperature-induced blueshift of emission energy for all

317

InGaN/GaN MQW MOCVD-Optkal Investigation

other samples at temperatures higher than 50 K, which is a fingerprint of the exciton localization effect. 2.6 ln

o.23Gao.77N/GaN

lno.23Gao.77N/GaN

d=5 nm

d=1.5 nm ;3.oo • E W2.99

E(0)=3.035eVa=15.4meV a=0.445meV/kp=830K 100 200 Temperature (K)

Ino^Gao^N/GaN d=2.5 nm.

E(0)-2.85eV 0=19.5 meV o=1.23meV/1
E(0)-3.118eV a=0meV a=0.47 meV/K p=900K

100 200 Temperature (K)

105 205 Temperature (K)

lno.23Gao.77N/GaN §2.55

o2.50

d=4nm

E(0)=2.591eVa=9.51 meV a=1.6meV/kp=900K 100 200 Temperature (K)

300

Fig. 7. Emission energy vs temperature for SQW structure with (a) 1.5nm Ino.23Gao.77N well (b) 2.5nm Ino.23Gao.77N well, (c) 4nm Ino.23Gao.77N well and (d) 5nm Ino.23Gao.77N well (e) 2.5nm Ino.10Gao.90N well. The solid lines are fitting curves based on the band-tail model, and the fitting parameters are also given in each figure. (From Ref.7 with permission of reprinting from American Institute of Physics).

The fitting is made based on Eq. (8) in each case, and the fitting parameters are also given in Fig. 7. These parameters are obtained based on several samples grown under identical conditions, yielding average values. In Fig. 8, the a value increases with increasing quantum well thickness up to 2.5 nm and then decreases with further increasing well thickness in the case of 23% indium mole fraction. Sample B (i.e., 2.5 nm well) shows the strongest exciton localization effect. There is no distinguished exciton localization effect to be observed in sample E due

318

T. Wang

to the lower indium mole fraction (-10%), although the quantum well thickness of sample E is also 2.5 nm. 32

3.0 JJ

•—DD

DD—DC

25nmltl).10Q%.90N<»J

lrv 23 G% 77 N«GaN

-•-t5hm , -m-2am —%—irm -O-am

«?

Si

00

Biergy

3

w

-•

p_

UOjSSj

IS

fi« >

'8^

^

10 Excitation Power (mW)

' 1

10 Excitation power (mW)

30

Fig. 8. Emission energy as a function of excitation power for sample A, B, C, D in (a) and sample E in (b), measured at 10K. (From Ref.7 with permission of reprinting from American Institute of Physics).

Figure 8 depicts the emission energy as a function of excitation power for all samples measured at 10K, which is generally used to investigate QCSE. The emission energy of sample A is almost independent of the excitation power, while sample B shows a slight blue shift of emission energy with increasing the excitation power. With further increasing quantum well thickness, the blue-shift optically induced is greatly enhanced. In the case of 10% Indium concentration (i.e. sample E), there is no any distinguished blue-shift optically induced to be observed, as shown in Figure 8(b). In addition, the study in section V indicates that the PL intensity decreases quickly with increasing well thickness due to QCSE. Consequently, from the magnitude of the blueshift measured under identical conditions, we can judge strength of QCSE existing in these samples. Therefore, the QCSE in the sample A

InGaN/GaN MQW MOCVD-Optical Investigation

319

and sample E is so weak that it can be safely ignored, while the strongest QCSE is shown in sample D. Based on above experimental results, in the case of 23% indiummole-fraction, QCSE can be monotonically enhanced by increasing wellthickness, while the exciton localization effect shows the strongest in the sample with 2.5 nm well (i.e. sample B). In the case of 10% indium concentration and 2.5nm well (i.e. sample E), there is no distinguished QCSE and excitation localization effect to be observed.

•

50 -

g-4o -

3. a 30 -

o Q.

~

20 •

/

Q.

O 10 -

•

ln 0 , 10 Ga 0 . 90 N'GaNLED

—•

— • — ln 0 23 Ga„ 7 7 WGaN LHJ

1

2

3

4

5

Quantum well thickness (nm) Fig. 9. Output power of LED A, B, C, D, E, which are measured under the condition of 20 mA injection current at room temperature. (From Ref.7 with permission of reprinting from American Institute of Physics).

Next, we compare the output power of LED samples, which has same active layer structures as samples A, B, C, D and E, respectively. Figure 9 shows the dependence of the output power on quantum well thickness, which is measured under the condition at 20 mA injection current at room temperature. In the case of 23% Indium, the output power decreases with increasing well-thickness, which is in good agreement with the well-thickness dependent PL intensity. Combined with Figure 8(b), one can understand that the decrease of output power with increasing well-thickness can be attributed to QCSE. Firstly, we consider the case of the influence of well thickness on optical power. Although LED B shows the stronger exciton localization

320

T. Wang

effect than LED A in Figure 9, the output power of LED B is only half of the LED A output power. Furthermore, Figure 8 indicates that the QCSE of LED B is stronger than that of LED A since the well thickness in LEDB is thicker than that in LEDA. It means that QCSE plays more important role than the exciton localization effect. Secondly, we compare both QCSE and exciton localization effect on optical power. The output power of LED B is lower than that of LED E. In this case, on the one hand, Figure 7 indicates that the exciton localization effect in LED B is further stronger than that in LED E, on the other hand, Figure 8 indicates that the QCSE in LED B show stronger than that in LED E. Therefore, the lower output power of LED B than that of LED E means that QCSE has stronger influence on LED performance than the exciton localization effect, which confirms the above conclusion once again. Of course, if QCSE is weak enough to be safely ignored, the performance of LED can be improved by the exciton localization effect, which can be understood from the comparison of the output power between LED A and E. From Figure 7, the exciton localization effect in sample A is much stronger than that in sample E, while Figure 8 shows that QCSE in both samples is very weak. Correspondingly, the output power of LED A is larger than that of LED E in Figure 9, which means that the exciton localization effect can be helpful for the enhancement of LED output power if QCSE is weak enough. Therefore, the above conclusion should be taken into account in designing InGaN-based blue/green LEDs. 4. Optical Investigation of InGaN/GaN MQWS under High Excitation The investigated samples in this section were grown on (0001) sapphire substrates by a low-pressure metalorganic chemical vapor deposition (LP-MOCVD) system. All the samples consist of 10-period Ino.nGao.sgN/GaN MQW. In each case, the barrier is 7.5 nm thick, but the quantum well thickness is 1.4, 1.9, 2.4, 3.3 and 3.9 nm, denoted as sample A, B, C, D and E, respectively. The thickness and indium mole fraction were determined by X-ray diffraction using calibration samples. PL measurements under high excitation were performed using a regeneratively amplified femtosecond Ti: sapphire laser with a repetition

InGaN/GaN MQW MOCVD-Optical Investigation

321

rate of 1 kHz operating around 800 nm. The 130-fs pulses were tripled to provide a 267 nm excitation source. The sample was placed into a flowing helium gas cryostat with a temperature controller ranging from 5 K to room temperature. The luminescence was dispersed by a 0.55 m monochromator, and was detected by CCD camera cooled by liquid nitrogen. 3.3 (a)

|

T=5K

A

P=500uJ'cm2

•

measured calculation

3.2

«, I 2 - 3.0

1

2.5 3.0 3.5 4.0 Emission Energy (eV)

1

-J

2 3 4 Well thickness (nm)

CO

• 2.9

Fig. 10. (a) PL spectra of sample A, B, C, D and E under high excitation, measured at 5K. The emission peaks denoted by the arrow correspond to the transition between the first electron and first heavy-hole subbands (el-hi). In the sample E, there is another peak around 3.02 eV, attributed to quantum-dot-like states. In the sample D, there is also a peak from quantum-dot-like states, appearing as a shoulder on the high-energy side of the el-hi peak. The peaks at 3.18 eV in sample D and E correspond to stimulated emissions. (From Ref.9 with permission of reprinting from American Institute of Physics).

Figure 10(a) shows the 5 K PL spectra of these samples under an excitation power density of 500 (aJ/cm2. In each spectrum, there appears a strong emission peak indicated by an arrow, which decreases in energy as the quantum well thickness increases. In addition to these emission peaks, in sample D and E, there is another peak in sample D and E appearing at 3.18 eV due to stimulated emission. A detailed investigation on the stimulated emission will be made in next section. These arrowed emission peaks in Figure 10(a) are due to the transition between the first

322

T. Wang

electron and first heavy-hole subbands denoted by el-hl, based on our calculation given in Figure 10(b). Under such high excitation, the sheet carrier density is in the order of 1013/cm2. n If a two dimensional screen length X=2ekT/e2n (where n is the carrier density, £ is a static permittivity, and other symbols have their usual meaning) was chosen,44 giving X«hz (where Lz is well thickness). This means that the piezoelectric field should be completely screened out, and can be ignored safely. Our calculation of the transition energy between the first electron and first heavy-hole subbands was based on a single particle model using a band-offset ratio of AEC: AEV=0.54:0.46 (Ref.50). The material parameters used in this calculation can be found in the review paper given by Vurgaftman et al.45 Figure 10(b) shows a good agreement with the measured values obtained from Figure 1(a). However, in the case of sample E, there is another strong emission peak appearing at about 3.02 eV, whose emission intensity is higher than that of the el-hl peak mentioned above. By the way, careful examination of sample D shows that there is also a shoulder appearing on the high-energy side of the elhl peak. The feature is more clearly observed under excitation-power dependent PL, where there is a clear intensity change between these two peaks. This behaves in a same way as that of sample E. However, our calculations indicate that there is no allowed transition in this energy region. In order to explore the mechanism of this feature, the excitationpower dependent PL measurements were made on all samples. Figure 11(a) and (b) show a series of spectra recorded at different excitation densities between 8.3 and 500 )jJ/cm2 for sample E and A, respectively. Sample B and C exhibit a very similar behavior to sample A, indicating that there appears only an emission from the el-hl transition for the excitation densities used in this study. For sample E, under low excitation densities, there only appears the el-hl peak. However, with increasing excitation density, the peak appears on the high-energy side of the el-hl peak, starting to show as a shoulder and then a clear peak at 3.02 eV. Finally, this peak becomes dominant over the el-hl emission. This is most likely the feature of quantum dots (QDs) due to carrier localization and more efficient recombination through QDs.46

InGaN/GaN MQW MOCVD-Optical Investigation

T

H~I—'—i—'—i—I

Brission&ergy(eV)

323

I—i—T—i—'—i—•—r

Brission Bergy(eV)

Fig. 11. Excitation power density dependent PL spectra of sample A (a) and sample E (b). (From Ref.9 with permission of reprinting from American Institute of Physics).

In order to get further evidence, the temperature dependent PL has been carried on sample E, which is given in Figure 12(a). With increasing temperature, the intensity of 3.02 eV emission decreases much more slowly compared to the el-hi peak, which is typical of quantumdot-like states.46 Furthermore, the temperature-induced peak shift of the 3.02 eV emission peak is much weaker than that of the el-hi emission peak, which can be seen clearly by comparing to the temperature dependent PL of the el-hi peak of sample A measured under identical conditions shown in Figure 12(b) . Therefore, both trends (power and temperature dependencies) point towards the 3.02 eV emission peak as originating from the quantum-dot-like states.46 Furthermore, our study indicates that there are no quantum-dot-like states to be observed in the thin well samples A, B, C, meaning that quantum-dot-like states easily appear in wide well samples.

324

T. Wang

2.6 2.8 3.0 3.2 3.4 3.6 3.8 Bnission Biergy (eV)

2.8 3.0 3.2 3.4 3.6 3.8 4.0 Bnission Biergy (eV)

Fig. 12. Temperature dependent PL spectra of sample E (a) and sample A (b) (From Ref.9 with permission of reprinting from American Institute of Physics).

The origin of quantum-dot-like states is most likely due to the growth model transition with increasing well thickness, i.e., twodimensional growth at a thin layer, and then three-dimensional growth, by which QDs can be formed. The QDs might be depleting the surrounding matrix of indium to provide a high barrier in three dimensions, which has been observed in InGaAs/GaAs system. The QDs must be very small, probably in the order of 1 nm in diameter. Previously, the QDs of a composition approaching InN with an estimated size from 0.6 to 3 nm in InGaN layer have been reported.31 Based on a single particle model, the estimated emission energy of QDs with a size of 1 nm is higher than 3 eV, approaching our data. Therefore, compared to the el-hi, the emission energy of our QDs can show a blue-shift rather than red-shift observed in other QD systems that are generally quite large. In next section, the investigation based on these series of samples indicates that wide-well sample has much lower threshold pumping for lasing compared to thin-well sample, which also proves the existence of

InGaN/GaN MQW MOCVD-Optical Investigation

325

quantum-dot-like states since the only extremely small nitride dots are predicted to have a low threshold for stimulated emission.47 Further experiments to confirm the QD behaviour has been made using micro-PL by means of a shadow mask technique, and E-beam lithography to give arrays of apertures with sizes ranging from 100 nm to a few micrometers. Micro-PL spectra measured at 4 K indicated that the emission peaks with FWHM of around 1.5 meV at the high-energy side of the el-hi peak has been observed,48 confirming the existence of quantum-dot-like states as seen in a recent paper by H Schomig et al. — i — • — i — • -

10

o

CD W ji. D

r

SanpleE

.

2

-

-•-330uJ/cm —•— 33 (iJ/cm2

I

50

100

150

200

.

I

—i—•—i—•—i—'—

Internal quantum c

•g 0.8

—i

.

250

300

1000/T(K-1)

Fig. 13. Temperature dependent internal-quantum-efficiency of sample E under low excitation (33 uJ/cm2) and high excitation (330 (iJ/cm2). (From Ref.9 with permission of reprinting from American Institute of Physics).

Figure 13 gives the internal quantum efficiency of sample E as a function of temperature under different excitation conditions (33 uJ/cm2 and 330 uJ/cm2). Generally, the internal quantum efficiency can be evaluated by the temperature dependence of the integrated PL intensity assuming that the internal quantum efficiency is unity at low temperature.26'27 Under low excitation, the internal quantum efficiency decreases very quickly with increasing temperature, finally it drops below 10% at RT. In contrast, the internal quantum efficiency under high excitation decreases slowly with increasing temperature. At RT, the internal quantum efficiency is still about 40%. Part of the reason for the

326

T. Wang

high internal quantum efficiency is that the emission is dominated by quantum-dot-like states under high excitation, which leads more efficient recombination through quantum-dot-like states. In turn, it also supports our above conclusion. Of course, one of other reasons is that the internal quantum efficiency under low excitation can also be reduced by the piezoelectric field induced QCSE, as discussed in section III. 5. Study of Stimulated Emission from InGaN/GaN Multiple Quantum Well Structures The samples used in this section are same as those in section IV. In order to achieve optically pumped stimulated emission. A high excitation is necessary, which can be obtained by a regeneratively amplified femtosecond Ti: sapphire laser with a repetition rate of 1 kHz operating around 800 nm. The 130-fs pulses were tripled to serve as the excitation source at a wavelength of 267 nm. The stripe-pumping geometry was employed, and the emission was collected from the edge of the samples, which was detected by CCD array detector. A 10 mW cw He-Cd laser was also used for the investigation as a low excitation source for comparison. Figure 14 gives the data of the stimulated emission of all samples. All samples exhibited a similar behavior. As a representative case, Figure 14(a) shows the room temperature (RT) emission spectra of sample D recorded under optical pumping from 33 |xl/cm2 to 190 |iJ/cm2. Under low optical pumping, there appears only one emission peak due to the spontaneous emission. However, when the optical pumping increases to around 120 uJ/cm2, another peak on the high-energy side appears. Moreover, with further increasing optical pumping this peak dramatically becomes narrow and very strong, indicating a stimulated emission process. The integrated PL intensity as a function of excitation density is shown in the inset of Fig 14(a), indicating that the threshold optical power is around 120 uJ/cm2. Such stimulated emission has been observed on all samples, and the threshold of optical pumping with an experimental error of ±3% is plotted as a function of well thickness in Figure 14(b), showing the threshold decreases monotonically with decreasing well thickness. All samples show good reproducibility.

InGaN/GaN MQW MOCVD-Optical Investigation

1

—

•

1

/ r

^J '

50

100

150

200

Optical pumping (uj/cm 2 )

H

R

190^J/cnf

n'

¥ 300

1

11 Integrated inter)!

Emissior itensity(a u.)

(a)

•*

327

1 33nJ/OT?

400

500

Wavelength (nm)

600

2

3

4

Quantum well thickness (nm)

Fig. 14. (a) Emission spectra of sample D under the excitation of a high-power pulse laser with different optical pumping densities, in which the stimulated emission can be clearly observed, and inset gives the integrated emission intensity as a function of the optical pumping, from which the threshold of optical pumping can be determined, (b) The stimulated emission threshold as a function of well thickness. (From Ref.10 with permission of reprinting from Elsevier)

In order to observe the stimulated emission of InGaN/GaN MQW, a very high excitation is necessary. Therefore, such a high optical pumping produces a strong screening effect, as expected. In this case, we would like to examine the influence of QCSE on the emission mechanism of InGaN/GaN MQW. Under optical excitation, the piezoelectric field in the strained InGaN well layer is weakened due to photo-generated carriers, which results in a blue shift of emission peak. Figure 15(a) shows the energy of the spontaneous emission peak as a function of optical pumping for all samples, from which the blue shift of the emission energy of all samples is found to saturate when the excitation power is higher than 60 uJ/cm2. This means that the QCSE can be completely screened out at the excitation powers of above 60 uJ/cm2. An estimation of two-dimensional screening length can also support this conclusion. From above data, the threshold is around 100 (oJ/cm2 or even higher.

328

T. Wang

i

•

i

•

Room Temperature

i—•—i

•

(a)

(b)

d=1.4nm

=SBBBese=o=e e.

415

d=1.9nm .410

405

V., A

d=2.4nm

A—A^S&AAAA—A—A A

KN

d=3.3nm

400 ~T> 395•

d=3.9nm

O—OQ3DOOOO—O—O O

• • ' • •, 50 100 150 200 250 300 Optical pumping (nJ/cm 2 )

350

-1 0

-

5

0 5 Growth dlrection(nm)

1C

Fig. 15. (a) Spontaneous emission energy as a function of optical excitation power. The curves from bottom to top are in the order of increasing well thickness. The blue shift of the emission energy of all samples saturates when the excitation power is higher than 60 (iJ/cm2. (b) Electron wave-functions in conduction band for all samples based on single particle model. (From Ref.10 with permission of reprinting from Elsevier).

Based on the estimation,11 the sheet carrier density is between 3xl012/cm2 and 8xl012/cm2, which is similar to a threshold current density of tens kA/cm2 for InGaN/GaN-based LD. If we follow Ridley44 and choose a two dimensional screening length X=2ekT/e2n (where n is the carrier density, e is a static permittivity, and other symbols have their usual meaning), giving X«LZ (where Lz is well thickness). This means that QCSE can be completely screened out. In other word, the QCSE has no any influence on the stimulated emission mechanism of InGaN/GaN MQWs. In this case, we can calculate the wavefunction of carrier based on a single particle model without considering the piezoelectric field using a band-offset ratio of AEC: AEv=0.54:0.46 (Ref.50). The material parameters used in this calculation can be found in the review paper given by Vurgaftman et al.26 In particular, Figure 15(b) shows the electron wavefunctions of all samples, indicating that the confinement of the electron wavefunction inside well can be enhanced by increasing well thickness. Furthermore, wide well thickness is also helpful for an enhancement of the optical confinement. Both can cause the threshold to

InGaN/GaN MQW MOCVD-Optical Investigation

329

drop, like AlGaAs/GaAs system. Therefore, the threshold decreases with increasing well thickness. '

l

'

l

'

A / \

l

•

l

•

l

1.4nm / \ * o ^

A

,u.)

A

1.9nm /

w .¥

'

1

Rxmterrps-aire lOOWcn? (a)

\

J //A\\\

c

jjpse!

2.3 nm / / \V

—y V-

(0

y\ V

E o Z •

360

•

3.3 nm / - ^

> - -

3.9 nmy

\

•

i^-i

•

•

•

PS—i

380 400 42) 440 460 480

Wavelength (nm)

1

2

3

4

Qjantum well thickness (nm)

Fig. 16. (a) Room temperature PL spectra of InGaN/GaN MQW samples under the excitation of a 100 W/cm2 used by a He-Cd laser, and (b) Dependence of PL intensity on well thickness. {From Ref.10 with permission of reprinting from Elsevier).

As a comparison, the PL measurements were also carried out on all the samples under a low excitation, in which a 10 mW He-Cd laser is generally used. The incident excitation power density is estimated to be around 100 W/cm2, two or three orders of magnitude lower than that of above femtosecond pulse laser. The room temperature PL spectra are shown in Figure 16(a). For samples A, B and C, there appears a strong emission peak accompanied by a weak peak on the low energy side. Both increase in wavelength as the quantum well thickness increases. In each case, the main peak is due to the emission of spontaneous emission from InGaN well. The weak peaks are due to the phonon replica, which is not our interest here. For samples D and E, except a weak peak due to the phonon replica mentioned above there appear two peaks. This is caused by the indium non-uniformity induced exciton localization effect, which

330

T. Wang

seems to become clearer in the sample with wide well thickness, studied by the previous study. 3 The PL intensity of main peaks is plotted as a function of well thickness, as shown in Figure 16(b). The PL intensity is found to increase with increasing quantum well thickness up to 2 nm, and then decreases with further increasing the quantum well thickness.51 The dependence of integrated intensity on well thickness shows a similar trend. All the data show good reproducibility from batch to batch. The dependence of RT PL intensity on well thickness is generally accepted to be caused by the piezoelectric field induced QCSE, which is enhanced by increasing well thickness. In the case of a thin well such as the 1.4—1.9 nm samples, the QCSE is weak. Consequently, like AlGaAs/GaAs system, the PL intensity increases as a result of increasing well thickness. In this case, without taking QCSE into account, the confinement of the wavefunction of carrier can be enhanced by increasing well thickness, similar to Figure 16(b). Furthermore, thermal energy can also produce the carrier's diffusion over barrier potential, leading the PL intensity to drop, which depends on the ratio of thermal energy to potential barrier. In the case of 1.4 nm well, the ratio is estimated to be 0.23, while this ratio is about 0.17 for 1.9 nm well. (In both cases, only one confined subband in conduction band is predicted). This also causes PL intensity of the sample of 1.4nm well lower than that of 1.9 nm well. In the case of wide well, the QCSE dominates the emission mechanism. Hence, the PL intensity is determined by the QCSE, resulting in the decrease of PL intensity with increasing well. Here, an important point should be taken into account, i.e., since the stimulated emission mechanism under a high optical pumping is different from the spontaneous emission mechanism under a low excitation, the optical quality of InGaN/GaN MQW for a laser design can not be adjudged by RT PL using a He-Cd laser, a low excitation source to study the optical properties of InGaN system, which is useful for LED design demonstrated in section III. From the present study, the sample with the highest PL intensity (sample B) has a high threshold for the stimulated emission, while the sample with lowest PL intensity (sample E) has the lowest threshold. The results indicated that InGaN/GaN MQW behaves in the same manner as the classical AlGaAs/GaAs system under a high excitation or

InGaN/GaN MQW MOCVD-Optical Investigation

331

high injection current that is necessary to produce stimulated emission. Therefore, like classical AlGaAs/GaAs system, dislocation may strongly affect the performance of InGaN-based LDs. Consequently, it is easy to understand that cw LDs with a long life-time can be achieved only on the low-dislocation-density substrate obtained by additional technologies. These above results strongly suggested that high crystal quality with a low dislocation density is crucial to achieve nitride laser, whiles this is not so important for nitride blue/green LEDs. 6. Femtosecond Studies of Electron Capture Times in InGaN/GaN MQWs Significant progress has been made in nitride-based optoelectronics during the last ten years, following the demonstration of high brightness blue/green LEDs and blue LDs containing InGaN/GaN quantum wells as the active region. Despite the rapid advances in device development, there are a number of important outstanding questions concerning the carrier dynamics that remain to be answered. One such issue is the quantum well capture time for carriers initially excited in the barriers. Theoretical studies of GaN/Al(Ga)N quantum wells have highlighted the strong dependence of the carrier capture times on key physical parameters, for example, the carrier density, the electron temperature and the quantum well parameters.52"54 However, previous investigations of carrier capture by ultrafast laser spectroscopy have been restricted so far to a limited range of experimental conditions.55'56 The samples studied were grown on (0001) oriented sapphire substrates by a low-pressure vertical MOCVD. Two structures consisting of nominally undoped 7.5 nm/2.5 nm InGaN/GaN QWs were grown. Both structures contained 10 QWs and were capped with a 20 nm thick GaN layer. The indium compositions were estimated to be 8% for sample I and 10% for sample II. Initial characterisation of the samples consisted of CW-PL spectroscopy and photoluminescence excitation (PLE) spectroscopy using a 0.25 m monochromator with a Xenon lamp as the excitation source. Degenerate and non-degenerate pump-probe measurements were performed using a regenerative amplified femtosecond Ti: sapphire laser

T. Wang

332

operating between 750 and 850 nm. The fundamental pulse width was less than 130 fs, and the repetition rate was 1 kHz. Frequency-doubled pulses were used to probe the QW absorption, with either frequencydoubled or frequency-tripled pulses as the pump pulse. Differential transmission (DT) signals were obtained by chopping the pump beam and recording the probe signal with a lock-in amplifier in phase with the pump. The probe beam was focussed to a tighter spot than the pump to ensure that only the centre of the excited region was probed. Figure 17 shows the cw-PL and PLE spectra obtained at 5 K for sample 1. The PLE signal shows a broadened QW absorption edge between 3.0 and 3.4 eV and a continuous absorption above 3.49 eV due to the GaN barriers. The fact that the PLE signal remains high for photon energies above the GaN absorption edge indicates that efficient carrier capture into the QWs is occurring. The efficient capture of carriers excited above the GaN barriers compared to Ref. 55 is related to the thin capping layers used in our samples. The same difference between PLE spectra for thin and thick capping layers has been observed by other authors.57

Sample 1 "55 c a>

UJ _l

a. •o

c ra

_i

a. 2.6

2.8

3.0 3.2 3.4 3.6 3.8 Photon Energy (eV)

4.0

4.2

Fig. 17. CW-PL and PLE spectra for sample 1 at 5 K. The arrow indicates the probe energy used in the pump-probe experiments. The inset shows a schematic band diagram neglecting the piezoelectric field and indicates the origin of the PLE signal when exciting carriers in the barriers. {From Ref. 12 with permission of reprinting from American Institute of Physics).

InGaN/GaN MQW MOCVD-Optical Investigation

333

We confirmed that the capture process is efficient in a separate experiment, in which we compared the amplified spontaneous emission (ASE) threshold for the two samples in stripe-pumping geometry with frequency-tripled pulses at 260 nm. At RT, the ASE peak appeared on the high-energy side of the PL peak: at 386 and 398 nm for samples 1 and 2, respectively. As indicated schematically in the inset of Fig. 17, the ASE threshold for sample 1 is around 110 mJ/cm2, which is comparable to that obtained by other authors.58 •

r~"^^^

I™"^^^"

|

"

"

^

^

^

^

^

^

r

Fig. 18. Normalized DT signals for sample 1 for either 390 nm or 260 nm pumping at 300 K for 140 tiJ/cm2 pump fluence. The probe wavelength was 390 nm in both cases. The inset shows a schematic of the carrier capture dynamics following 260 nm pumping. (From Ref.12 with permission of reprinting from American Institute of Physics).

Figure 18 shows typical DT results at room temperature for sample 1. The figure compares the DT signals obtained either with excitation of carriers in the barriers using frequency-tripled pump pulses at 260 nm (4.77 eV), as shown in the inset, or with excitation of carriers directly into the quantum wells, with frequency-doubled pump pulses at 390 nm (3.18 eV). In both cases, the probe wavelength was 390 nm (3.18 eV),

334

T. Wang

which corresponds to an energy below the GaN bandgap but within the QW absorption band, as indicated by the arrow in Fig. 17. The DT signal measures the saturation of the QW absorption caused by carriers within the quantum wells. For direct excitation of the quantum wells at 390 nm, a nearinstantaneous effect is expected, but for excitation at 260 nm, a delay will occur, reflecting carrier capture into the wells. The rise time of the DT signals shown in Fig. 18 is clearly much slower for the case of the 260 nm pump, and the difference gives direct information about the carrier capture times. The hole capture time in nitride quantum wells has been calculated to be three orders of magnitude faster than that of the electrons due to the larger density of states, lower effective barrier height and greater number of confined states to act as capture channels.53 >

—

i

Normalized DT

bulk

•

i

-,....

|

,

4

. InGaNvf InGaN/GaN MQW Pump: 260 nm Probe: 390 nm 1

0

1 2 Time delay (ps)

3

Fig. 19. Nodegenerate pump-probe measurements on bulk InGaN and InGaN/GaN NQWs with 260 nm pumping and 390 nm probing at 300K for 140 uJ/cm2 pump fluence. The solid curves are fit assuming rise times of 160 fs for the InGaN epilayer and 470 fs for MQWs (From Ref.12 with permission of reprinting from American Institute of Physics).

InGaN/GaN MQWMOCVD-Optical Investigation

335

Carrier Density (x 1019 cm'3) Normal zed DT

700 600 CD

500

E

300 K \ K 140(iJ/cm2 1 0

400

1 2

3

4

Time delay (ps)

+-»

Q. (0

o

300 200 I

100 200 300 400 500 Excitation Energy Density (jiJ/cm2)

Fig. 20. Capture time for sample 1 at 300 K with 260 nm pumping versus excitation energy density. The top axis gives the initial density in the barriers averaged over the MQW region. The inset compares the DT signals at 5 K and 300 K for an excitation energy density of 140 |xl/cm2. (From Ref.12 with permission of reprinting from American Institute of Physics).

The DT rise time is therefore expected to be determined primarily by the electron dynamics. The measured rise time is a convolution of the laser pulse width, the electron relaxation time to the GaN band edge, and the electron capture time Tenure. We were able to determine the convolution of the first two parameters separately by performing a pumpprobe experiment on an Ino.11Gao.g9N epilayer using pump pulses at 260 nm and probe pulses at 390 nm. Figure 19 compares the DT signal for the bulk epilayer with that obtained for the MQW under identical conditions. The DT rise time for the epilayer was 160 fs, which has to be compared to the value of 470 fs obtained for the MQW. On the assumption that the electron cooling times of the InGaN epilayer and bulk GaN are similar, we were then able to determine ^capture

by deconvolving the pulse width/carrier cooling time through a rate equation analysis of the DT rise time for the MQW samples. Figure 20 shows the excitation energy dependence of the electron capture time for sample 1 at 300 K. The corresponding carrier density

336

T. Wang

^barrier is plotted on the top axis. The carrier densities represent the initial density in the GaN barriers averaged across the MQW region and were calculated by assuming an absorption coefficient of 1.8 x 105 cm-1 at the pump wavelength.58 Carrier densities ~1019 cm"3 occur frequently in laser diodes, and hence the present results are directly relevant to the carrier dynamics in the stimulated emission regime. The results in Fig. 19 indicate that the capture time decreases with increasing carrier density, in agreement with theoretical calculations for GaN/AlGaN quantum wells.53 In the model of ref 53, the decrease in ^capture occurs through the increased overlap between the confined and continuum states caused by band bending in the barrier. This band bending arises from the variation of the local charge density caused by the different electron and hole capture times. Another factor that may also be significant, and has yet to be considered theoretically, is increased carrier-carrier scattering at higher densities. The inset of Fig. 19 compares the results obtained at 140 uJ/cm2 (Awarder ~ 1 .OxlO19 cm"3) for temperatures of 5 K and 300 K. The capture time at 5 K is 780 ± 40 fs, significantly slower than that the room temperature value of 470 ± 40 fs. This behaviour agrees with calculations performed for GaN/AIN quantum wells which predict that the capture velocity (and hence the capture rate) generally increases with temperature due to the thermal broadening of the electron distribution.54 An alternative explanation is that the electron relaxation is slower at 5 K. However, we consider this second explanation to be unlikely because the electron-phonon scattering rate in GaN has been determined to be 4xl0 13 s' at 25 K,59 which implies that the carrier relaxation time is much faster than the time resolution of our experiment. Our results concur with CW electroluminescence measurements which show that carrier capture is less efficient at low temperatures in blue nitride QW LEDs,61'62 although it is important to realize that the CW studies are also affected by the temperature dependence of carrier transport processes. The results obtained for sample 2 were broadly similar to those for sample 1, but with some important differences. The peak emission wavelength at 5 K was 418 nm as opposed to 413 nm for sample 1, which confirmed the slightly larger indium mole fraction determined by the XRD and TEM measurements. The capture times were somewhat

InGaN/GaN MQW MOCVD-Optical Investigation

337

faster than those for sample 1. For example, capture times of 680+30 fs and 440+30 fs are determined at 140 mJ/cm2 for temperatures of 5 and 300 K, respectively. The faster capture times are consistent with cw-PL measurements, which indicate that the capture is more efficient in deeper wells.63 Moreover, theoretical work predicts that the capture time should be an oscillatory function of the QW width, with resonances whenever a new bound state is formed at the top of the QW.52"54 The same argument would imply a similarly strong dependence on the indium mole fraction at constant well width. The faster capture times and weaker temperature dependence measured for sample 2 could then be explained by assuming that sample 2 is closer to one of the electron resonances than is sample 1. The capture times measured in both samples are comparable to, but slightly longer than, the capture times reported for Si-doped Ino.05Gao.95N/Ino.15Gao.85N and Ino.07Gao.93 N/In0.i2Gao.8gN MQWs at RT. 5556 One possible reason for this slight discrepancy is the band bending caused by the Si doping. Alternatively, the difference may simply be related to the strong sensitivity of the capture time to the precise QW parameters that we have noted earlier.

7. Summary The quantum-well number dependence of strain-relaxation in InGaN/GaN MQW has been investigated by XRD and photoluminescence measurements. With increasing quantum-well number, the emission energy shows a clear shift, which is attributed to the strain-relaxation. Based on the QCSE model, the residual strain is estimated for our X-ray kinetic simulation, agreeing well with the measured XRD patterns. The quantum-confined-Stark effect and the exciton localization effect in PnGaN/GaN-based LEDs were systematically studied. Based on the measurement of LED output power, the exciton localization effect is confirmed to be helpful for improving the output power of LED. However, QCSE shows much stronger influence on the output power of LEDs than the exciton localization effect, which should be taken into account in designing blue/green LEDs. The emission mechanism of InGaN/GaN MQWs with different well thickness including the internal quantum efficiency has been investigated

338

T. Wang

under high excitation. For the MQWs with wide-well, strong emission from quantum-dot-like states has been observed, confirmed by the temperature- and power- dependent PL measurements. The studies indicates that the emission mechanism under high excitation is dominated by quantum-dot-like states from room temperature to low temperature, which can partly explain the enhanced internal quantum efficiency under high excitation compared to low excitation. This result is of critical importance for understanding the emission mechanism during lasing process for InGaN/GaN MQWs based laser diodes. The investigation of the stimulated emission under optical pumping has been carried out on InGaN/GaN MQW structures at room temperature. The threshold of optical pumping decreases monotonically as a result of increasing well thickness. The stimulated emission is generated under high optical pumping that results in the completely screening of the piezoelectric field. In this case, the stimulated emission mechanism is not dominated by the QCSE, while the QCSE is generally accepted to play an important role in the process of spontaneous emission in InGaN/GaN-based LEDs. Consequently, InGaN/GaN MQW behaves in the same manner as the classical AlGaAs/GaAs system, and the threshold decreases with increasing well thickness. This different mechanism between spontaneous and stimulated emissions must be highly taken into account in designing InGaN/GaN LD and LED. The carrier capture times for InGaN/GaN quantum wells were by non-degenerate femtosecond pump-probe measurements. The capture times were deduced from the rise time of the QW differential transmission signal following photoexcitation of carriers in the barriers by femtosecond UV pulses. It is found that the capture time varies significantly with carrier density, temperature and indium mole fraction. These results suggest that the capture time is a complicated function of the excitation conditions and sample design, which clearly has implications for the design of LEDs and low threshold, high efficiency lasers.

InGaN/GaN MQW MOCVD-Optical Investigation

339

Acknowledgments Experimental results reported here are parts of work in Satellite Venture Business Laboratory, the University of Tokushima in Japan, and in Department of Electronic and Electrical Engineering and Department of Physics, the University of Sheffield in the UK. The author is grateful to Prof. S Sakai, Drs J Bai, P J Parbook, M Fox and W H Fan for their contributions to this work. References 1. S Nakamura, M Senoh, S Nagahama, N Iwasa, T Yamada, T Matsushita, H Kiyoku, Y Sugimoto, T Kozaki, H Umemoto, M Sano and K Chocho, "InGaN/GaN/AlGaNbased laser diodes with modulation-doped strained-layer superlattices", Jpn. J. Appl. Phys. 36, LI 568 (1997). 2. T Wang, D Nakagawa, J Wang, T Sugahara and S Sakai, "Photoluminescence investigation of InGaN/GaN single quantum well and multiple quantum wells", Appl.Phys.Lett. 73,3571 (1998). 3. T Wang, D Nakagawa, M Lachab, T Sugahara and S Sakai, "Optical investigation of InGaN/GaN multiple quantum wells", Appl.Phys.Lett. 74, 3128 (1999). 4. T Wang, H Saeki, J Bai, M Lachab, T Shirahama and S Sakai, "Effect of silicondoping on the optical and transport properties of InGaN/GaN multiple quantum well structures", Appl.Phys.Lett 76, 1737 (2000). 5. T Wang, J Bai and S Sakai, "Influence of InGaN/GaN quantum-well structure on the performance of light-emitting diodes and laser diodes grown on sapphire substrate", J. Cryst. Growth 224, 5 (2001). 6. J Bai, T Wang and S Sakai, "The influence of well-thickness on the mechanism of radiative recombination in InGaN/GaN quantum well structure", J. Appl. Phys. 88, 4729 (2000). 7. T Wang, J Bai, S Sakai and J K Ho, "Investigation of the emission mechanism in InGaN/GaN-based light-emitting diodes", Appl.Phys.Lett. 78, 2671 (2001). 8. J Bai, T Wang and S Sakai, "Investigation of the strain-relaxation in InGaN/GaN multiple-quantum-well structures", J.Appl.Phys. 90, 1740(2001). 9. T Wang, P J Parbrook, W H Fan and A M Fox "Optical Investigation of InGaN/GaN multiple-quantum wells under high excitation", Appl.Phys.Lett. 84, 5159(2004). 10. T Wang, P J Parbrook, M A Whitehead, W H Fan, S M Olaizola and A M Fox, "Study of stimulated emission from InGaN/GaN quantum well structure", J. Cryst. Growth 273/1-2, 48-53 (2004).

340

T. Wang

11. W H Fan, S M Olaizola, T Wang, P J Parbrook, M A Whitehead and A M Fox, "Temperature dependence of carrier capture time in InGaN/GaN multiple quantum wells", Phys. Stat. Sol. (b), 240, 364 (2003). 12. W H Fan, S M Olaizola, J P R. Wells, A M Fox, T Wang, P J Parbrook, D J Mowbray, and M S Skolnick, "Femtosecond studies of electron capture times in InGaN/GaN multiple quantum wells", Appl.Phys.Lett. 84, 3052(2004). 13. T Wang, J Bai and S Sakai, International Workshop on Nitride semiconductors (IWN2000) 09/24-27/2000, Nagoya, Japan. 14. See website: http://www.uknc.org. 15. T Takeuchi, S Sota, M Katsuragawa, M Komori, H Takeuchi, H Amano and I Akasaki, "Quantum-confined stark effect due to piezoelectric fields in GalnN strained quantum wells", Jpn. J. Appl. Phys. 36, L382 (1997). 16. M Koike, S Yamasaki, S Nagai, N Koide, S Asami, H Amano and I Akasaki, "Highquality GalnN/GaN multiple quantum wells ", Appl.Phys.Lett. 68, 1403 (1996). 17. T Takeuchi, S Sota, M Katsuragawa, M Komori, H Takeuchi, H Amano and I Akasaki, "Optical properties of strained AlGaN and GalnN on GaN", Jpn. J. Appl. Phys. 36, L177 (1997). 18. N A Shapiro, P Perlin, C Kisielowski, L S Mattos, J W Yang and E R Weber, "The effects of indium concentration and well-thickness on the mechanisms of radiative recombination in InxGa].xN quantum wells", MRS Internet J. Nitride Semicond. Res. 5 1 (2000) 19. P Perlin, C Kisielowski, V Iota, B A Weinstein, L Mattos, N A Shapiro, J Kruger, E R Weber, Jinwei Yang, "InGaN/GaN quantum wells studied by high pressure, variable temperature, and excitation power spectroscopy", Appl.Phys.Lett. 73, 2778 (1998). 20. T Takeuchi, C Wetzel, SYamaguchi, H Sakai, H Amano and I Akasaki, "Determination of piezoelectric fields in strained GalnN quantum wells using the quantum-confined Stark effect", Appl. Phys. Lett. 73, 1691 (1998). 21. M D Nardelli, K Rapcewicz and J Bernholc, "Polarization field effects on the electron-hole recombination dynamics in In0.2Ga0.8N/lni_xGaxN multiple quantum wells", Appl.Phys.Lett. 71, 3135 (1997). 22. A Hangleiter, "Optical properties and polarization fields in the nitrides" J.of Lumin. 87-89, 130 (2000). 23. J K Sheu, G C Chi, Y K Su, C C Liu, C M Chang, W C Hung and M J Jou, "Luminescence of an InGaN/GaN multiple quantum well light-emitting diode", Solid-State Electronics, 44,1055 (2000). 24. O Mayrock, H J Wiinsche, and F Henneberger, "Polarization charge screening and indium surface segregation in (In,Ga)N/GaN single and multiple quantum wells", Phys.Rev. B 62, 16870 (1990). 25. S F Chichibu, A C Abare, M S Minsky, S Keller, S B Fleischer, J E Bowers, E Hu, U K Mishra, L A Coldren, S P DenBaars and T Sota, " Effective band gap

InGaN/GaN MQW MOCVD-Optical Investigation

26.

27. 28.

29. 30.

31. 32.

33.

34.

35.

36.

37.

38.

39.

341

inhomogeneity and piezoelectric field in InGaN/GaN multiquantum well structures", Appl. Phys. Lett. 73, 2006 (1998). R Singh, D Doppalapudi and T D Moustakas, "Phase separation in InGaN thick films and formation of InGaN/GaN double heterostructures in the entire alloy composition", Appl. Phys. Lett. 70,1089 (1997). S Chichibu, K Wada and S Nakamura, "Spatially resolved cathodoluminescence spectra of InGaN quantum wells", Appl. Phys. Lett. 71, 2346 (1997). Y Narukawa, Y Kawakami, M Funato, S Fujita and S Nakamura, "Recombination dynamics of localized excitons in In02oGag,soN-In0osGa09s multiple quantum wells", Phys. Rev. B 55, R1938 (1997). P G Eliseev, P Perlin, J Lee and M Osinski, "Blue temperature-induced shift and band-tail emission in InGaN-based light sources", Appl. Phys. Lett. 71, 569 (1997). L Nistor, H.Bender, A Vantomme, M F Wu, J Landuyt, K P O'Donnell, RW Martin, K Jacons and I Moerman, "Direct evidence of spontaneous quantum dot formation in a thick InGaN epilayer", Appl.Phys.Lett. 77, 507 (2000). K P O'Donnell, R W Martin and P G Middleton, "Origin of luminescence from InGaN diodes", Phys.Rev.Lett. 82, 237 (1999). Y Narukawa, K Sawada, Y Kawakami, Shizuo Fujita, Shigeo Fujita, and S Nakamura, "Emission mechanism of localized excitons in InxGaj.xN single quantum weir, J. Cryst. Growth 189-190, 606 (1998). Y Kawakami, Y Narukawa, K Sawada, S Saijyo, Shizuo Fujita, Shigeo Fujita and S Nakamura, "Recombination dynamics of localized excitons in self-formed InGaN quantum dots", Materials Science and Engineering B 50, 256 (1997). E L Piner, M K Behbehani, S X Liu, N A El-Masr and S M Bedair, "Phase separation and ordering coexisting in Infiaj^N grown by metal organic chemical vapor deposition", Appl.Phys.Lett. 75, 2202 (1999). A Satake, Y Masumoto, T Miyajima, T Asatsuma, F Nakamura and M Ikeda, "Localized exciton and its stimulated emission in InGaN multiple quantum wells", J. Cryst. Growth 190, 601 (1998). P A Crowell, D K Young, S Keller, E L Hu and D D Awschalom, "Near-field scanning optical spectroscopy of an InGaN quantum well", Appl.Phys.Lett. 72, 927 (1998). K L Teo, J S Colton, P Y Yu, E R Weber, M F Li, W Liu, K Uchida, H Tokunaga, N Akutsu and K Matsumoto, "An analysis of temperature dependent photoluminescence line shapes in InGaN\ Appl.Phys.Lett. 73, 1697 (1998). E S Jeon, V Kozlov, Y K Song, A Vertikov, M Kuball, A V Nurmikko, H Liu, C Chen, R S Kern, C P Kuo, and M G Craford, "Recombination dynamics in InGaN quantum wells", Appl.Phys.Lett. 69,4194 (1996). S Nakamura, in Introduction to Nitride Semiconductor Blue Lasers and Light Emitting Diodes, edited by S Nakamura and S G Chichibu (Taylor and Francis, London, 2000), pp. 325.

342

T. Wang

40. J Bai, W H Liu, Z Q Wu, Y T Wang, L Xiu, and X M Jiang, "X-ray diffraction study of GaAs/InAs/GaAs ultrathin single quantum well", J. Appl. Phys. 79, 7627 (1996). 41. D A B Miller, D S Chemla, T C Damen, A C Gossard, W Wiegmann, T H Wood and C A Burrus, "Electric-field depedence of optical absorption near the band-gap of quantum-well structures", Phys.Rev. B32, 1043 (1985). 42. D A B Miller, D S Chemla, T C Damen, A C Gossard, W Wiegmann, T H Wood and C A Burrus, "Band-edge electroabsorption in quantum welll structures-the quantum confined Stark-effect', Phys. Rev. Lett. 26, 2173 (1984). 43. B E Warren, X-ray diffraction (Adison-Wesley, MA, 1969), P.28. 44. B K Ridley, "Kinetics of radiative recombination in quantum-wells", Phys. Rev. B 41,12190(1990). 45. I Vurgaftman and J R Meyer, "Band parameters for nitrogen-containing semiconductors", J. Appl. Phys. 94, 3675 (2003). 46. J Ristic, E Calleja, M A Sanchez-Garcia, J M Ulloa, J Sanchez-Paramo, J M Calleja, U Jahn, A Trampert and K H Ploog, "Characterization of GaN quantum discs embedded in AlxGal-xN nanocolumns grown by molecular beam epitaxy", Phys. Rev. B 68,125305 (2003). 47. Y Arakawa, "Progress and prospect of quantum dot lasers", Proceeding of SPIE Vol.4580, 179(2001). 48. J Rice, Private communication. 49. H Schomig, S Halm, A Forchel, G Bacher, J Off, and F Scholz, "Probing Individual Localization Centers in an InGaN/GaN Quantum Well", Phy. Rev. Lett. 92,106802 (2004). 50. O Ambacher, "Growth and applications of Group III nitrides", J.Phys.D: Appl.Phys. 31, 2653 (1998). 51. In fact, the PL measurements of some samples with 1 nm well thickness also indicated that their PL intensities are lower than that of sample A, the sample with 1.4 nm well. Therefore, we can say the PL intensity increases with increasing well thickness up to 2 nm. 52. N S Mansour, K W Kim and M A Littlejohn, " Theoretical study of electrontransport in Gallium Nitride", J. Appl. Phys. 77, 2834 (1995). 53. J Wang, K W Kim and M A Littlejohn, "Carrier capture in pseudomorphically strained wurtzite GaN quantum-well lasers", Appl. Phys. Lett. 71, 820 (1997). 54. N A Zakhleniuk, C R Bennett, V N Stavrou, M Babiker and B K Ridley, "Quantum capture of injected electrons in GaN-based laser heterostructures", Phys. Stat. Sol. A 176, 79 (1999). 55. U Ozgur, M J Bergmann, H C Casey Jr., H O Everitt, A C Abare, S Keller and S P DenBaars, "Ultrafast optical characterization of carrier capture times in InxGaj_xN multiple quantum wells", Appl. Phys. Lett. 77, 109 (2000). 56. U Ozgiir, H O Everitt, S Keller and S P DenBaars, "Stimulated emission and ultrafast carrier relaxation in InGaN multiple quantum wells", Appl. Phys. Lett. 82, 1416 (2003).

InGaN/GaN MQW MOCVD-Optical Investigation

343

57. For example, S F Chichibu, Y Kawakami, and T Sota in Introduction to Nitride Semiconductor Blue Lasers and Light Emitting Diodes, edited by S Nakamura and S G Chichibu (Taylor and Francis, London, 2000), pp. 153-270. 58. J F Muth, J H Lee, I K Shmagin, R M Kolbas, H C Casey Jr., B P Keller, U K Mishra, and S P DenBaars, "Absorption coefficient, energy gap, exciton binding energy, and recombination lifetime of GaN obtained from transmission measurements", Appl. Phys. Lett. 71, 2572 (1997). 59. KT Tsen, D K Ferry, A Botchkarev, B Sverdlov, A Salvador and H Morkoc, "Direct measurements of electron-longitudinal optical phonon scattering rates in wurtzite GaN", Appl. Phys. Lett. 71, 1852 (1997). 60. A Hori, D Yasunaga, A Satake and K Fujiwara, "Temperature dependence of electroluminescence intensity of green and blue InGaN single-quantum-well lightemitting diodes", Appl. Phys. Lett. 79, 3723 (2001). 61. A Hori, D Yasunaga, A Satake and K Fujiwara, "Temperature and injection current dependence of electroluminescence intensity in green and blue InGaN singlequantum-well light-emitting diodes", J. Appl. Phys. 93, 3152 (2003). 62. X A Cao, S F Leboeuf, L B Rowland, C H Yan and H Liu, "Temperature-dependent emission intensity and energy shift in InGaN/GaN multiple-quantum-well lightemitting diodes", Appl. Phys. Lett. 82, 3614 (2003). 63. F Binet, J Y Duboz, C Grattepain, F Scholz, and J Off, "Carrier capture in InGaN quantum wells and hot carrier effects in GaN", Mater. Sci. Eng. B 59, 323 (1999).

This page is intentionally left blank

CHAPTER 10 CLUSTERING NANOSTRUCTURES AND OPTICAL CHARACTERISTICS IN INGAN/GAN QUANTUM-WELL STRUCTURES WITH SILICON DOPING

Yung-Chen Cheng, Cheng-Yen Chen, and C. C. Yang Graduate Institute of Electro-Optical Engineering and Department of Electrical Engineering, National Taiwan University, 1, Roosevelt Road, Sec. 4, Taipei, Taiwan, R.O.C. E-mail: [email protected] The results of photoluminescence (PL), detection-energy-dependent photoluminescence excitation, excitation-energy-dependent photoluminescence, amplified spontaneous emission (ASE), cathodoluminescence (CL) and strain state analysis (SSA) of three InGaN/GaN quantum-well (QW) samples with un-doped, well-doped, and barrier-doped conditions are compared for understanding the silicon doping effects on nanostructure and photon emission mechanism. Based on the SSA and CL images, a nanostructure model is built for describing the potential fluctuation differences between the three samples. In the barrier-doped sample, strongly clustering nanostructures with individual steep potential minima, which generate significant quantum confinement effects, are assumed. In the undoped and well-doped samples, relatively weaker composition fluctuations, in which carriers relax through a cascading process, are proposed. Between the undoped and well-doped samples, the potential fluctuation in the well-doped sample is relatively steeper such that a certain extent of quantum confinement existed. Such variations in nanostructure result in different carrier transport processes between the coexistent quantum dot and quantum well states, which well explain the PL, DEDPLE, EEDPL, ASE, and CL observations. In particular, the PL results provide us clues for speculating that the S-shape behavior of PL peak position is dominated by the quantum-confined Stark effect (QCSE) in an undoped InGaN/GaN QW structure. However, carrier

345

Cheng, Chen, Yang

346

localization is more effective in blue-shifting luminescence and improving radiative efficiency of a sample, when compared with the relaxation of QCSE. Also, the ASE results show the temperaturedependent evolution of gain spectrum due to the liquidation of thermalized carriers. Different nanostructures result in different spectral variation trends. 1. Introduction Silicon doping in InGaN/GaN quantum well (QW) structures has been widely used in fabricating high-performance light-emitting devices. By appropriately doping silicon in certain layers, photon emission efficiency can be significantly enhanced. The effects of silicon doping have been widely studied. Several models were proposed for interpreting the optical and material characteristics of InGaN/GaN QWs with silicon doping.1"10 The most commonly used model is the carrier screening effect of the strain-induced piezoelectric field and hence the reduction of quantumconfined Stark effect (QCSE). This effect leads to the reduced Stokes shift,1,2'5 the decrease of radiative recombination lifetime,2'3'510 the blue shift of PL spectrum,4'5 and the enhancement of photon emission efficiency.1'5'6 Also, material analyses have led to the conclusions of growth mode change,6 nanostructure alternation,7 formation of abrupt QW interfaces,2'8 strain relaxation,7'9 and higher potential uniformity in QWs.5 These conclusions are not necessarily mutually consistent, particularly in the material structures of such silicon-doped QW samples. Although these models can interpret certain observed phenomena in a silicon-doped sample, they are essentially based on phenomenological viewpoints. The fundamental issues, such as the variations of nanostructure upon silicon doping of different conditions and their implications in radiative mechanisms of such samples have not been well studied yet. The study on their nanostructures is important because it is usually difficult to grow uniform InGaN alloy, particularly with high indium contents, due to the solid phase immiscibility and phase separation between GaN and InN,11"13 which originate from their large lattice constant mismatch (-11 %). In such a compound, the clustering nanostructures and their derivative effects dominate the photon emission

InGaN/GaN Quantum Wells - Clustering Nanostructures, Silicon Doping

347

process.14"16 Understanding the nanostructures can help us in designing the optimum growth and process conditions for high-efficiency lightemitting devices. In this chapter, we demonstrate the comparisons of nanostructure and associated optical characteristics between three green-emitting InGaN/GaN QW samples of the same geometry and composition, but different silicon doping conditions. The three samples of no doping, well doping, and barrier doping were prepared for comparison. Several optical characterization methods, including temperature-dependent photoluminescence (PL), excitation-intensity-dependent photoluminescence, excitation-energy-dependent photoluminescence (EEDPL), detection-energy-dependent photoluminescence excitation (DEDPLE), and amplified spontaneous emission (ASE), were utilized to study the differences of photon emission mechanism between the three samples. To understand their nanostructures more directly, the method of strain state analysis (SSA), based on a high-resolution transmission electron microscopy (HRTEM) technique17, and cathodoluminescence (CL) measurement were performed. The results of SSA and CL show the strongest indium clustering structure, which leads to significant quantum confinement effects, in the barrier-doped sample. On the other hand, relatively milder potential fluctuations exist in the undoped sample, in which shallow indium-rich clusters contribute to photon emission when the sample is weakly excited. In the well-doped sample, the degree of potential fluctuation is essentially between the other two samples. Other optical characterization results, including the blue-shifted trend of PL spectral peak and the improvement of radiative efficiency in the barrierdoped sample, can be well interpreted with the nanostructure model based on the SSA and CL observations. Meanwhile, the ASE spectra show a distinct temperature dependence of optical gain characteristics in the barrier-doped sample. The liquidation of thermalized carriers homogenizes the gain spectrum as temperature increases, leading to the evolution of ASE spectrum from a multiple-peak feature into a single major peak distribution. This chapter is organized as follows: In section II, the sample structures and experimental conditions are described. In section III, the measurement results of PL, EEDPL, and DEDPLE are reported. Then, in

348

Cheng, Chen, Yang

sections IV-VI, the observations of ASE, CL, and SSA, respectively, are discussed. A nanostructure model and the interpretations for the optical characteristics are proposed in section VII. Finally, conclusions are drawn in section VIII. 2. Sample Preparation and Measurement Conditions The three InGaN/GaN QW samples of different doping conditions were prepared with MOCVD growth on sapphire (0001) substrate. They all consisted of five QW periods, with 2.5 nm in well width and 7.5 nm in barrier width, on top of an un-doped GaN buffer layer of 1.52 yuan, in thickness. The growth temperatures were 1100 and 800 °C for the GaN barriers and InGaN wells, respectively. The doping concentration of silicon was 5 x 1018 cm" , either in barriers or wells. The samples of undoped, well-doped and barrier-doped conditions were denoted with HU, HW, and HB, respectively. The nominal indium contents of the three samples are the same at 20 %. Continuous-wave (CW) PL measurements were carried out with the 325 nm line of a 35 mW He-Cd laser. The samples were placed in a cryostat for temperature-dependent measurements ranging from 10 to 300 K. PLE experiments were conducted using a quasi-monochromatic excitation light source from a xenon lamp dispersed by a 0.15-m monochromator. A 0.5-m monochromator was also used for detecting the luminescence intensity at selected wavelengths. In the measurements of EEDPL, a xenon lamp, dispersed by a 0.15-m monochromator, was used for varying the excitation wavelength. Regarding the ASE measurements, the fourth-harmonic (266 nm) of a Q-switched Nd:YAG laser at 100 Hz was focused with a cylindrical lens to form a line-shaped excitation beam of 2.5 mm in length. The excitation pulse width was 3 ns and its peak intensity was 10 MW/cm2. The HRTEM investigations (for SSA) were conducted with a 300 keV JEM 3010 microscope. All the high-resolution micrographs were taken with two-electron-beam interference. The CL images and spectra were acquired using a Gatan monoCL3 spectrometer in a JEOL JSM 6700F SEM system. The acceleration voltage of electron beam was 2 kV

InGaN/GaN Quantum Wells - Clustering Nanostructures, Silicon Doping

349

3. Photoluminescence and Photoluminescence Excitation In Figure 1, we show the temperature-dependent variations of PL spectral peak energy of the three samples. Here, a significant blue shift of spectral peak energy in sample HB, when compared with the other two samples, can be observed. A clear S-shape variation can be seen only in sample HU. The PL spectral peak energies of samples HU and HW are quite close below 150 K. They split at higher temperatures because of the significant blue-shift trend in sample HU. It is interesting to note that the curves of samples HW and HB are almost parallel in Figure 1. In Figure 2, we show the temperature-dependent variations of normalized integrated PL intensity of the three samples.

*

2.45

I 2.30

50

100

150

200

250

300

Tem p e r a t u r e (K)

Fig. 1. Temperature-dependent PL spectral peak energies of the three samples.

50

100

150

200

250

300

T e m p e r a t u r e (K)

Fig. 2. Temperature-dependent integrated PL intensities of the three samples.

Cheng, Chen, Yang

350

Here, one can see that the radiative efficiency of sample HB is higher than those of the other two samples. Such results are quite similar to what we have reported previously with similar samples emitting violet photons.7 Barrier-doped samples always result in higher photon emission efficiencies. In the inset of Figure 2, we show the curve fitting of the Arrhenius plots. Here, the continuous curves represent the best fitting results with the equation18 I(T)

= a l + yffexp

E AX

-{/iT) T

e xPp{ l ++ rrcx

E A2 " /kT

(1)

Here, I(T) is the temperature-dependent normalized integrated PL intensity and the two activation energies EA1 and EA2 correspond to two different thermal quenching processes. The notations /? and /represent the probability ratios of non-radiative over radiative transitions in the two processes, respectively. The constant or is a fitting parameter. In the low temperature range, the shallower decay trend of the integrated PL intensity is dominated by the first exponential term with a smaller activation energy EA1, ranging from 4.2 to 9.1 meV, and f5 value around unity for all the three samples, as shown in Table 1. This portion of decay is supposed to originate from carrier consumption by the defects during carrier transport within a cluster or between coupled clusters.1 The smallest /? and the largest EA1 in sample HB imply the strongest carrier confinement in this sample. The second exponential term in (1) describes mainly the intensity decay in the higher-temperature range. The values of y (the corresponding activation energies, EA2) are 130.2 (44.6 meV), 250.8 (56 meV), and 66.6 (63.7 meV) for samples HU, HW, and HB, respectively. The results show again the strongest carrier localization effect in sample HB. Table 1. Results of the best curve fitting for the parameters in (1) of the three samples. sample

HU

HW

HB

a

0.989

0.994

0.989

P

1.25

1.08

0.78

7 (meV)

130.2

250.8

66.6

5.1

4.2

9.1

EA2 (meV)

44.6

56

63.7

EM

InGaN/GaN Quantum Wells - Clustering Nanostructures, Silicon Doping

351

2.415

Pump Intensity (W/cm )

Fig. 3. Excitation-intensity-dependent PL spectral peak energies of the three samples at 10 K.

Figure 3 shows the excitation-intensity dependent PL peak energies of the three samples at 10 K. Here, one can observe the increasing trends with excitation intensity in samples HU and HW, although their variation ranges are quite different. It is interesting to see the decreasing trend in sample HB. The different trends can be attributed to the different nanostructures, piezoelectric fields, and carrier densities in these three samples. In Figure 4, we show the EEDPL spectral peak positions as functions of excitation photon energy for the three samples at 10 K. In sample HU, the spectral peak energy of EEDPL is almost unchanged when carriers are excited in the quantum well layers. However, the peak energy blue shifts by around 10 meV if the carriers are excited in the barrier level. Such results imply that potential fluctuation exists in this sample and the carrier transport path determines the final states for carrier recombination. In sample HW, EEDPL peak energy decreases with increasing excitation energy. However, the opposite trend is observed in sample HB. Note that the EEDPL peak energies in samples HW and HB with carriers excited in the barrier levels are almost unchanged when the excitation energy is

352

Cheng, Chen, Yang

varied. These observations suggest that the structures of potential fluctuation among the three samples are quite different. W avelength 500

450

2.5

(nm )

400

3.0

Excitation

350

3.5

4.0

Energy

(eV)

Fig. 4. EEDPL spectral peak energies as functions of excitation photon energy of the three samples at 10 K. Wavelength 660 550

440

(nm) 330 --2.380 eV 2 .465 e V PLE

P LE 2.0

2.5

3.0

Excitation

3.5

4.0

Energy

4.5

(eV)

Fig. 5. PL and DEDPLE spectra of the three samples at 10 K. The energy values in eV shown in the legends represent the detection photon energies.

InGaN/GaN Quantum Wells - Clustering Nanostructures, Silicon Doping

353

In Figure 5, we show the PL and DEDPLE spectra of the three samples at 10 K. the DEDPLE spectra were normalized to the level of the GaN absorption peak at 3.5 eV. The energy values in eV shown in the legends represent the detection photon energies. Here, one can see that in sample HU, the luminescence intensity level with the InGaN absorption band decreases with increasing detection photon energy. However, the opposite variation trend is observed in sample HW. Meanwhile, almost the same absorption spectra, when the detection photon energy is varied, were measured in sample HB. Such major differences imply again the significant variations in sample nanostructure upon silicon doping of different conditions. The other important observation in Figure 5(c) is the maximum InGaN absorption in all PLE measurements of sample HB. The stronger InGaN absorption than the GaN barrier implies that no clear interfaces exist between wells and barriers in this sample. Note that the oscillating features in Figures 5(a)-(c) originate from the Fabry-Perot resonance in a cavity consisting of the multiple quantum well layers and the GaN buffer layer. 3.1. Amplified Spontaneous Emission In Figures 6(a)-(c), we show the temperature-dependent ASE spectra of the three samples. In these figures, the thick solid curves at the lower-left and upper-left corners represent the PL spectra at room temperature and 10 K, respectively. For temperatures above 220 K, the ASE spectra show clear Fabry-Perot resonance modes in samples HU and HW with the period corresponding to the thickness of the epi-layers of the samples. The sharp peaks at 2.33 eV in the ASE spectra correspond to the 532 nm line or the second-order diffraction of the 266 nm line from the pumping laser. Below 220 K, the two emission peaks in samples HU and HW can be identified as the PL peak (the lower-energy one, which corresponds to the localized or quantum dot - QD states) and the stimulated emission peak (the higher-energy one, which corresponds to the "free-carrier" or QW states). The PL peaks at 10 K and room temperature in the ASE spectra of samples HU and HW are quite consistent with those in Figure 1, as indicated with the thick curves in Figures 6(a) and (b).

Cheng, Chen, Yang

354

Wavelength (nm) 650600 550 500

450

400

350

10K 20K

(A C

a> •o

a> N

75 E >_ o z

Photon Energy (eV) Fig. 6. (a) Temperature-dependent ASE spectra of samples HU.

Wavelength (nm) 600

540

480

420

360

(0

c

I i

2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8

Photon Energy (eV) Fig. 6. (b) Temperature-dependent ASE spectra of samples HW .

4.0

InGaN/GaN Quantum Wells - Clustering Nanostructures, Silicon Doping

355

Wavelength (nm) 550

500

450

400

350

Photon Energy (eV) Fig. 6. (c) Temperature-dependent ASE spectra of samples HB.

It is noted that the PL peak position of either sample HU or HW in the ASE measurement is almost independent of temperature; however, the stimulated emission peak shows the similar variation with temperature to the result of the CW PL measurement.20 In other words, the stimulated emission peaks of samples HU and HW show the S-shape temperature variation and the red-shift trend, respectively. Such an Sshape behavior in sample HU has been attributed to either the carrier localization effect due to the cluster structures21 or the quantum-confined Stark effect due to the strain-induced piezoelectric field22. Although the ASE phenomena in samples HU and HW are similar, quite different behaviors of temperature-dependent ASE spectrum in sample HB can be observed in Figure 6(c). At low temperatures, four clear peaks with the spacing corresponding to the free spectral range of Fabry-Perot resonance can be observed. As shown in the thick curves, at 10 K the CW PL spectrum is essentially located between the two peaks on the lower energy side with its side-lobe coinciding with the lowest-energy peak in the ASE measurement. As temperature increases, the separations

356

Cheng, Chen, Yang

among the four peaks become smaller. Also, the two peaks at the two spectral ends diminish with increasing temperature. The two peaks in the middle evolve into a major peak around 2.7 eV at room temperature. The CW PL spectrum at room temperature coincides essentially well with the diminishing lowest-energy peak in the ASE spectrum. 3.2. Cathodoluminescence In Figure 7(a)-(c), we show the CL images of the three samples. In samples HU and HW, the sharp sub-micron bright light spots correspond to the indium-rich clusters. The background distributions of contrast correspond to slowly varying potential fluctuations in the two samples. One can clearly see that slightly more indium-rich clusters and significantly stronger potential fluctuation exist in sample HW. Compared with samples HU and HW, the CL image of sample HB shows even more sharp bright spots and an even stronger potential fluctuation. In particular, a more discrete-like potential distribution exists in sample HB. In other words, distributions of island-like structures exist around the QW layers. It should be noted that the acceleration voltage of the used electron beam is 2 kV, which corresponds to the kinetic energy that can just excite the quantum wells, not into the GaN buffer layer. In Figures 8(a)-(c), we show the CL spectra from selected areas of the sample surfaces. The smooth, thin, continuous curves (labeled "large area" in the figures) were obtained from photon emission regions of 1 jj.m . Noisy continuous curves, dashed, and dash-dotted curves (labeled as "small spots") were obtained from sub-micron bright spots, as shown in Figure 7(a)-(c). Those bright spots are supposed to correspond to the indium-rich clusters, as to be further demonstrated by the red spots in Figures 9(a)-(c). For comparison, the CW PL spectra at room temperature were also plotted as thick continuous curves in Figures 8(a)(c). They are always located on the low-energy side of the CL spectra. In sample HU, shown in Figure 8(a), the spectra from the sub-micron bright spots contribute to the lower energy side of the overall spectrum (the small kink on the lower-energy side). These contributions are consistent with the CW PL spectrum.

InGaN/GaN Quantum Wells - Clustering Nanostructures, Silicon Doping

Fig. 7. (a) Typical CL image of sample HU at room temperature.

Fig. 7. (b) Typical CL image of sample HW at room temperature.

357

358

Cheng, Chen, Yang

Fig. 7. (c) Typical CL image of sample HB at room temperature.

This result implies that the indium clusters in sample HU, which contribute to PL, are not sharp enough to generate significant quantum effects. They simply have higher indium compositions such that the emitted spectra are red shifted. In this sample, the major light emission originates from the background area, instead of the sub-micron bright spots. On the other hand, in sample HW, the small bright spots contribute more significantly in the overall photon emission intensity. Also, they contribute to the high-energy side of the overall spectrum (see Figure 8(b)). This result implies that the clusters in sample HW are sharp enough for generating a certain quantum effect. Then, in sample HB, the small bright spots contribute to both sides of the overall spectrum, implying that the clusters in this sample have various sizes, shapes, and compositions (see Figure 8(c)). The overall spectrum is basically composed of the contributions from those clusters. Because the photon emission from sample HB is significantly blue-shifted, we speculate that the quantum effect plays an important role in the clusters of this sample.

InGaN/GaN Quantum Wells - Clustering Nanostructures, Silicon Doping

359

Wavelength (nm) 600 575 550 525 500

475 450

. Room Temp.

C

I I 0 20 21 22 23 24 25 26 27 28

Photon Energy (eV)

Fig. 8. (a) Normalized CL spectra of sample HU (a) detected under different conditions.

Wavelength (nm) 600 575

550

525

500

475 i

Room Temp.

.»

*

i

HW

i [i

Norma lized Int ens

450 •

IV X20.5

CWPL^jpjf 1m

I1)

\\ X7.4 1\ \ CL (small spots, l\V .r large area)

\ v> \ — •

2.0

^ i

2.1

• — i — • — i — • — i —

2.2

2.3

2.4

i—r-

2.5

V T—

2.6

2.7

2.8

Photon Energy (eV) Fig. 8. (b) Normalized CL spectra of sample HW (b) detected under different conditions.

360

Cheng, Chen, Yang

W a v e l e n g t h (n m ) 575

550

525

500

475

450

HB c o

CW

X10.1 X8.3

PL

CL (sm all spots, large area)

0 N

H E

2.1

2.2

2.3

2.4

2.5

2.6

2.7

2.8

2.9

Photon Energy (eV) Fig. 8. (c) Normalized CL spectra of sample HB (c) detected under different conditions.

3.3. Strain State Analysis In Figures 9(a)-(c), we show the typical SSA images of samples HU, HW, and HB, respectively. In these SSA images, line scans were performed along the shown white lines. Here, the line scan values 1 and 1.1, respectively, represent indium compositions of 0 and 60 % (estimated based on the assumption of a specimen thickness larger than 30 nm). Different colors stand for various ranges of indium composition, as shown in the legends. As shown in Figure 9(a) for sample HU, although the QW interface is blurred, indium is basically confined within the well. Here, within the well a couple red spots of indium aggregation can be observed. From the line-scan results, one can observe quite a weak fluctuation in indium composition either along the well layer or in the growth direction. The fluctuation contrast (the difference between the maximum and minimum in the scan range) is about 0.03 along the QW layer. Then, in Figure 9(b) for sample HW, the QW is not as well shaped as sample HU. The indium composition fluctuations in both directions are relatively stronger in comparing with sample HU. In particular, more

InGaN/GaN Quantum Wells - Clustering Nanostructures, Silicon Doping

361

indium-aggregated clusters can be observed within the well layer. The fluctuation contrast is now around 0.04 along the QW layer. As shown in Figure 9(c), the SSA image of sample HB shows quite a different nanostructure from the other two samples. Here, the QW layer becomes unclear. Instead, a distribution of clusters of different sizes and shapes exists. The indium composition fluctuation (the contrast is now around 0.08 along the QW layer) is much stronger than those of the other two samples, implying stronger carrier localization for effective recombination. 4. Discussions From the strain-state analysis results in Figures 9(a)-(c), we can build a model, as schematically shown in Figure 10, for potential variation along a quantum well layer. Here, in part (a), which is drawn for describing sample HU, potential fluctuations are weak. A potential minimum, which stands for a cluster, is surrounded by secondary minima of shallow barriers in between. In part (b), which is drawn for sample HW, potential variations are relatively stronger. In this situation, a potential minimum is also surrounded by secondary minima. However, the barriers between the minima are now quite steep such that certain energy is required for carrier transport between the minima. Then, in Figure 10(c) for sample HB, sharp minima with few secondary minima are drawn. Therefore, in the cases of Figures 10 (a) and (b), carriers can relax down to the absolute potential minima with a cascading process through the secondary minima. The difference between samples HU and HW is that more energy is required for overcoming the barriers in sample HW. However, such a cascading process is less significant in the case of Figure 10(c). In other words, carriers in sample HB can relax down to the individual steep potential minima directly. Because of the relatively shallow potential minima in sample HU, the quantum effect is insignificant in this sample that is consistent with the observation in Figure 9(a). Also, because of the relatively sharper secondary minima or local absolute minima in samples HW and HB, more bright light spots are observed in Figures 8(b) and (c). The model shown in Figure 10 will be used for interpreting the measurement results in Figures 1-7.

362

Cheng, Chen, Yang

lam

11)15

1031)

l.(M5

1J0S0

1.1QQ

Fig. 9. (a) Typical SSA image of sample HU (a). The color legend indicates the estimated indium mole fractions (1.0 = 0 %, 1.1 = 60 %).

1.000 1.015

1.030

1.045

1.060

1.100

Fig. 9. (b) Typical SSA image of sample HW (b). The color legend indicates the estimated indium mole fractions (1.0 = 0 %, 1.1 = 60 %).

1.0011 1.1115 1.1)30

1.045

I.Odd

1.100

Fig. 9. (c) Typical SSA image of sample HB (c). The color legend indicates the estimated indium mole fractions (1.0 = 0 %, 1.1 = 60 %).

InGaN/GaN Quantum Wells - Clustering Nanostructures, Silicon Doping

363

Fig. 10. Schematic drawings for the potential fluctuations in samples HU (a), HW (b), and HB (c).

Based on the model in Figure 10, potential fluctuates mildly in sample HU. As shown in Figure 1, the temperature-dependent PL spectral peak energy reveals a typical S-shape behavior, which has been widely reported. In sample HW, a stronger potential fluctuation is observed. The slight blue shit of PL peak below 125 K and the small blue shifts of the CL spectra from the bright spots, as shown in Figure 8(b), imply that a certain degree of quantum confinement effect may exist in sample HW. However, the most significant feature of this sample in Figure 1 is the absence of the S-shape behavior. The S-shape has been attributed to the carrier-localization effect, QCSE or both. With silicon doping in either wells or barriers (carriers in barriers tend to transport to wells), carrier screening and possibly strain energy relaxation can reduce the QCSE, leading to a less prominent S-shape behavior. Therefore, the S-shape behaviors in samples HU and HW essentially disappear. With this observation, one may speculate that in the widely reported studies on undoped InGaN/GaN QW samples, the S-shape behavior is mainly caused by the QCSE. On the other hand, the strong carrier localization effect in a strongly clustering sample, such as sample HB, is mainly responsible for the significant blue shift of PL peak and the enhancement of radiative efficiency. Such results are clearly shown in Figures 1 and 2. In this sample, strong carrier localization with a significant quantum confinement effect is expected.

364

Cheng, Chen, Yang

The analysis results of the Arrhenius plots, as shown in Table 1, can also be used for the similar interpretations. The values of EA2 represent the degree of carrier localization. The strongest localization effect in sample HB, followed by sample HW, can be clearly seen. From the values of y one can see that the density of non-radiative recombination center in sample HW is the highest among the three samples. This result can be attributed to the silicon doping in the wells, which are the light emission regions. Silicon doping in barriers does not seem to increase the density of non-radiative recombination center. The increase of this density due to silicon doping in sample HW may cancel the positive effect of carrier localization such that the radiative efficiency is not significantly improved in this sample, as shown in Figure 2. In Figure 3, the excitation intensity dependent PL peak energies are shown. Here, the increasing trend (only about 2.5 meV in the measurement range) of sample HU is simply due to the carrier screening effect for reducing the QCSE. In sample HW, because a plenty of carriers due to doping exist in the wells already and the QCSE has been essentially relaxed, further increase of carrier density results in the band filling effect and the significant blue shift of PL peak (about 20 meV shift in the measurement range). However, in sample HB, photo-generated carriers added to the even higher carrier density already existent (due to a larger volume of barrier doping), which are strongly localized in clusters, lead to a certain degree of local renormalization effect. Therefore, a red shift (about 7 meV) can be observed, as shown in Figure 3. As shown in Figure 5, the DEDPLE signal intensity decreases with increasing detection photon energy in sample HU. This trend can be attributed to the relatively weaker potential fluctuation in this sample. With such an energy level distribution, photo-generated carriers in higher energy levels can easily transport to the absolute potential minimum for recombination within a certain region. Hence, when the detection photon energy is low, most carriers can contribute to photon emission. When the detection photon energy becomes higher, fewer carriers can actually recombine at this relatively higher energy level and hence the DEDPLE signal becomes weaker. In this situation, PL spectral peak is always located at the absolute potential minimum and is independent of the excitation level, as shown in Figure 4.

InGaN/GaN Quantum Wells - Clustering Nanostructures, Silicon Doping

365

Then, as mentioned earlier for sample HW, it requires a certain amount of energy and hence is more difficult for carriers to transport from a secondary minimum to another of a deeper level. Therefore, the majority of photo-generated carriers can be trapped by secondary minima of relatively higher energy. When the detection photon energy is high, stronger DEDPLE signals are recorded, as shown in Figure 5. In this case, as the excitation energy becomes lower in the EEDPL measurement, more carriers can actually be trapped in secondary potential minima of relatively higher levels such that the PL spectral peak energy increases with decreasing excitation energy, as shown in Figure 4. In sample HB, because of the strongly clustering structure with an island-by-island configuration, when carriers are generated at high InGaN energy levels, they can transport directly into individual potential minima without a cascading relaxation process. In this situation, carrier distributions after relaxation among the shallow and deep potential minima can be quite even such that the DEDPLE signal intensity is almost independent of the detection photon energy (see Figure 5(a)). Also, as the excitation photon energy decreases, the local potential minima of relatively lower levels can collect more carriers and hence the EEDPL peak position decreases, as shown in Figure 4. The multiple-peak feature of ASE spectra in high-indium-content InGaN/GaN QWs has been observed and attributed to various inter-band transitions between different quantized levels in the QWs23. However, in our results only sample HB shows the multi-peak feature. The ASE spectra of samples HU and HW at low temperatures show only two peaks. These two peaks are identified as the contributions of QD (cluster) and QW states, respectively. Because of the high photo-generated carrier density in the sample in the ASE measurement, the QD states are completely filled in samples HU and HW. In this situation, the QCSE in sample HU is also relaxed. Therefore, either effect of carrier localization or QCSE is ineffective in generating the S-shape behavior for the lowerenergy peak at low temperatures in Figure 6(a). However, because the QW states are not completely filled, the carrier localization effect can still produce the S-shape behavior for the higher-energy peak in the low temperature range. The similar arguments can be applied to sample HW. In sample HW, more carrier supply, from both doping and photo-

366

Cheng, Chen, Yang

generation, leads to more complete filling of QW states such that the Sshape behavior of the higher-energy peak in Figure 6(b) becomes less significant. In both Figures 6(a) and (b), as temperature increases, the higher-energy peak becomes relatively stronger due to the flow of thermalized carriers from the QD states into the QW states. When temperature approaches 200 K, the liquidation of carriers among the QD and QW states leads to the formation of a gain spectrum. In this situation, the peak positions are determined by the Fabry-Perot resonance in the crystal growth direction. The multiple-peak feature in Figure 6(c) is quite unique among the three samples. The multiple-peak ASE feature of sample HB is attributed to the formation of indium-rich clusters of quantum nature with broad distributions in size, shape, indium composition, and hence a broad distribution of photon emission spectrum. This interpretation is confirmed by the spectral distribution shown in Figure 8(c). Therefore, when a plenty of carriers are generated, a broad gain spectrum is produced. In the low temperature range, the peak positions are still dominated by the Fabry-Perot resonance wavelengths within the gain spectrum. As temperature increases, the liquidation of thermalized carriers tends to homogenize the gain spectrum. In this situation, the gain spectrum tends to be centralized with increasing temperature. Eventually, a major peak appears in the ASE process, as shown at the 300 K curve of Figure 6(c). Based on this observation, it is speculated that carrier thermalization may lead to a certain advantage for a laser system in such a material structure to reach its threshold condition, if the effect of carrier localization is strong enough to overcome the luminescence decay with temperature and/or the defect density is low. 5. Conclusions In summary, we have compared the results of PL, DEDPLE, EEDPL, ASE, CL and SSA of three InGaN/GaN QW samples with un-doped, well-doped, and barrier-doped structures. Based on the SSA and CL images, a nanostructure model was built for describing the potential fluctuation differences between the three samples. In the barrier-doped sample, strongly clustering nanostructures with individual steep potential

InGaN/GaN Quantum Wells - Clustering Nanostructures, Silicon Doping

367

minima, which generated significant quantum confinement effects, were assumed. In the undoped and well-doped samples, relatively weaker composition fluctuations, in which carriers relaxed through a cascading process, were proposed. Between the undoped and well-doped samples, the potential fluctuation in the well-doped sample is relatively steeper such that a certain extent of quantum confinement existed. Such variations in nanostructure resulted in different carrier transport processes between the coexistent quantum dot and quantum well states. The proposed model well explained the PL, DEDPLE, EEDPL, ASE, and CL observations. In particular, the PL results provided us clues for speculating that the S-shape behavior of PL peak position is dominated by the QCSE in an undoped InGaN/GaN QW structure. However, carrier localization is more effective in blue-shifting luminescence and improving radiative efficiency of a sample, when compared with the relaxation of QCSE. Also, the ASE results showed the temperaturedependent evolution of gain spectrum due to the liquidation of thermalized carriers. Different nanostructures resulted in different spectral variation trends. Acknowledgments This research was supported by National Science Council, The Republic of China, under the grant of NSC 93-2210-M-002-006 and NSC 942215-E-002-015, and by US Air Force under the contracts AOARD-044026 and AOARD-05-4085. References 1. S. Chichibu, D. A. Cohen, M. P. Mack, A. C. Abare, P. Kozodoy, M. Minsky, S. Fleischer, S. Keller, J. E. Bowers, U. K. Mishra, L. A. Coldren, D. R. Clarke, and S. P. DenBaars, Appl. Phys. Lett. 73,496 (1998). 2. Y. H. Cho, J. J. Song, S. Keller, M. S. Minsky, E. Hu, U. K. Mishra, and S. P. DenBaars, Appl. Phys. Lett. 73, 1128 (1998). 3. M. Y. Ryu, Y. J. Yu, E. J. Shin, P. W. Yu, J. I. Lee, S. K. Yu, E. S. Oh, O. H. Nam, C. S. Sone, Y. J. Park, and T. I. Kim, Solid State Commun. 116, 675 (2000). 4. J. Dalfors, J. P. Bergman, P. O. Holtz, B. E. Sernelius, B. Monemar, H. Amano and I. Akasaki, Appl. Phys. Lett. 74, 3299 (1999).

368

Cheng, Chen, Yang

5. M. S. Minsky, S. Chichibu, S. B. Fleischer, A. C. Abare, J. E. Bowers, E. L. Hu, S. Keller, U. K. Mishra and S. P. DenBaar, Jpn. J. Appl. Phys. 37, L1362 (1998). 6. K. Uchida, T. Tang, S. Goto, T. Mishima, A. Niwa, J. Gotoh, Appl. Phys. Lett. 74, 1153(1999). 7. Y. C. Cheng, Cheng-Hua Tseng, Chen Hsu, Kung-Jen Ma, Shih-Wei Feng, EnChiang Lin, C. C. Yang, and Jen-Inn Chyi, J. Electron. Mater. 32, 375 (2003). 8. Y. H. Cho, F. Fedler, R. J. Hauenstein, G. H. Park, J. J. Song, S. Keller, U. K. Mishra and S. P. DenBaars, J. Appl. Phys. 85, 3006 (1999). 9. S. Ruvimov, Z. Liliental-Weber, T. Suski, J. W. Ager III, J. Washburn, J. Krueger, C. Kisielowski, E. R. Weber, H. Amano and I. Akasaki, Appl. Phys. Lett. 69, 990 (1996). 10. C. K. Choi, Y. H. Kwon, B. D. Little, G. H. Gainer, J. J. Song, Y. C. Chang, S. Keller, U. K. Mishra and S. P. DenBaars, Phys. Rev. B, 64, 245339-1 (2001). 11. I. Ho, G. B. Stringfellow, Appl. Phys. Lett. 69, 2701 (1996). 12. Y. S. Lin, K. J. Ma, C. Hsu, S. W. Feng, Y. C. Cheng, C. C. Liao, C. C. Yang, C. C. Chuo, C. M. Lee, and J. I. Chyi, Appl. Phys. Lett. 77, 2988 (2000). 13. Y. S. Lin, K. J. Ma, C. Hsu, Y. Y. Chung, C. W. Liu, S. W. Feng, Y. C. Cheng, M. H. Mao, C. C. Yang, H. W. Chuang, C. T. Kuo, J. S. Tsang, and T. E. Weirich, Appl. Phys. Lett. 80, 2571 (2002). 14. S. W. Feng, Y. C. Cheng, Y. Y. Chung, E. C. Lin, C. C. Yang, C. C. Yan, Y. S. Lin, C. Hsu, K. J. Ma and H. X. Jiang, Appl. Phys. Lett. 82, 1377 (2003). 15. Y. Y. Chung, Y. S. Lin, S. W. Feng, Y. C. Cheng, E. C. Lin, C. C. Yang, K. J. Ma, H. W. Chuang, C. T. Kuo, and J. S. Tsang, J. Appl. Phys. 93, 9693 (2003). 16. S. W. Feng, E. C. Lin, T. Y. Tang, Y. C. Cheng, H. C. Wang, C. C. Yang, K. J. Ma, C. H. Shen, L. C. Chen, K. H. Kim, J. Y. Lin, and H. X. Jiang, Appl. Phys. Lett. 83, 3906 (2003). 17. D. Gerthsen, B. Neubauer, A. Rosenauer, T. Stephan, H. Kalt, O. Schon and M. Heuken, Appl. Phys. Lett. 69, 2701 (1996). 18. R. Seitz, C. Gaspar, T. Monteiro, E. Pereira, M. Leroux, B. Beaumont, and P. Gibart, J. Crystal Growth 189/190, 546 (1998). 19. S. W. Feng, Y. C. Cheng, Y. Y. Chung, C. C. Yang, Y. S. Lin, C. Hsu, K. J. Ma, and J. I. Chyi, J. Appl. Phys. 92,4441, (2002). 20. C. C. Liao, S. W. Feng, C. C. Yang, Y. S. Lin, K. J. Ma, C. C. Chuo, C. M. Lee, and J. I. Chyi, Appl. Phys. Lett. 76, 318, (2000). 21. Y. H. Cho, G. H. Gainer, A. J. Fischer, J. J. Song, S. Keller, U. K. Mishra and S. P. DenBarrs, Appl. Phys. Lett. 73, 1370 (1998). 22. S. F. Chichibu, A. C. Abare, M. S. Minsky, S. Keller, S. B. Fleischer, J. E. Bowers, E. Hu, U. K. Mishra, L. A. Coldren, and S. P. DenBaars, T. Sota, Appl. Phys. Lett. 73, 2006 (1998). 23. C. C. Chen, H. W. Chuang, G. C. Chi, C. C. Chuo and J. I. Chyi, Appl. Phys. Lett. 11, 3758 (2000).

CHAPTER 11 III-NITRIDES MICRO- AND NANOSTRUCTURES

Hock M. Ng and Aref Chowdhury Bell Laboratories, Lucent Technologies 600 Mountain Avenue, Murray Hill, NJ 07974, U.S.A. E-mail: [email protected] Nanotechnology applied to the realm of compound semiconductors can result in novel physics as well as interesting applications, particularly in the field of optoelectronics. In this chapter, we will review the current status of Ill-nitrides (GaN, InN, A1N, and related alloys) micro- and nanostructures. These structures can be formed either during the growth process or post-growth stage using dry or wet etching techniques. The polarity of GaN, whether Ga- or N-polar, plays a major role in the behavior of the material when subjected to a chemical etchant. Certain chemical etchants have been shown to be selective towards one polarity of GaN but not the other. This selectivity can be exploited as a means to form interesting microstructures. Examples of applications for IIInitride micro- and nanostructures will also be discussed. 1.

Introduction and Overview

Nanotechnology applied to semiconductors can be divided into nanomaterials - the synthesis of structures possessing nanoscale dimensions and nanophotonics - the interaction of light with nanostructures having dimensions comparable to or smaller than the wavelength of light in the medium. The ability to control the dimensions of semiconductor materials at the nanoscale allows one to essentially tailor the macroscopic properties of the material by manipulating its physical dimensions.

369

370

Ng, Chowdhury

With the advent of the information age, the demand for information storage and displays has been the driving force for much of compound semiconductor development. For GaN and InGaN, the commercial applications of visible light-emitting diodes (LEDs) and blue lasers for information storage and retrieval have driven the pace of research in both industry and academia. AlGaN alloys with larger bandgap energies are being developed for emitters and detectors in the ultraviolet spectral region for a wide variety of applications including biosensing and solidstate lighting. However, the development of nanotechnology for IIInitrides is still in its infancy. The objective of this chapter is to provide a broad overview of the development of Ill-nitrides nanotechnology. Different approaches to forming micro- and nanostructures will be reviewed. The study of microstructures is the intermediate step towards scaling down the dimensions to approach the nanoscale. In general, there are two ways of forming semiconductor nanostructures - the bottom-up and top-down approaches. In section 2, nanostructures formed by epitaxial growth will be discussed. This will be followed by section 3 that discusses another route to forming nanostructures by etching or regrowth, or a combination of the two. Examples of selective chemical etching based on the polarity of GaN will be discussed. Finally, a few different examples of the applications of these structures will be exemplified in section 4 and the chapter is concluded with an outlook for the future. 2. Nanostructures by Epitaxy Epitaxial growth of nanostructures provides a way to create structures where electrons or photons can be confined in one, two or three dimensions corresponding to quantum wells, quantum wires and quantum dots. Nanostructures formed by epitaxial growth are assembled from the bottom-up. This is in contrast with the top-down approach where a thin film is grown first and then processed into nanostructures. With the bottom-up process, unique conditions during epitaxy are exploited to induce three-dimensional growth. For example, GaN quantum dots can be formed during molecular beam epitaxy (MBE) by the strain driven Stranski-Krastanov growth mode.1 The compressive

Ill-Nitrides Micro- and Nanostructures

371

stress generated when a material with a larger in-plane lattice constant is grown on top of a layer with a smaller in-plane lattice constant drives the formation of the quantum dots. Quantum dots have also been studied in other semiconductor systems such as InAs/GaAs,2 InGaAs/GaAs,3 InP/InGaP,4 and Ge/Si.5 Recently, semiconductor nanowires have become an area of growing interest. For instance, GaN nanowires have been formed with the direct reaction of metallic gallium and ammonia gas at a temperature of 900 °C using nickel as a catalyst.6 This growth mechanism was given the name Vapor-Liquid-Solid (VLS) by Wagner and Ellis at Bell Laboratories.7 In some instances when forming compound semiconductor nanowires, one of the constituents of the compound can be used as a self-catalyst thus alleviating the need for a separate metal catalyst. For example, in growing GaN nanowires, Ga droplets have been used for self-catalytic VLS growth.8'9 GaN nanowires have even been observed to lase under optical pumping.6 However, one major challenge that remains is finding a reliable technique of processing these nanowires into useful electrically driven devices. In this chapter, we will not go into further details about nanowires formed by chemical synthesis. For a recent review of semiconductor nanowires and nanotubes, readers are referred to Reference 10. Another example of fabricating nanostructures by epitaxy is selective area growth (SAG). Using metalorganic chemical vapor deposition (MOCVD), SAG is commonly combined with epitaxial lateral overgrowth (ELO) as a means to achieving a reduction in threading dislocation density for GaN layers.11 However, SAG can also form interesting structures such as GaN hexagonal pillars and pyramids.1213 Typically, a thin film of GaN is grown on a substrate such as (0001) sapphire or (111) silicon. The sample is then removed from the growth reactor and a Si0 2 layer is deposited. This is followed by photolithography and etching of the Si0 2 to create an array of openings in the shape of holes or stripes. The sample is then reinserted into the growth chamber for a second growth of GaN. During the high temperature growth process, epitaxy of GaN occurs within the openings in the Si0 2 layer but not on top of the Si0 2 mask. The resulting structures can be controlled by adjusting growth parameters such as the

372

Ng, Chowdhury

temperature, reactor pressure and IE/V ratio. Lasing by optical pumping has been observed in GaN pyramids as a testament to their crystal quality.14 Another example of the bottom-up approach is the self-assembled nanocolumnar growth occurring under certain growth conditions during MBE. In particular, the formation of nanocolumns is favored when the V/III ratio is high corresponding to growth in the nitrogen-rich regime. Under these conditions, columnar growth is preferred to two-dimensional film growth. The group-III adatom mobility is greatly reduced under high N flux. The proposed mechanism of formation is the nucleation induced group III metallic droplets similar to the self-catalytic VLS mechanism. Nanocolumns of GaN,15 AlGaN,16 and InN17 have been demonstrated. Some examples of InN nanocolumnar growth are highlighted here using results from the authors' laboratory. For InN, there are only a few studies of nanowires grown primarily by the VLS mechanism where a catalyst metal is first deposited on the substrate.18"20 Compared to VLS, epitaxial growth techniques offer an alternative to obtain better control over nanowire alignment and formation of heterostructures. For instance, AlGaN/GaN heterostructure nanocolumns were reported by Yoshizawa et al. by MBE growth.21 Figure 1 shows an example of a dense aggregate of InN nanocolumns grown on (0001) sapphire by plasma-assisted MBE with growth temperatures in the range of 380 to 430 °C. The cross-sectional scanning electron microscope (SEM) image shows that the InN nanocolumns are aligned along the growth axis which is the [0001] direction. The polarity of the InN cannot be ascertained at this time. The average diameter of the nanocolumns was about 75 nm. X-ray diffraction measurements showed that the nanocolumns possess the wurtzite crystal structure and are oriented with the c-axis perpendicular to the substrate surface. The wurtzite crystal structure was also confirmed by selected area electron diffraction measurements. The measured c-lattice constant of 5.70 A is in good agreement with the value reported by Yamaguchi et al.21 The InN nanocolumns do not have very good adhesion to the sapphire substrate. They can be easily delaminated as seen in Fig. 1(b) where part of the underlying sapphire substrate is exposed. This may be due to the

Ill-Nitrides Micro- and Nanostructures

373

large in-plane lattice mismatch between InN and sapphire (-25%) resulting in a high built-in stress at the interface between the nanocolumns and the substrate. The effect of the In/N ratio plays a major role in maintaining the growth of the nanocolumns. Figure 2 shows the progressive changes in the surface morphology of four different InN samples (N668, N671, N672, N581) as the In/N flux ratio was increased (decreasing the V/III ratio). The average diameter of the InN nanocolumns becomes larger and eventually the neighboring columns start to coalesce and form a twodimensional (2D) film.

Fig. 1. (a) A cross-sectional scanning electron microscope image of InN nanocolumns grown on (0001) sapphire, (b) A lower magnification top view of the InN nanocolumns.

Going from sample N668 to N671, the In temperature was increased by 10 °C corresponding to an increase in the In flux by about 60 %. The average diameter of the nanocolumas increased from 75 to 200 nm. For sample N672, the In flux was increased along with a slight decrease in the growth temperature (from 422 to 382 °C). As seen in Fig. 2(c), the diameters of the nanocolumns became less uniform with the onset of the early stages of coalescence. The growth mode went through a transition from 3D to 2D growth as the In flux was further increased while keeping the N flux constant shown in Fig. 2(d). When the In flux was increased further, In droplets

374

Ng, Chowdhury

were formed on the sample surface similar to the case of Ga droplets forming under extremely Ga-rich growth conditions.23 In general, the growth temperature was found to be a less influencing factor for the formation of nanocolumns than the In/N ratio. (a)

(b)

Fig. 2. A series of SEM images showing the changes in the surface morphology of the InN samples as the In cell temperature (and the corresponding In flux) was increased, (a) N668: TIn = 740 °C, (b) N671: TIn = 750 °C, (c) N672: TIn = 770 °C, (d) N581: Tln = 790 °C. [from Ref. 17] Reproduced by permission of The Electrochemical Society, Inc.

When the growth temperature increases, competition between growth and decomposition of InN comes into play. One InN sample that was left in vacuum at 406 °C for several hours without striking the nitrogen plasma was completely decomposed, leaving only metallic In droplets on the surface of the sapphire substrate. The fact that there were In droplets

Ill-Nitrides Micro- and Nanostructures

375

left on the surface was an indication that the decomposition rate of InN is higher than the desorption of In from the sample surface at that temperature. This observation is consistent with early work by MacChesney et al. showing that the decomposition of InN in a nitrogen ambient at atmospheric pressure occurs in the vicinity of 500 °C.24 3. Nanostructures by Selective Etching and Regrowth We now turn our attention to the issue of forming micro- and nanostructures by means of etching. Due to the large bond strengths of the Ill-nitrides, it is difficult to etch the material using chemical solutions. To give an idea of the relative magnitude of the bond strengths, we can compare the atomic bond strengths of GaN, A1N and InN with several other compound semiconductors as shown in Table I. To break the strong bonds between the group III and nitrogen atoms, a highly energetic process is required. Dry etching techniques involving a chlorine-based plasma are most commonly employed for etching IIInitrides. The energetic ions in the plasma assist in the etching process by means of sputtering. Among the different variants of dry etching include reactive ion etching (RIE), inductively-coupled plasma (ICP) etching or a combination of the two. A review comparing different dry etching techniques for Hi-nitrides can be found in References 26 and 27. Table I: Bond strengths of various compound semiconductors. [From Ref. 25]

Semiconductor

Bond Strength (eV)

GaN

8.96

A1N

11.52

InN

7.72

GaAs

6.52

InP

6.96

376

Ng, Chowdhury

3.1. Photoelectrochemical Etching We now turn our attention to chemical etching of GaN. The use of chemical etchants has been primarily limited to bases such as KOH28 and acids such as H 3 P0 4 . 29 In photoelectrochemical (PEC) etching, a light source with photons having energies exceeding the bandgap energy of the material is required. For GaN, this requires an ultraviolet (UV) light source with wavelengths shorter than 365 nm. A mercury lamp is commonly used as the light source. A platinum wire or grid serves as the counter electrode with the GaN sample as the working electrode.30 The reaction that occurs for etching in KOH is as follows: GaN + 60H " + 3h+ - » GaO\~ + 0.5N2 + 3H20 For etching in phosphoric acid, the reaction is: GaN + 3h+ ->Ga 3 + + 0.5 N2

Fig. 3. A scanning electron micrograph of an etched GaN film after PEC etching in KOH. The whiskers have diameters between 10 and 50 nm and lengths of 1 \im or more. Reprinted with permission from C. Youtsey, L. T. Romano, and I. Adesida Applied Physics Utters, 73, 797 (1998). Copyright 1998, American Institute of Physics.

Ill-Nitrides Micro- and Nanostructures

311

The above-bandgap light generates electron-hole pairs in GaN. The holes participate in surface oxidation of GaN which is then dissolved in the KOH. This process works only for n-type GaN. Under certain conditions, nanoscale whisker-like structures were obtained after PEC etching.31 An example of this is shown in Fig. 3. In cross-sectional transmission electron microscopy (TEM) studies, individual whiskers were found to have a threading dislocation in the center.31 The explanation is given that the dislocation acts as a recombination center for the photo-generated electrons and holes thus reducing the availability of free holes for the oxidation process. As a result the material around the dislocation was not etched. For p-type GaN, the minority carriers are electrons which cannot participate in the oxidation process. In order for PEC to work for p-type GaN, a positive bias has to be applied to the sample to reduce the surface band bending and draw holes to the surface.32 3.2. Polarity Selective Chemical Etching (PSCE) In the case of chemical etching without illumination of above-bandgap light, the situation is quite different. There is a strong dependence on the polarity of the GaN sample. Being a non-centrosymmetric crystal, GaN has two different polarities, namely Ga-polar (0001) and N-polar (000 1). When exposed to a KOH solution, only N-polar GaN is etched whereas Ga-polar GaN remains intact.33 Because of the selectivity of the etchant towards the polarity of the material, we term this as polarity selective chemical etching (PSCE).34 In order to better illustrate the effects of PSCE, a "polarity-patterned" GaN sample with adjacent Ga- and N-polar regions on a single substrate was studied. Figure 4 shows the flow chart of the process for fabricating a polarity-patterned GaN sample. GaN grown on top of an AIN buffer layer has Ga-polarity whereas N-polarity material results from growth directly on the sapphire substrate. An SEM image depicting two adjacent regions of the Ga- and Npolarity GaN before etching in KOH is shown in Fig. 5(a). It is seen that

378

Ng, Chowdhury

the as-grown surface morphology of the Ga-polar GaN is smooth while the N-polar surface displays roughness at the submicron scale. The difference in the surface morphology can be attributed to the fact that the Ga-polar material grows on top of an A1N buffer layer (lattice mismatch ~ 2.5%) whereas the N-polar material is nucleated directly on sapphire (lattice mismatch ~ 14%) at the high growth temperature. It is noted also that the interface between the Ga- and N-polar GaN is sharply defined. There are some small columnar growths in both the Ga- and N-polar regions that are believed to be inversion domains. (d)

(a)

(b) (e) (c)

•

A1N

| Ga-polar GaN

Photoresist

I | Sapphire ^ | N-polar GaN Fig. 4. Process flow for fabricating GaN microstructures by PSCE: (a) GaN/AIN growth on sapphire by MBE, (b) formation of stripe/hole pattern by optical lithography, (c) plasma etch to expose the sapphire substrate, (d) second growth of GaN by MBE, and (e) PSCE step to remove N-polar GaN. [From Ref. 34]

III-Nitrides Micro- and Nanostructures

379

Fig. 5(b) displays an SEM image after the sample was etched for 45 minutes at 90°C in 2M KOH. As seen in the figure, the KOH solution selectively etched the N-polar GaN but not the Ga-polar GaN. Comparing Figs. 5(a) and (b), it can be seen that the surface morphology of the Ga-polar regions remains smooth while nanotip pyramids are formed in the N-polar regions.35

Fig. 5. (a) SEM image taken at a 45° angle of the sample before PSCE showing the smooth and rough surfaces of Ga- and N-polar GaN, respectively, (b) After PSCE, the Ga-polar regions remain smooth while hexagonal pyramids were formed in the N-polar regions. Reprinted with permission from Hock M. Ng, Nils G. Weimann, and Aref Chowdhury, Journal of Applied Physics, 94, 650 (2003). Copyright 2003, American Institute of Physics.

380

Ng, Chowdhury

The PSCE process occurs through the interaction of the OH" ions with the Ga atoms to form Ga 2 0 3 , which then dissolves in KOH [36]. The process can be described as: KOH 2GaN + 3 H 2 0

•

Ga 2 0 3 + 2NH 3

The reason the Ga-polar GaN was not etched by the KOH solution is because the OH' ions are repelled by the negatively-charged triple dangling bonds of N at the surface, which was substantiated by x-ray photoelectron spectroscopy studies.36 Therefore, if the Ga-polar GaN is Ga-terminated, the etch stops after the initial Ga layer is removed. However, for N-polar GaN, each N atom at the termination layer only has a single dangling bond. Therefore, the probability of an OH" ion getting in to attack the Ga-N bond below is higher. This probability is increased at higher solution temperatures where the OH" ions have higher kinetic energy resulting in a higher etch rate. Increasing the KOH molarity also increases the etch rate as a consequence of the increase in the concentration of OH" ions in the solution. Individual pyramids can be isolated by etching the sample with a higher KOH concentration. In Figures 6(a) and (b), a high magnification SEM image of a single hexagonal pyramid is shown along with a schematic drawing. The faces of the pyramid are seen to be extremely smooth. Viewed from the top, each pyramid clearly shows a six-fold symmetry. The faces of the pyramids are the {10 1 1} planes as evidenced by the angle between the inclined edge and the base of the pyramid. The measured angle of around 58-60° is in good agreement with the calculated angle of 58.4° using the GaN lattice parameters of c = 5.185 A and a = 3.189 A. The formation of the pyramids indicates that the {10 1 1} surfaces are etched preferentially compared to the (000 1) surface. It is also noted that the tip of the pyramid is very sharp with a diameter measured to be less than 20 nm. The sharpness of the pyramid tips is maintained with prolonged etching without revealing the (000 1)

Ill-Nitrides Micro- and Nanostructures

381

planar surface. Thus, it is postulated that the {10 I T } surfaces have the lowest surface energy with respect to this etching process. The density of the pyramids decreases as a function of time and KOH concentration. For a fixed solution temperature and etch duration, increasing the KOH concentration from 1 to 8 M shows a saturation of the pyramid density (counting the number of pyramid tips) at around 3 x 10 cm"2. In addition, the pyramid size gets smaller with prolonged etching as well as with increasing the KOH concentration and solution temperature.

Fig. 6. (a) An SEM image showing an isolated pyramid on sapphire, (b) Schematic drawing of the hexagonal pyramid with six (10 1 1} sides. The shaded region denotes a cross-section through the apex of the pyramid. Reprinted with permission from Hock M. Ng, Nils G. Weimann, and Aref Chowdhury, Journal of Applied Physics 94, 650 (2003). Copyright 2003, American Institute of Physics.

382

Ng, Chowdhury

A parameter of interest for these etching experiments is the activation energy Ea. This is generally obtained by fitting an Arrhenius plot to R = R0 exp(- Ea lkBT), where R is the etch rate, R0 is a constant, kB is the Boltzmann's constant, and T is the temperature in Kelvin. In the case of the GaN etching, we believe that the etching rate is to first-order inversely proportional to the resultant pyramid density for a given etchant concentration and time. The physical explanation is that at the initial stage of the etching, the tips of the pyramids form throughout the surface. As the etching continues, many of the miniature pyramids form a cluster from where only a single pyramid emerges. This consolidation process of pyramids repeats itself until isolated pyramids are formed as shown in Fig. 6(a). Further etching will eventually etch away all the isolated pyramids. i—i—i—|—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—

E a =0.587 eV

10 7 E o

c

,

N

IO

8 X

-

T)

•g E CL-

10 9

i

2.5

2.6

.

i

2.7

•

i

2.8

.

i

2.9

•

i

3.0

.

i

.

3.1

i

3.2

.

i

3.3

•

i

3.4

3.5

1

1000/T(K" )

Fig. 7. Arrhenius plot of the inverse pyramid density. Reprinted with permission from Hock M. Ng, Nils G Weimann, and Aref Chowdhury, Journal of Applied Physics 94, 650 (2003). Copyright 2003, American Institute of Physics.

If the pyramid density is given by pA, the Arrhenius plot can now be fitted to

Ill-Nitrides Micro- and Nanostructures

=

exp{-Ea/kj)

383

(1)

where po is a constant. The fit to the Arrhenius plot is shown in Figure 7. For this particular analysis, four GaN samples of approximately equal area were separately etched in KOH of 2 M concentration for 15 minutes in a stir bath at 230 rpm. The temperature of the etching was 25°C, 50°C, 75°C and 100°C for the four samples, respectively. For each sample, the pyramid density was determined by averaging three randomly chosen areas. From Fig. 7, we determined that the activation energy is Ea ~ 0.587 eV. The etching activation energy of AIN was determined to be ~ 0.67 eV by other workers that used KOH as a constituent of their etchant.37 The ratio of AIN/GaN etching activation energy is ~ 1.141 which is in good agreement with the AIN/GaN atomic bond strength ratio (11.52 eV/8.96 eV = 1.286). Based on these findings, we predict that the etching activation energy for InN is - 0.51 eV. The behavior of GaN in KOH is independent of the method of growth. Similar observations that only N-polar GaN etches in KOH have been reported for MOCVD grown material.38 These pyramidal structures have also been integrated into LED structures in order to enhance the light extraction by reducing total internal reflection.39 3.3. ID and 2D Periodic Structures Formed by PSCE Structures having a periodic modulation of refractive indices are useful for a variety of photonic applications. An example of a periodic structure in one dimension is a distributed Bragg reflector (DBR) where repeating pairs of A/4 layers with high refractive index contrast are used as mirrors. High reflectivity DBRs have been demonstrated using AIN and GaN A/4 layers with maximum reflectivity over 99% and a bandwidth of 45 ran.40 For the case where the periodicity is extended to two dimensions, a photonic crystal can be formed that has been shown to be useful in the manipulation of light propagating within the crystal.41'42 GaN-based photonic crystal structures have been demonstrated43 and incorporated into InGaN/GaN LEDs.44'45 LEDs incorporating photonic crystal

384

Ng, Chowdhury

structures take advantage of the Purcell effect for increased internal efficiency. They also have better directionality that can be dictated by the design of the photonic crystal. These GaN photonic crystal structures were fabricated using a combination of electron beam lithography and plasma etching. The fabrication process for GaN photonic crystals still poses a major challenge. The feature size for photonic crystals has to be comparable to the wavelength of operation in the material. For GaN devices working in the blue and UV spectral region, this means that feature sizes around 100 nm would have to be achieved. These preliminary reports are encouraging and should pave the way for further research.

6.0kV

X1.700

10/um

WD 7.1 mm

Fig. 8. A ID periodic GaN/air grating. [From Ref. 34]

In view of the aforementioned applications, we have investigated fabricating both ID and 2D periodic structures using PSCE. Complete etching of the N-polar GaN in the sample shown in Fig. 5 (a) resulted in a periodic array of GaN stripes consisting of Ga-polar GaN. Figure 8 shows the SEM image of a 4 |i.m thick GaN sample with a periodic stripe pattern (17.2 u.m period) after the PSCE step. As seen in the images after etching, the Ga-polar surface remained unaffected while the N-polar

Ill-Nitrides Micro- and Nanostructures

385

regions were completely etched down to the sapphire surface. Furthermore, vertical sidewalls can be maintained for the large etch depths. It should be pointed out that the initial GaN stripes have to be oriented parallel to the [11 20] direction of GaN in order to have the smoothest sidewalls with {1 100} planes. This may lead to reduced scattering losses for structures such as GaN optical waveguides.

Fig. 9. SEM images of a sample with (a) an array of circular holes filled with N-polar GaN before etching and (b) an array of hexagonal holes after PSCE; the inset is a higher magnification view of one of the hexagonal holes.

386

Ng, Chowdhury

In the second case, instead of a stripe pattern, an array of circular holes was formed by photolithography and dry etching after the first MBE growth. The SEM image of the resulting sample after the second MBE growth can be seen in Fig. 9(a) where the N-polar regions having a diameter of 2.8 (J.m are embedded in a matrix of Ga-polar GaN. Even though circular holes were formed from the initial dry etch, PSCE revealed hexagonal-shaped holes in Fig. 9(b) reflecting the hexagonal symmetry of wurtzite GaN. The walls of the hexagonal holes are extremely smooth as shown in the higher magnification image in the inset of Fig. 9(b). The facets that are revealed by the etch are the {1 100} planes of GaN. In this sample, the etch depth is about 1.7 |J.m. Alternatively, the inverse pattern of circular Ga-polarity GaN regions in a matrix of N-polarity GaN has also been studied. The resulting SEM image of the sample after PSCE is shown in Fig. 10. Circular posts or pillars of Ga-polarity GaN were left after the N-polarity material was completely removed by the etch. It is not clear at this time why the posts do not have the hexagonal shape.

Fig. 10. An array of Ga-polar GaN posts after PSCE removal of the N-polar matrix.

Ill-Nitrides Micro- and Nanostructures

387

4. Applications of GaN Micro- and Nanostructures

4.1. Second Harmonic Generation Second harmonic generation by quasi-phase matching (QPM) has been demonstrated in a number of different materials such as LiNb0 3 , and GaAs. The principle of quasi-phase matching can be found in the seminal paper by Armstrong et al.46 In a bulk sample, the second harmonic wave and the fundamental wave have different phase velocities due to differences in the refractive indices at the two wavelengths, resulting in a phase mismatch. The phase mismatch leads to the second harmonic wave periodically interfering constructively and destructively as a function of position. In order to overcome this effect with QPM, the sign of the second order nonlinear susceptibility, Xm > is inverted periodically to reestablish the phase relationship between the two waves. Since GaN is a non-centrosymmetric crystal, it has a non-zero x(T>• Reported measured values of ^333 range from 7.4 to 10.7 pm/V 47'48 with one report quoting a value as high as 33 pm/V.49 Theoretical calculations show a value of 11.52 pm/V 50 which is closer to the lower range of measured values. The periodic inversion of the polarity of GaN is the key to achieving QPM. The periodically poled GaN (PPGaN)51 sample that was studied had a periodicity of 17.2 |j.m with each Ga- and N-polar stripe having a width of 8.6 11m. The dispersion of the refractive index of GaN was measured using a reference sample by spectroscopic ellipsometry. Fabrication of the sample followed the flowchart illustrated in Fig 4 stopping at step (d). The sample was completed after the second MBE growth step. The second harmonic power measured as a function of input wavelength is shown in Fig. 11. For this graph, the peak occurs at A, = 1658.6 nm which is close to our calculated wavelength of 1633 nm based on a periodicity of 17.2 Jim. It should be pointed out that an error in estimating (n2[o- nM) by as small as 7xl0"4 results in a shift of 25 nm for

388

Ng, Chowdhury

the required fundamental wavelength and such an error is beyond the resolution limit of the ellipsometry measurements. For an input fundamental of X - 1658.6 nm, we see 9 p.W at the second harmonic. The plot in Fig. 11 is, however, very broad and does not have the characteristic sine-squared function that is expected from periodically poled structures. The broadening of the spectrum comes from using an ultrafast laser with a pulse duration of 130 fs. As the center wavelength of the input laser is changed, the whole spectrum is shifted with its peak at the new wavelength position. When the wavelength of the laser is moved from X = 1658.6 nm to X = 1638.6 nm, we still see second harmonic power (about 2 |iW) since there is still a significant amount of power in the spectrum at X = 1658.6 nm. The polarization dependence of the second harmonic was also tested and it was found that no second harmonic was produced when the input polarization was aligned to the ordinary axis of the crystal. After accounting for Fresnel losses (~ 30 %) and the fact that only a portion of the spectrum at the fundamental is used for producing SHG (~ 10 %), the amount of useful fundamental power entering the PPGaN is about 7 mW.

1630 1640 1650 1660 1670 1680 1690 1700 1710

Peak laser wavelength (nm) Fig. 11. Second harmonic power as a function of the peak laser wavelength of a femtosecond pulsed source. Reprinted with permission from A. Chowdhury, H. M. Ng, M. Bhardwaj, and N. G. Weimann, Applied Physics Letters 83, 1077 (2003). Copyright 2003, American Institute of Physics.

Ill-Nitrides Micro- and Nanostructures

389

The normalized conversion efficiency T)nor [P2a/(P
100n

— *c*

3a>

10n

•o

o c <

1n

100p 300 400 500 600 700 800 900 10001100 Anode Voltage (V)

Fig. 12. Current-voltage measurements of the GaN pyramids with increasing sampleanode spacing, in steps of 50 (im, for curves going from left to right. [From Ref. 54]

390

Ng, Chowdhury

such as low electron affinity, high thermal conductivity, high melting point, and chemical robustness that make it a promising candidate for field emitters.52 Indeed, the GaN nanotip pyramids described in Section 3.2 have been tested for field emission. The sharp tips with diameters less than 20 nm are ideal for electric field enhancement and thus efficient field emission of electrons.53 The chemical stability of GaN also makes it a promising material for field emitters with long lifetimes. Measurements of the resulting electron field emission current from these nanotip pyramids were measured by placing a stainless steel anode close to the top of the pyramids and applying a high voltage between the anode and a probe tip on the sample surface.54 -26

-28

^

-30

^

-32

-34

-36 0.0010

0.0015

0.0020

1/V Fig. 13. Fowler-Nordheim plot showing the intercept (A) and slope (B). [From Ref. 54]

The current-voltage characteristics are shown in Fig. 12. The different curves going from left to right correspond to increasing anodesample spacing (in steps of 50 |J.m). According to theory, field emission current can be described by

Ill-Nitrides Micro- and Nanostructures

391

3/2 1.5x10^ xSx{j3Vf Kf Tn/. I(V) = ^ ' x e x p -6.8x10 i(/>

P*\

0

(2)

where V is the applied voltage, 5 is the emitting surface area, jB is the field enhancement factor, and c# is the electron affinity. To obtain Fowler-Nordheim plots, In (I/V2) is plotted against 1/V. The resulting plot is shown in Fig. 13, indicating that the emission current is indeed due to field emission rather than thermionic emission. From the slope of the curve, the field enhancement factor can be obtained. We normalized the field enhancement factor by multiplying /3 with the sample-anode spacing d. resulting value was found to be about 1500 assuming that the electron affinity of GaN is 3.5 eV.55 In order to obtain the turn-on electric field, we have taken the threshold voltage to be the applied voltage needed to produce an emission current of 10 nA. I

|

"I"'" |

I

I""™ I

|

'¥'

|

•

|

1000 ^ 900 a>

Slope = 1.6 V/nm

& 800 o > 1

700

CO

CD

£

600 500 - • _i

50

Ji

100

•i

i_

150

200

250

300

350

Relative anode-sample spacing (urn) Fig. 14. The threshold voltage as a function of the anode-sample spacing. The turn-on field is obtained from the slope. [From Ref. 54]

The threshold voltages plotted as a function of the relative anodesample spacing is shown in Fig. 14. From the slope of this curve, the

392

Ng, Chowdhury

average turn-on field was found to be 1.6 V/u.m. This value is comparable to those reported for carbon nanotube field emitters.56 Table II summarizes a number of reports found in the literature for field emission from GaN formed by various techniques. In order to make a relevant comparison, the turn-on fields are quoted for the same condition as described above, i.e. I = 10 nA. The range of turn-on field previously reported falls between 7 to 25 V/u,m. Therefore, these nanotip pyramids have the lowest value of turn-on field and highest normalized field enhancement factor for ungated GaN structures. Preliminary lifetime testing showed a 50% reduction of the emission current over a period of 3 hours. Further testing will be required to identify the degradation mechanism. Table II. Comparison of various ungated GaN field emitters reported in the literature. [From Ref. 54]

Sample preparation technique

Turn-on field @ 10 nA (V/(xm)

Normalized field enhancement factor

Selective area growth pyramids58

25

Selective area growth pyramids59

7

GaN surface roughened by hydrogen plasma treatment61

12.4

150

Polycrystalline GaN grown on Mo substrate61

12

200-610

GaN nanotips formed by reactive ion etching62

12

300

This work

1.6

1500

Given the promising results, we now discuss strategies by which the emission current can be further increased. These GaN pyramids were

Ill-Nitrides Micro- and Nanostructures

393

nominally undoped and therefore have a low background electron concentration. In order to increase the emission current, we can introduce Si donors to obtain a higher electron concentration. Furthermore, according to equation (2), an improvement in the emission current can be accomplished by a combination of increasing the emitting surface area, increasing the field enhancement factor or choosing a material with lower electron affinity. Assuming that most of the emission occurs at the tip of the pyramids, the surface area can be increased by having a higher density of tips. The field enhancement factor is related to geometrical factors and in this case, we have already obtained very sharp tips. Finally, AlxGai_xN can be used to replace GaN as the pyramid material since the electron affinity is smaller for larger values of x. It should also be pointed out that the use of selectively etched nanotip pyramids placed between regions of Ga-polar GaN with flat surfaces will simplify the fabrication of gated field emitters.57 Gate electrodes can be directly deposited on the flat GaN regions without the need for planarization.

5. Summary and Future Outlook In this chapter, an overview of the recent developments in the area of micro- and nanostructures of Ill-nitrides semiconductors was presented. The different methods of fabricating these structures were discussed. The ultimate choice whether to use the bottom-up or top-down approach will depend on the degree of control required for precise placement of these structures. A combination of the two methods may be the best way forward. There are still many challenges ahead for nanomaterials and nanophotonics. However, the addition of Ill-nitrides with their unique material properties to the scientist's/engineer's toolbox will definitely promise a future with devices exhibiting new functionality.

394

Ng, Chowdhury

Acknowledgments We would like to recognize valuable discussions and contributions of our current and former collaborators including W. Parz, N.G. Weimann, M. Bhardwaj (Bell Laboratories), J. Shaw (Naval Research Laboratory), R. Liu, A. Bell, and F.A. Ponce (Arizona State University). References 1. B. Daudin, F. Widmann, G. Feuillet, Y. Samson, M. Arlery, and J.L. Rouviere, "Stranski-Krastanov growth mode during the molecular beam epitaxy of highly strained GaN," Phys. Rev. B 56, R7069 (1997). 2. L. Goldstein, F. Glas, J.Y. Marzin, M.N. Charasse, and G. Le Roux, "Growth by molecular beam epitaxy and characterization of InAs/GaAs strained layer superlattices," Appl. Phys. Lett. 47, 1099 (1985). 3. D. Leonard, M. Krishnamurthy, CM. Reaves, S.P. Denbaars, and P.M. Petroff, "Direct formation of quantum-sized dots from uniform coherent islands of InGaAs on GaAs surfaces," Appl. Phys. Lett. 63, 3203 (1993). 4. M.K. Zundel, P Specht, K. Eberl, N.Y. Jin-Phillipp, and F. Phillipp, "Structural and optical properties of vertically aligned InP quantum dots," Appl. Phys. Lett. 71, 2972 (1997). 5. D.J. Eaglesham and M. Cerullo, "Dislocation-free Stranski-Krastanow growth of Ge on Si(100)," Phys. Rev. Lett. 64, 1943 (1990). 6. J.C. Johnson, H. Choi, K.P. Knutsen, R.D. Schaller, P. Yang, and R.J. Saykally, "Single gallium nitride nanowire lasers," Nature Materials, 1, 106 (2002). 7. R.S. Wagner, and W.C. Ellis, "Vapor-Liquid-Solid mechanism of single crystal growth," Appl. Phys. Lett. 4, 89 (1964). 8. M. He, P. Zhou, S.N. Mohammad, G.L. Harris, J.B. Halpern, et al. "Growth of GaN nanowires by direct reaction of Ga with NH3," J. Cryst. Growth 231, 357 (2001). 9. S.M. Zhou, Y.S. Feng, and L.D. Zhang, "A physical evaporation synthetic route to large-scale GaN nanowires and their dielectric properties," Chem. Phys. Lett. 369, 610 (2003). 10. M. Law, J. Goldberger, and P. Yang, "Semiconductor nanowires and nanotubes," Annu. Rev. Mater. Res. 34, 83 (2004). 11. A. Usui, H. Sunakawa, A. Sakai, and A.A. Yamaguchi, "Thick GaN epitaxial growth with low dislocation density by hydride vapor phase epitaxy," Jpn. J. Appl. Phys. 36, L899 (1997). 12. Y. Kato, S. Kitamura, K. Hiramatsu, and N. Sawaki, "Selective growth of GaN and AlxGa!.xN on GaN/sapphire substrates by metalorganic vapor phase epitaxy," J. Cryst. Growth 144, 133 (1994).

Ill-Nitrides Micro- and Nanostructures

395

13. K. Tachibana, T. Someya, S. Ishida, and Y. Arakawa, "Selective growth of InGaN quantum dot structures and their microphotoluminescence at room temperature," Appl. Phys. Lett. 76, 3212 (2000). 14. S. Bydnik, B.D. Little, Y.H. Cho, J. Krasinski, J.J. Song, W. Yang, and S.A. McPherson, "Room temperature laser action in laterally overgrown GaN pyramids on (111) silicon," MRS Internet J. Nitride Semicond. Res. 4S1, G6.48 (1999). 15. E. Calleja, M. A. Sanchez-Garcia, F. J. Sanchez, F. Calle, F. B. Naranjo, E. Munoz, U. Jahn, and K. Ploog, "Luminescence properties and defects in GaN nanocolumns grown by molecular beam epitaxy" Phys. Rev. B 62, 16826 (2000). 16. J. Ristic, E. Calleja, M.A. Sanchez-Garcia, J.M. Ulloa, J. Sanchez-Paramo, J.M. Calleja, U. Jahn, A. Trampert, and K.H. Ploog, "Characterization of GaN quantum discs embedded in AL,Gai_xN nanocolumns grown by molecular beam epitaxy," Phys. Rev. B 68, 125305 (2003). 17. H.M. Ng, R. Liu, and F.A. Ponce, "Self-assembled indium nitride nanocolumns grown by molecular beam epitaxy," Electrochem. Society Proceedings, 2004-06, 372 (2004). 18. C.H. Liang, L.C. Chen, J.S. Hwang, K.H. Chen, Y.T. Hung, and Y.F. Chen, "Selective-area growth of indium nitride nanowires on gold-patterned Si(100) substrates," Appl. Phys. Lett. 81, 22 (2002). 19. H. Parala, A. Devi, F. Hipler, E. Maile, A. Birkner, H.W. Becker, and R.A. Fischer, "Investigations on InN whiskers grown by chemical vapour deposition," J. Cryst. Growth 231, 68(2001). 20. B. Schwenzer, L. Loeffler, R. Seshadri, S. Keller, F.F. Lange, S.P. DenBaars, and U.K. Mishra, "Preparation of indium nitride micro- and nanostructures by ammonolysis of indium oxide," J. Mater. Chem. 14, 637 (2004). 21. M. Yoshizawa, A. Kikuchi, N. Fujita, K. Kushi, H. Sasamoto, and K. Kishino, "Self-organization of GaN/Al018Gao.82N multi-layer nano-columns on (0001) A1203 by RF molecular beam epitaxy for fabricating GaN quantum disks," J. Cryst. Growth. 189/190, 138(1998). 22. S. Yamaguchi, M. Kariya, S. Nitta, T. Takeuchi, C. Wetzel, H. Amano, and I. Akasaki, "Structural properties of InN on GaN grown by metalorganic vapor-phase epitaxy," J. Appl. Phys. 85, 7682 (1999). 23. B. Heying, R. Averbeck, L.F. Chen, E. Haus, H. Riechert, and J.S. Speck, "Control of GaN surface morphologies using plasma-assisted molecular beam epitaxy," J. Appl. Phys. 88,1855(2000). 24. J.B. MacChesney, P.M. Bridenbaugh, and P.B. O'Connor, "Thermal stability of indium nitride at elevated temperatures and nitrogen pressures," Mater. Res. Bull. 5, 783 (1970). 25. W.A. Harrison, Electronic Structure and Properties of Solids (Freeman, San Francisco, CA, 1980).

396

Ng, Chowdhury

26. J. Lee, H. Cho, D.C. Hays, C.R. Abernathy, S.J. Pearton, R.J. Shul, G.A. Vawter, and J. Han, "Dry etching of GaN and related materials: comparison of techniques," IEEE J. Selected Topics in Quantum Electronics 4, 557 (1998). 27. C.R. Eddy, Jr. "Etch processing of III-V nitrides," MRS Internet J. Nitride Semicond. Res. 4S1, G10.5 (1999). 28. T. Kozawa, T. Kachi, T. Ohwaki, Y. Taga, N. Koide, and M. Koike, "Dislocation etch pits in GaN epitaxial layers grown on sapphire substrate," J. Electrochem. Soc. 143, LI 7 (1996). 29. A. Shintani and S. Minagawa, "Optical properties of GaN light-emitting diodes," J. Electrochem. Soc. 123,1725 (1976). 30. C. Youtsey, I. Adesida, and G. Bulman, "Highly anisotropic photoenhanced wet etching of n-type GaN," Appl. Phys. Lett. 71, 2151 (1997). 31. C. Youtsey, L.T. Romano, and I. Adesida, "Gallium nitride whiskers formed by selective photoenhanced wet etching of dislocations," Appl. Phys. Lett. 73, 797 (1998). 32. J. Borton, C. Cai, M. Nathan, P. Chow, J. Van Hove, A. Wowchak, and H. Morkoc, "Bias-assisted photoelectrochemical etching of p-GaN at 300 K," Appl. Phys. Lett. 77, 1227 (2000). 33. J.L. Weyher, S. Muller, I. Grzegory, and S. Porowski, "Chemical polishing of bulk and epitaxial GaN," J. Cryst. Growth 182, 17 (1997). 34. H.M. Ng, W. Parz, N.G. Weimann, and A. Chowdhury, "Patterning GaN microstructures by polarity-selective chemical etching," Jpn. J. Appl. Phys. part 2 42, L1405 (2003). 35. H.M. Ng, N.G. Weimann, and A. Chowdhury, "GaN nanotip pyramids formed by anisotropic etching," J. Appl. Phys. 94, 650 (2003). 36. D. Li, M. Sumiya, S. Fuke, D. Yang, D. Que, Y. Suzuki, and Y. Fukuda, "Selective etching of GaN polar surface in potassium hydroxide solution studied by x-ray photoelectron spectroscopy," "J. Appl. Phys. 90,4219 (2001). 37. J.R. Mileham, S.J. Pearton, C.R. Abernathy, J.D. MacKenzie, R.J. Shul, and S.P. Kilcoyne, "Wet chemical etching of A1N," Appl. Phys. Lett. 67, 1119 (1995). 38. Y. Gao, M.D. Craven, J.S. Speck, S.P. DenBaars, and E.L. Hu, "Dislocation- and crystallographic-dependent photoelectrochemical wet etching of gallium nitride," Appl. Phys. Lett. 84, 3322 (2004). 39. T. Fujii, Y. Gao, R. Sharma, E.L. Hu, S.P. DenBaars, and S. Nakamura, "Increase in the extraction efficiency of GaN-based light-emitting diodes via surface roughening," Appl. Phys. Lett. 84, 855 (2004). 40. H.M. Ng, T.D. Moustakas, and S.N.G. Chu, "High reflectivity and broad bandwidth AIN/GaN distributed Bragg reflectors grown by molecular-beam epitaxy," Appl. Phys. Lett. 76, 2818 (2000). 41. E. Yablonovitch, "Inhibited spontaneous emission in solid-state physics and electronics," Phys. Rev. Lett. 58, 2059 (1987).

Ill-Nitrides Micro- and Nanostructures

397

42. S. John, "Strong localization of photons in certain disordered dielectric superlattices," Phys. Rev. Lett. 58, 2486 (1987). 43. T.N. Oder, J. Shakya, J.Y. Lin, and H.X. Jiang, "Ill-nitride photonic crystals," Appl. Phys. Lett. 83, 1231 (2003). 44. J.J. Wierer, M.R. Krames, J.E. Epler, N.F. Gardner, M.G. Craford, J.R. Wendt, J.A. Simmons, and M.M. Sigalas, "InGaN/GaN quantum-well heterostructure lightemitting diodes employing photonic crystal structures," Appl. Phys. Lett. 84, 3885 (2004). 45. T.N. Oder, K.H. Kim, J.Y. Lin, and H.X. Jiang, "Ill-nitride blue and ultraviolet photonic crystal light emitting diodes," Appl. Phys. Lett. 84, 466 (2004). 46. J.A. Armstrong, N. Bloembergen, J. Ducuing, and P.S. Pershan, "Interactions between light waves in a nonlinear dielectric," Phys. Rev. 127, 1918 (1962). 47. J. Miragliotta, D.K. Wickenden, T.J. Kistenmacher, and W.A. Bryden, "Linear- and nonlinear-optical properties of GaN thin films," J. Opt. Soc. Am. B 10, 1447 (1993). 48. N.A. Sanford, A.V. Davydov, D.V. Tsvetkov, A.V. Dmitriev, S. Keller, U.K. Mishra, S.P. DenBaars, S.S. Park, J.Y. Han, and R.J. Molnar, "Measurement of second order susceptibilities of GaN and AlGaN," J. Appl. Phys. 97, 053512 (2005). 49. H.Y. Zhang, X.H. He, Y.H. Shih, M. Schumann, Z.C. Feng, and R.A. Stall, "Study of nonlinear optical effects in GaN:Mg epitaxial film," Appl. Phys. Lett. 69, 2953 (1996). 50. J. Chen, Z.H. Levine, and J.W. Wilkins, "Calculated second-harmonic susceptibilities of BN, A1N, and GaN," Appl. Phys. Lett. 66, 1129 (1995). 51. A. Chowdhury, H.M. Ng, M. Bhardwaj, and N.G. Weimann, "Second-harmonic generation in periodically poled GaN," Appl. Phys. Lett. 83, 1077 (2003). 52. V.V. Zhirnov, G.J.Wojak, W.B. Choi, J.J. Cuomo, and J.J. Hren, "Wide band gap materials for field emission devices," J. Vac. Sci. Technol. A 15, 1733 (1997). 53. R. Gomer, "Field emission and field ionization," American Institute of Physics, New York, 1993. 54. H.M. Ng, J. Shaw, A. Chowdhury, and N.G. Weimann, "Electron Field Emission From GaN Nanotip Pyramids," Proceedings of the MRS 2003 Fall Meeting, vol. 798 (2004). 55. I. Wu and A. Kahn, "Investigation of the chemistry and electronic properties of metal/gallium nitride surfaces," J. Vac. Sci. Technol. B 16, 2218 (1998). 56. K. Matsumoto, S. Kinosita, Y. Gotoh, T. Uchiyama, S. Manalis, and C. Quate, "Ultralow biased field emitter using single-walled carbon nanotube directly grown onto silicon tip by thermal chemical vapor deposition," Appl. Phys. Lett. 78, 539 (2001). 57. T. Kozawa, T. Ohwaki, Y. Taga, and N. Sawaki, "Field emission study of gated GaN and Alo.1Gao.9N pyramidal field emitter arrays," Appl. Phys. Lett. 75, 3330 (1999).

398

Ng, Chowdhury

58. O. Nam, M.D. Bremser, B.L. Ward, R.J. Nemanich, and R.F. Davis, "Growth of GaN and Al02Gao.8N on patterned substrates via organometallic vapor phase epitaxy," Jpn. J. Appl. Phys. Part 2, 36, L532 (1997). 59. B.L. Ward, O.-H. Nam, J.D. Hartman, S.L. English, B.L. McCarson, R. Schlesser, Z. Sitar, R.F. Davis, and R.J. Nemanich, "Electron emission characteristics of GaN pyramid arrays grown via organometallic vapor phase epitaxy," J. Appl. Phys. 84, 5238 (1998). 60. T. Sugino, T. Hori, C. Kimura, and T. Yamamoto, "Field emission from GaN surfaces roughened by hydrogen plasma treatment," Appl. Phys. Lett. 78, 3229 (2001). 61. H. Tampo, T. Yamanaka, K. Yamada, K. Ohnishi, M. Hashimoto, and H. Asahi, "Field emission from polycrystalline GaN grown on Mo substrate," Jpn. J. Appl. Phys. Part 2, 41, L907 (2002). 62. Y. Terada, H. Yoshida, T. Urushido, H. Miyake, and K. Hiramatsu, "Field emission from GaN self-organized nanotips," Jpn. J. Appl. Phys. Part 2, 41, LI 194 (2002).

CHAPTER 12 NEW DEVELOPMENTS IN DILUTE NITRIDE SEMICONDUCTOR RESEARCH

W. Shan1, W. Walukiewicz1, K.M. Yu1, J. Wu2, J.W. Ager III1, and E.E. Haller1'3 1. Materials Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720 2. Department of Chemistry and Chemical Biology, Harvard University, Cambridge, MA 02138 3. Department of Materials Science and Engineering, University of California, Berkeley, CA 94720

Abstract Dilute nitrides are considered as highly mismatched semiconductor alloys of great technological importance for their applications in telecommunication devices and photovoltaic solar cells. This chapter reviews recent developments in the dilute nitride semiconductors research with the emphasis on the understanding of N-induced change in the electronic structure of the materials. The unusual physical properties associated with the dilute nitrides can be readily explained by the change in the conduction-band structure originating from a band anticrossing interaction between the extended conduction-band states and the localized states of nitrogen. 1. Introduction Dilute nitrides, especially GaAsj^N^ and Gai_>,InyAsi_jtN.c, have recently attracted considerable attention from both scientific and technological perspectives. The advances in thin-film deposition

399

400

Shan, Walukiewicz, Yu, Wu, Ager, Haller

technology have allowed these materials to be grown with everimproving crystalline quality, which in turn improves their optical properties and electronic performance. This enables the fundamental study of the unusual properties of these materials. These include a reduction of the fundamental band-gap energy,1,2 a significant increase in electron effective mass and a decrease in electron mobility.3"5 Furthermore, a new optical transition (E+) above the fundamental band gap energy has been observed.6'7 Most notable is the large observed band-gap bowing. Incorporation of only one percent nitrogen into GaAs induces a strikingly large reduction of 0.18 eV in the fundamental bandgap energy.8 The mechanism by which the addition of nitrogen changes the properties of these III-N-V materials appears to be fundamentally different from that in other III-V alloy systems such as A^Ga^As. This creates new opportunities for band-gap engineering and optoelectronic device-structure designs with the direct band gaps of these dilute-nitride alloys accessible to the near-IR, which is of great importance for telecommunications and solar power conversion applications. 2. Material Properties The novel material properties of dilute nitrides were first discovered in the early 1990's. In the quest to close the gap between the nitrides and arsenides thus to achieve the goal of fabricating light emitting devices covering the entire visible spectral region, Weyers and coworkers succeeded in growing GaNxAsi_x alloys using plasma assisted metalorganic chemical vapor deposition (MOCVD).1 To their surprise, they found that these alloys exhibit a considerable red shift in photoluminescence and absorption edge rather than the anticipated blue shift. Furthermore, application of simple interpolation between the properties of the end point materials using first or second order polynomials within the virtual crystal approximation (VCA), in which the random alloy potential is approximated by a periodic lattice of average atomic potential9"11 and has the trend of increasing band-gap energy with decreasing lattice constant, led to large and composition dependent bowing parameters12"15 beyond common experience.

New Developments in Dilute Nitride Semiconductor Research

401

The uncommon physical properties are the consequence of the extraordinary chemical characteristics of nitrogen compared to other group-V elements. These chemical characteristics, in turn, cause difficulties in incorporating nitrogen in III-V semiconductor crystals to form randomly mixed III-N-V nitride alloys. The conventional nonnitride III-V compound semiconductors do not easily crystallize in the wurtzite form, the crystal structure of GaN. It is therefore expected that GaNAs and the other analogous alloys will crystallize in the zinc-blende structure. There is a large miscibility gap that makes it difficult to prepare the alloys with large N fractions. At present, growth of III-N-V alloys is still considered challenging and bulk crystals have not been grown. To date most reports on III-N-V alloys involve thin films grown by molecular beam epitaxy (MBE) using RF plasma nitrogen radical beam source or metalorganic chemical vapor deposition (MOCVD) with dimethylhydrazine as nitrogen source.16 The nitrogen contents in such samples were usually determined using secondary ion mass spectrometry (SIMS) and indirectly from the change of the lattice constant measured with the (004) reflection in double-crystal x-ray diffraction. One of the major challenges for the growth of dilute nitrides is the completely different set of boundary conditions that affect the choice of the epitaxial growth technique. Both MBE and MOCVD have been used to grow dilute nitride materials. However, the issues governing choice are appreciably more complex and challenging than for GaAs-based and InP-based alloy systems that are extensively used in optoelectronic applications. The situation for growth of GaNAs and GalnNAs is entirely different compared to the InGaAsP materials system. First of all, in order to incorporate sufficient N, the growth has to occur at much lower growth temperatures and under metastable growth conditions within the miscibility gap region of the GalnNAs alloy. This is due to the different basic crystal structures of the constituent alloys and their regions of growth compatibility: GaN is a hexagonal (wurtzite) crystal grown at relatively high temperatures while GaAs is cubic (zinc-blende) grown at significantly lower temperatures, creating a miscibility gap in the

402

Shan, Walukiewicz, Yu, Wu, Ager, Haller

alloys. Hence, as either or both N and growth temperature increase, phase segregation occurs. Kondow and coworkers used an N plasma source added to a gas source MBE system.18 This growth technique provided material which showed the potential for GalnNAs. However, issues related to N incorporation and strong growth temperature composition sensitivity due to the low arsine cracking efficiency set limitations, particularly for large scale deposition using this approach. Solid-source MBE with an atomic N plasma source has proven to be the most versatile system to allow growth at the lowest temperatures and over the largest range of N and In compositions.17 The single most critical parameter controlling growth is the growth temperature. When the growth temperature exceeds a critical value, MBE growth begins to change from 2D, layer-by-layer growth to 3D island growth with microphase segregation.17 There is a N composition dependence on suitable growth temperature, however, 420°C < T < 450°C maintains 2D epitaxial growth over the greatest range of N compositions. The V/III supply ratio also has an impact on growth, but much less so than temperature. Because growth must occur at much lower temperatures, MOCVD growth is far more challenging than MBE growth. Compared to MOCVD growth of N-based wide-band-gap systems, which use ammonia as the N source, the growth temperature for GalnNAs is too low to achieve reasonable cracking of either ammonia or arsine. Need to use new sources with complex precursor reactions and highly nonlinear incorporation ratios greatly complicate the growth compared to conventional III-V materials systems. The higher growth temperatures limit the N incorporation where micro-phase segregation begins and makes it extremely challenging to reach the N compositions needed to achieve 0.8-1.0 eV band gap.19 A new method for synthesizing dilute nitrides was developed during recent years. Nitrogen implantation followed by rapid thermal annealing (RTA) was found to be a practical and convenient method for the formation of diluted III-N-V alloys.20'21 The fundamental band-gap energies for the ion beam synthesized thin films of GaNxAsi_x, InNxPi-x

New Developments in Dilute Nitride Semiconductor Research

403

and AlyGai_yNxAsi_x obtained by N+ implantation into GaAs, InP and AlyGai.yAs were found to decrease with increasing N implantation dose in a manner similar to that observed in epitaxially grown thin films. In GaNxAsi_x the highest value of x achieved using N+-implantation and conventional RTA technique was 0.006 corresponding to an N activation efficiency of -15%. In the course of optimizing the annealing conditions in these studies, it was found that, in GaNAs formed in this way, the substitutional NAs is thermally unstable at temperatures higher than 850°C and will precipitate to form N2 filled voids.22 N(4%) impl. GaAs; PLM+950°C 10s AE=240meV 4% N

a:

1.2

1.3

1.4

energy (eV) Fig. 1. PR spectra measured from a series of samples implanted with increasing amounts of N (ximp) and processed by PLM at an energy fluence of 0.34J/cm2 and subsequent RTA at950°Cforl0sec.

More recently, it has been shown that pulsed laser melting (PLM) of N-implanted III-Vs dramatically improves the incorporation of N on the

404

Shan, Walukiewicz, Yu, Wu, Ager, Haller

group-V element site. ' In PLM, the near surface absorption of a single intense laser pulse instantaneously melts the implant-damaged or amorphized layer. This is followed immediately by rapid epitaxial regrowth from the liquid. Epitaxy is seeded at the solid-liquid interface by the crystalline bulk in a manner very similar to liquid phase epitaxy (LPE) but with the whole process occurring on a much shorter time scale, typically between 10"8-10"6 second.25'26 Figure 1 shows a series of photoreflectance (PR) spectra from GaAs implanted with increasing amounts of N processed by PLM with an energy fluence of 0.34 J/cm2 and subsequently by RTA at 950°C for ten seconds. Such PLM-RTA post-implantation treatments appear to represent the "optimum" process conditions found to date and the samples so formed have clear, sharp optical transitions. The amount of N incorporated in the As sublattice ("active" N) for the GaNxAsi_x layers formed by this method can be estimated using the BAC model and is -40-60% of the implanted value. This is over five times higher than the activation efficiency observed in samples processed by RTA only.21 Such a drastic improvement can be attributed to the extremely short melt duration (~2xl 0"7 s) and re-growth process that promotes N substitution in the As site and inhibits the formation of nitrogen voids.24 In addition to the enhanced N incorporation, the dilute nitride layers synthesized by N+-implantation followed by PLM-RTA were also found to be thermally stable up to annealing temperature > 950°C. This improved sample synthesis technique provides a convenient and reliable method, in addition to conventional epitaxial growth techniques,2'4'12 for preparing large variety of dilute nitride samples.

3. Device Applications

3.1. Long-wavelength Laser Diodes for Telecommunications The rapid growth of the internet and data transmission in recent years has driven the bandwidth of optical fiber networks, particularly in the areas

New Developments in Dilute Nitride Semiconductor Research

405

of metro area networks (MAN) and local area networks (LAN), to be continuously expanded to meet the ever increased demand. Low cost, single mode vertical cavity surface-emitting lasers (VCSELs) operating in longer wavelengths and at room temperature are an essential element for data communications in a gigabit MAN or LAN architecture because the most used GaAs-based 850-nm VCSELs for data communications suffers a severe drop in transmission distance as the bit-rate increases. Although InGaAsP/InP has long been the materials system for distributed feedback (DFB) edge-emitting lasers that have been the sources for long-haul, 1.55 jum optical-fiber backbone networks over the years, the inherent material problem of insufficient refractive index contrast, makes it impossible to meet the requirements for distributed Bragg reflector quarter-wave VCSEL mirrors, particularly in 1.3 /urn. It is also not feasible to use such expensive DFB lasers for data communications in MAN and LAN that would require millions of them. In addition, use of Raman amplifiers in the dense-wavelength division-multiplexing architectures within the S- and L-bands require high-power pump lasers similar to 0.98 jum strained InGaAs/GaAs QWs lasers, but at longer wavelengths (1.2 - 1.5 //m). The well-known materials limitations of the InGaAsP system make it nearly impossible for applications in long wavelength high-power pumping lasers. The small heterojunction conduction-band offset between InP and InGaAsP (-40% AEg) limits electron confinement in the QWs, resulting in a much lower characteristic temperature T0 compared to the InGaAs/GaAs materials used for EDFA pumps. Since the bandgap energy decreases for decreasing lattice constant in mixed group-V dilute nitrides, they can dramatically expand the range of applications of III-V alloy semiconductors and significantly increase freedom in designing semiconductor devices. Therefore, there are possibilities in which novel devices can be created or the performance of current devices can be drastically improved. It is Kondow and his coworkers who first proposed GalnNAs as an alternative active-region material for semiconductor laser diodes operating in the 1.3 //m and 1.55

406

Shan, Walukiewicz, Yu, Wu, Ager, Haller

jum regions, the telecommunications windows of optical fibers. It is possible to make devices with significantly superior performance than the ones based on the InGaAsP/InP materials system due to the unusual physical properties of the dilute nitride alloy semiconductors: Adding In to GaAs increases the lattice constant of InGaAs alloys, whereas adding N to GaAs decreases the lattice constant of GaNAs alloys. Therefore GalnNAs can be grown lattice-matched or pseudomorphically nearlymatched on a GaAs substrate. Both In and N have the effect of reducing the band-gap energy that makes GalnNAs suitable for long-wavelength laser diodes (1.3-1.55 jum and longer wavelengths). Furthermore, as will be discussed in the next section, incorporating of N into GaAs has a negligible effect on the valence band so that almost the entire change in the band gap between GaNAs and GaAs is accommodated by the conduction-band offset alone. All these unusual properties make GalnNAs one of the most attractive new materials for both VCSELs and high-power edge-emitting laser applications. By combining GalnNAs with GaAs or other wide-gap materials that can be grown on a GaAs substrate, a type-I band lineup is achieved and, thus, very deep quantum wells can be fabricated, especially in the conduction band.16 Since the electron overflow from the wells to the barrier layers at high temperatures can be suppressed, GalnNAs is highly attractive for overcoming the poor temperature characteristics of conventional InGaAsP/InP long-wavelength laser diodes. In the case of VCSELs, GalnNAs can utilize structures almost identical to 850 nm VCSELs that are now in large-scale production. In order to greatly expand broadband amplifier used in MANs, the key features that must be achieved are sufficient power and low cost. At this point it is clearly far from obvious that this new material will reach the leading position for the development of a broad range of VCSELs and edge-emitting lasers that will be the foundation of lower cost fiber optical networks for telecommunication.

New Developments in Dilute Nitride Semiconductor Research

407

3.2. Solar Cells for Photovoltaic Solar Power Conversion High-efficiency GalnP/GaAs/Ge monolithic series-connected threejunction solar cells are currently in production for space applications and are the leading candidates for terrestrial concentrator applications.27"29 However, the next generation of four-junction devices with considerably higher efficiencies require a set of III-V materials with a band gap energy lower than that of GaAs but higher than that of Ge, preferably lattice matched to GaAs to minimize strain-induced defects that severely degrade solar cell performance.30 The discovery by Weyers et al. of the anomalously large band-gap reduction in GaNAs1 and the introduction of GalnNAs as a 1 eV semiconductor, lattice matched to GaAs for laser applications by Kondow et al.l& has quickly led to the realization that GalnNAs could be also a suitable material for next-generation solar cell designs. A solar cell must convert photons to electron-hole pairs and separate them into electrons and holes. The voltage of a solar cell is limited by the lowest band-gap energy of the semiconductor. The absorption coefficient a(£) of a given semiconductor material is a measure of how strongly a photon with a particular wavelength will interact with the semiconductor, and is inversely proportional to the distance within which the photon will most likely be absorbed. Both the thickness of a solar cell and its carrier collection length (i.e., its combined diffusion length and depletion width) must be greater than l/a(E) in order to collect a significant fraction of photons of a particular wavelength. Only strongly absorbing direct-gap semiconductors are useful in thin-film solar cells. In a typical solar cell, electrons and holes are separated by the electric field generated by the diffusion potential of a p-n junction. In high-quality semiconductors, these carriers can also diffuse through a field-free region to the p-n junction, allowing collection from relatively thick layers of semiconductor. The currently most advanced and complicated 1.8 eV GaInP/1.4 eV GaAs/1.0 eV GaInNAs/0.7 eV Ge lattice matched four-junction structure has the potential for extremely high efficiencies.30 The novel component of this structure is the 1 eV GalnNAs third junction. Single-junction 1.0 eV GalnNAs cells lattice matched to GaAs have been studied most

408

Shan, Walukiewicz, Yu, Wu, Ager, Haller

extensively. For a 1.0 eV solar cell to be current-matched in the fourjunction devices described above, nearly unity internal quantum efficiencies (QE) are required in the energy range 1.0-1.4 eV. Unfortunately, internal QEs of actual devices have reached only about 70% because of poor minority-carrier diffusion lengths. These 70% QEs were achieved by using non-standard designs to minimize the effects of poor diffusion lengths, such as designs using p-i-n type structures with regions of low carrier concentration31,33 or designs using annealed thick n-type emitter layers.34 In contrast, p-n junction designs using a thin highly doped (~lxl0 18 cm"3) emitter and a thicker moderately doped (~l xlO17 cm"3) base, a design which works well for GaAs cells, typically result in <25% internal QE because of the low GalnNAs diffusion length.31 The cause for short diffusion lengths of minority carriers is a combined effect of low mobilities and short lifetimes that are still not fully understood. Further improvement on the material quality is inevitably needed in order to achieve nearly unity internal quantum efficiency. 4. Origin of Band-gap Reduction in Dilute Nitrides 4.1. Large Band-gap Bowing and Early Impurity Models The unexpectedly strong effect of the introduction of N in III-V compound semiconductors on the fundamental band gap is related to the fact that replacement of atoms such as As with the much smaller and more electronegative N atom leads to a large, local perturbation of the crystal lattice potential. The electronegativity of N is 3.00, while that of P, As, or Sb is in a range from 2.2 to 1.8.38 Although it has been known for a long time that small quantities of nitrogen form deep, localized impurity states in GaAs and GaP, the unexpected discovery of a considerable red shift in photoluminescence and absorption edge rather than the expected blue shift, together with a giant bowing parameter, in the GaNAs alloy was a surprise to many. An early theoretical study predicted a bowing parameter C-25 eV for the direct energy gap (Eg=A+Bx+Cx2) of GaNAs.39 C=18 eV was

New Developments in Dilute Nitride Semiconductor Research

409

obtained from PL measurements of GaNAs with N concentration less than 1.5%.13 The same result was inferred from studies of the GaNAs near the two binary limits, with the decrease in the energy gap being linear for N fractions as high as 3%. 40 A quadratic form with C=l 1 eV fit the results of an ellipsometry study for x=3.3% fairly well.41 Other studies of dilute GaNAs indicated bowing parameters as large as 22 eV. 4 M 4 A series of first-principles calculations examined various aspects of the band structure for GaNAs, including ordering effects.45'50 Those studies found that the band-edge wave functions in GaNAs tend to be localized impurity-like states, with the conduction-band wave function strongly localized on the As sublattice and the valence-band wave function on the N sublattice. In order to avoid the early semimetallic transition predicted by using C=20 eV, Uesugi et al. proposed a more gradual reduction of the bowing parameter with composition, and attributed the discrepancy to different strain conditions in the different studies.8 4.2. Band Anticrossing It is well known that an isolated N atom introduces a localized state with energy level EN in conventional III-V materials. In most cases, this level is located very close to the conduction band edge. It lies at about 0.25 eV above the conduction band edge in GaAs and less than 0.1 eV below the conduction band edge in GaP. The existence of such states has been predicted by theoretical calculations within the tight binding approximation framework,51 and confirmed by experimental measurements under hydrostatic pressure.52'53 As expected for a localized state the energy level of N shows pressure dependence much weaker than that of the conduction band edge of GaAs. The level was also found to move into the band gap when GaAs is alloyed with AlAs.54 The highly localized nature of the N states suggests that there is only weak hybridization between the orbits of N atoms and the extended states, E^k), of the semiconductor matrix. The electronic band structure of the host crystal is not significantly affected by these low nitrogen concentrations.

410

Shan, Walukiewicz, Yu, Wu, Ager, Haller

However, alloying a few atomic percentage of nitrogen into III-V compounds drastically modifies the electronic band structure. A pressure dependent study6 found that all the Ga^InxNyAsi.y samples exhibit a much weaker dependence of the band gap energy at low pressures, and a tendency of the energy gap to saturate at high pressures. The saturation is clearly visible in the samples with lower N contents. The pronounced change in the pressure dependence of the energy gap can only be understood in terms of a pressure-induced transformation of the nature of the lowest conduction band states, namely from extended to highly localized. The gradual nature of the transformation cannot be associated with a pressure-induced crossover of non-interacting T and X conduction-band valleys but rather suggests that it is a manifestation of an anticrossing behavior of two strongly interacting energy levels with distinctly different pressure dependencies. In order to explain the observed pressure dependence, a simple band anticrossing (BAC) model of two interacting energy levels; one associated with extended states of the GalnAs matrix and the other with the localized N states, was proposed to explain the large changes in the band structure of the resulting dilute III-N-V nitrides. By assuming that N atoms are randomly distributed over the group V sites and are only weakly coupled to the extended states of the conduction band of the host semiconductor matrix, the dispersion relations of the two interacting bands takes the form6

E±{k) = \{EM(k)

+

EN)±J{EM(k)-Ej+4Vm2\

(1)

Where EiJJk) and EN are the energies of the unperturbed conduction band and of the localized N state relative to the top of the valence band, respectively. The matrix element VMN=CMNxm, where CMN is a constant describing the coupling between localized states and the extended conduction-band states and x is the alloy composition. A very important inference of this model is that it predicts that the interaction of the conduction-band edge with the dispersionless N level results in a splitting of the conduction band into two highly nonparabolic subbands, E_(k) and E+(k). The energy positions of the subband edges E_ and E+

New Developments in Dilute Nitride Semiconductor Research

411

given by Eq.(l) depend on alloy concentration x and the coupling parameter CMN, as well as the location of EN with respect to the conduction band edge EM. Although the density-functional calculation of Jones et al. also predicted a reduced pressure dependence,55 the formation of E+(k) subband is a definitive result of band anticrossing.

/ // // // y 'rE

P \ —. \

.

/

^---—

E_

CD

c

LU

(«)// -20

-10

\\lh so\ \ 10

20 -1i

Wavevector (10 cm )

-20

Wavevector (10 6 cm"1)

Fig. 2. Illustration of the effects of band anticrossing on the conduction band structure in the vicinity of T-point minimum, (a) The N induced localized state resonant with the conduction band; (b) The localized state located below the conduction band. The solid lines are the restructured E_ and E+ subbands resulting from the band anticrossing interaction between the localized states (dash-dotted line) and the extended states of the conduction band (broken line).

Figure 2 shows schematic examples of the calculated band structure based on the BAC model. The interaction between the localized isoelectronic states and the extended conduction-band states has a pronounced effect on the dispersion relation of the two conduction subbands E_ and E+. The effect of the interaction is most pronounced for the states located close to EN.56 If the localized state is located within the conduction band of the matrix, as depicted in Fig.2(a), the conductionband states at the E_ edge retain mostly the extended EM-like character

412

Shan, Walukiewicz, Yu, Wu, Ager, Haller

and those at the E+ edge have a more localized and En-like character. The lower conduction subband narrows drastically as the EN level moves below the bottom of the conduction band. The narrowing of the band leads to a highly nonparabolic dispersion relationship and to a large enhancement of the effective mass and the density of states If the localized state is located below the conduction-band edge then as illustrated in Fig.2(b), the E_ subband states become more localized wheras E+ subband states acquire more of the extended state character. The conduction band dispersion relations predicted by BAC model have been recently confirmed in an elegant magnetotunneling experiment.57

1.4

—I—1— i

<

|

i

i

i

i

|

•

•

•

i

i—i

i

i

|

i

i

i

i

GaAs N @295K

\

1- X

-

X 1 1

1.5

II 1

1

> 1.3 a> a §> 1.2 is.

• 0 A •

0.9

•

K A

Uesugi, et. al. Keyes, et. al. Malikova, et. al. Bhat, et. al. - BAC theory J—i—i—i—I_I

0.01

i—i—

0.02

A A

1 1

>» • D) 0> c 1.1 LU

\

.

,

,

I

J

• ^\^

1—1—1—1—1—1

0.03

0.04

1

1

0.05

Nitrogen fraction, x Fig. 3. Comparison between the experimentally observed and calculated band-gap reduction of GaNjAsi.* as a function of N concentration. The calculations are based on the BAC model with CUN=2.1 eV, £M=1.42 eV, and £^=1.65 eV. Recent theoretical considerations using the tight binding approximation have provided additional refinements for the BAC model.58"60 The BAC model can be extended to treat ten bands (spin-

New Developments in Dilute Nitride Semiconductor Research

413

doubled conduction, valence, and nitrogen impurity bands) by modifying the 8-band k-p theory to include two extra spin-degenerate nitrogen states to describe the electronic band structure of GaNAs/GaAs and related heterostructures.5862 It was also argued that the electronic structure of GaNAs alloys is determined by interactions between nitrogen, X, L and T states63'64 This provided more parameters to afford greater flexibility in fitting the experimental data. Although the BAC model does not consider anything more complicated than an interaction of randomly distributed localized nitrogen states with the extended states of the conduction band it properly describes all the main characteristics of the electronic structure of dilute nitrides. It ignores possible complex behavior of nitrogen in the semiconductor matrix, e.g., the formation of nitrogen pairs and clusters. Detailed studies on the nearest-neighbor environment of the substitutional N atoms in GalnNAs have shown that the fundamental band-gap energy in quaternary dilute nitride alloys is fairly sensitive to the local environmental conditions especially in the case of quantum well structures.65'66 A more complicated modeling of the electronic structure based on pseudopotentials67'68 requires a substantial computational effort. The numerical results are difficult to use. This is why simple band anticrossing has been a method of choice in the design of GalnNAs based optoelectronic devices69 The recent work by Lindsay et al. points out that the most amply verified prediction of the BAC model, the dependence of the band gap on the N content, is be unaffected by the multiplicity of higher-lying states.70

4.3. E_ and E+ Transitions The downward shift of the E_ transition relative to the top of the valence band represents the fundamental band-gap reduction in dilute nitrides. It has a nonlinear dependence on the N concentration. Figure 3 shows that BAC model with a single fitting parameter CMN provides a very good agreement with the experimental data reported by a number of research groups.8'71"73 The coupling parameter CMN=2J eV was obtained from the fitting of Eq. (1) to the experimentally determined pressure dependence of the band gap of GaNxAsi_x.6'74

414

Shan, Walukiewicz, Yu, Wu, Ager, Haller

1/1 DC Q_

1.2

1.3

1.4

1.5

2.0

2.1

2.2

P h o t o n E n e r g y (eV) Fig. 4. PR spectra of GaNxAsi.x samples with different N concentrations.

The conduction band splitting into two nonparabolic subbands with energy minima at E_ and E+ has been unambiguously observed in GaNxAsi_x and Gai„yInyNxAsi_x by various groups using a variety of methods.6,7,75'76 Shown in Fig. 4 are PR spectra recorded with MOCVDgrown GaNxAsi_x samples. For N containing samples, in addition to the PR spectral features related to the transition across the fundamental band gap (E_ transition) and the transition from the top of the spin-orbit splitoff valence band to the bottom of the conduction band (E_+Ao transition) as displayed on the PR curve of GaAs, an extra feature (E+) appears at higher energies in the PR spectra. With increasing N concentration, the E_ and E_+Ao transitions shift to lower energy and the E+ transition moves in the opposite direction toward higher energy.74 The predictions of the BAC model were further verified by the measurement of the E_ and E+ transitions under applied pressure.6'74 The effect of hydrostatic pressure on the optical transitions associated with the E_ and E+ subband edges in a GaNo.015Aso.985 sample and a

New Developments in Dilute Nitride Semiconductor Research

415

Gao.95hio.05No.012Aso.988 sample is shown in Fig.5. Application of hydrostatic pressure shifts the bottom of the conduction band above the localized N level, gradually changing the character of the £_-subband edge from extended / ^ l i k e to localized Zsjv-like, and the character of the £+-subband edge from the localized-like to extended-like. Such a transformation is exactly what the BAC model predicted, schematically depicted in Fig.2.

Pressure (kbar) Fig. 5. Effects of pressure on the optical transitions associated with the E_ and E+ transitions in GaNaol5As0.985 and Gao.95Ino.05No.012Aso.9g8-

416

Shan, Walukiewicz, Yu, Wu, Ager, Haller

The dispersion relations of the E_(k) and E+(k) conduction subbands shown in Fig.2(a) represent the case at low pressures where the bottom of the conduction band of the host matrix (EM) is below the EN level. Fig.2(b) represents the case at high pressures where EM is shifted to above the EN level. The anti-crossing behavior of two strongly interacting energy levels with distinctly different pressure dependencies is unmistakably observed. The E_ transition has a strong dependence at low pressures and gradually saturates at high pressures, whereas the E+ transition has a weak dependence upon pressure at low pressures and displays a much stronger dependence at high pressures. The solid lines through the experimental data in the figure are the results of calculations using Eq.(l). The best fits to the data yield the energy of the nitrogen state, EN= Ev+ 1.65 eV for both samples at atmospheric pressure, and it is independent of In concentration.6'74 These results prove that the effects of alloying with In on the band gap can be separated from the shifts produced by the interaction with N states, allowing for an independent determination of EM from a given In concentration in Gai_yInyNxAs!_x alloys. 4.4. Enhancement in Maximum Free Electron Concentration As has been discussed above, the BAC model not only explains the band gap reduction in dilute III-N-V nitrides but it also predicts that the Ninduced modifications of the conduction band will have profound effects on the transport properties of those material systems.57 In particular, the downward shift of the conduction band edge and the enhancement of the DOS effective mass will lead to much enhanced maximum free electron concentration n^^. The electron effective mass as a function of energy can be calculated using the standard definition of the density of states effective mass, 2

m (k F ) = /z

dE_{k)/dk

= mn 1 + fc=k.

C

x

(EN-E_(kF)f

(2)

New Developments in Dilute Nitride Semiconductor Research

All

Where m0* is the effective mass of the semiconductor matrix and E is the energy in the lower or upper subband measured from the valence band edge. It is seen from Eq.(2) that the effective mass diverges for the electron energy approaching EN. This is a result of the increasing contribution of the localized N states to the electron states in the lower and upper subbands. The maximum achievable electron and/or hole concentration is of great importance in semiconductor devices engineering. A universal rule that governs the maximum free carrier concentration achievable by doping has been developed based on an amphoteric native defect model and demonstrated to be valid for a wide variety of semiconductor materials.77'78 In that model, the type and concentrations of the defects compensating intentionally introduced dopants depends on the location of the Fermi level relative to a material-independent common energy reference called Fermi level stabilization energy EFS- GaAs is predicted, for example, to exhibit limitations on the maximum free electron concentration. Indeed, the maximum electron concentration n^^ in GaAs achievable under equilibrium conditions has been experimentally confirmed to be limited to about 1018-1019cm"3.79 Shown in Fig. 6 are the free electron concentrations in Se doped MOCVD-grown Gai_3xIn3xNxAsi_x films with x=0 to 0.033 measured by Hall effect and the electrochemical capacitance-voltage (ECV) technique.80 Since the Se atomic concentrations in these films are at least an order of magnitude higher than the free electron concentration (in the range of 2~7xl020cm~3), the measured free electron concentration should be regarded as the maximum achievable free electron concentration, n,^. The result shown in Fig. 6 indicates that the n^ increases strongly with the N concentration. A maximum value of 7xl0 19 cm"3 was observed for x=0.033. This value is ~ 20 times that found in a GaAs film (3.5xl0 18 cm"3) grown under the same conditions. The much-enhanced nrmx in Gai_ 3xIn3xNxAsi.x films can be explained by considering the conduction band modifications by N-induced anticrossing interaction. Since the maximum free electron concentration is determined by the Fermi energy with respect to £PS79 and because the position of the valence band in

418

Shan, Walukiewicz, Yu, Wu, Ager, Haller

GalnNAs is independent of N concentration, the downward shift of the conduction band edge toward .Eps and the enhancement of the DOS effective mass in GalnNAs leads to a much larger concentration of uncompensated, electrically active donors for the same location of the Fermi energy relative to EFS- The calculated n^ as a function of x for Gai_3XIn3xNxAsi_x only considering the downward shift of the conduction band caused by the band anticrossing, as well as that also including the increase in the effective mass can be obtained using81

Iff

21

:

:

Se concentration " i 11II

.2 SB

J '

CD O

c. o o

o -#—» o

/

10'

— '

"

s<1 * * I *--' ,.---" Band edge shift only

*

i

band shift+eff, mass__

E

-

i t

I * 1 *

0) •

10'

•

•

0.01

•

i

i

i

i

i

i

•

•

•

•

i

•

0.02 0.03 0.04 Nitrogen fraction, x

•

0.05

Fig. 6. Comparison of the measured maximum electron concentration with the calculated values as a function of N fraction in GaI.3xIn3xNxAs1.x. Two different cases of the calculated n m are shown: one includes effects of downward shift of the conduction band only (dashed curve) and the other includes both the band shift and the enhancement of the density of states (solid curve). The calculated n M for samples without N (i.e. when only the effects from the band-gap lowering produced by In incorporation are considered) are also shown (dotted curve) for comparison. The shaded area indicates the range of Se concentration in these samples.

New Developments in Dilute Nitride Semiconductor Research

n(EF)=j-

p{E)dE + exp[(£-£j/£Br]'

419

(3)

where p(E) is the perturbed density of states. The results are shown in Fig. 6. Comparison of the experimental data with the calculation shows that in order to account for the large enhancement of the doping limits in III-N-V alloys both effects, the downward shift of the conduction band and the increase of the effective mass have to be taken into account.

10 20

.

> -1—

1019: 9 is * •'

>

-I—

|

F

|

1—7

F

1

1

1

F

1 —F—

I

F

.

\ band edge+eff. mass \ : -band edge only ;

°ees„

E c 2

1

w

1!

c CD o c o o 10' ^

S(cal.) —— SI-GaAs --9-GaN As 0.008

10

0.0

0.5

®

0.992

1.0

1.5

2.0

depth (urn) Fig. 7. Ionized net donor concentration profiles for the GaNxAs[.x films and the SI-GaAs standard measured by the electrochemical capacitance-voltage (ECV) technique. The short-dashed curve is the calculated distribution of implanted S atoms. The dashed horizontal lines indicate the theoretical free electron concentrations in Ga!_xNxAs by considering only the effects of the downward shift of the conduction band (band edge only) and both the effects of band-gap reduction and density of states effective mass enhancement (band edge + effective mass).

420

Shan, Walukiewicz, Yu, Wu, Ager, Haller

While Se doped Gai.3xIn3xNxAsi_x alloys grown by MOCVD have shown enhanced n m x in accordance with the BAC model, similar behavior is also observed in S+-implanted GaNjAsj.x thin films.82 Figure 7 displays the carrier concentration profiles measured by the ECV technique for S implanted GaNxAsi_x (x~0.008) and SI-GaAs samples after RTA. A striking difference in the free electron concentration n measured in the SI-GaAs and the GaNxAsi_x samples is observed. In the S+-implanted SI-GaAs sample, n ~2.5xl017cm"3 was measured in the bulk of the implanted layer, with a higher n ~5xl017cm"3 towards the end of the implantation profile. The theoretical n^x in GaNxAsi_x due to the N-induced conduction band modification within the framework of the BAC model and the amphoteric native defect model is -lxlO 1 9 cm"3 for the GaNo.oo8Aso.992 sample. This value is in a reasonably good agreement with the measured concentration of 6xl0 18 cm"3 shown in Fig. 7. Attempts to form n-type GaNxAsi_x thin films with high electron concentration were also made by co-implantation of N and a dopant element in GaAs.83 Figure 8 shows a comparison of the ECV determined free electron concentration profiles for GaAs samples implanted with S alone and co-implanted with S and N (S+N) after RTA at 945°C for 10s. The calculated as-implanted S and N atomic distributions are also shown in the figure. The most prominent difference in the electron concentration profiles between the S only and (S+N) samples is the much enhanced electron concentration in the (S+N) sample in a narrow region (-500A) near the surface. The region with lower electron concentration at -0.1-0.2 |J,m below the surface coincides with a region with excess As due to the implantation process that makes the substitution of S atoms into the As sites more difficult.84 In addition, larger concentrations of the compensating VGa acceptors are also expected in the As-rich region. A reduced availability of group V sites and an increased VGa concentration in the region lead to the minimum in the electron concentration. The effect is exacerbated in the (S+N) sample where both S and N compete for the same group V element sites.

New Developments in Dilute Nitride Semiconductor Research

0.0

0.2

0.4 depth (Ltm)

0.6

421

0.8

Fig. 8. The ECV measured net donor concentration profiles for the GaAs samples implanted with S alone and S+N after RTA at 945°C for 10s. The calculated atomic profiles for both the implanted S and N are also shown.

Considering both band gap reduction and large increase in the electron effective mass, the high ntmx in the near-surface region of the (S+N) sample (~1.5xl019cm~3) implies that the N content in this thin near-surface diluted nitride layer is x=0.0032. This value is in good agreement with the calculated N concentration in the surface region (x«0.003-0.01). With this N content the conduction band edge is shifted downward by 77 meV and the conduction band effective mass at the Fermi energy becomes about three times larger than that of GaAs.8

422

Shan, Walukiewicz, Yu, Wu, Ager, Haller

5. Concluding Remarks: From Dilute III-N-V Nitrides to Dilute IIO-VI Oxides Although this chapter has been devoted to III-N-V alloys, we would like to conclude it by pointing out that the experimental and theoretical methods developed for the dilute nitrides should be generally applicable to other semiconductor alloys with isoelectronic substitution of elements with highly mismatched electronegativities. It has been demonstrated that group II-O-VI alloys in which highly electronegative O partially replaces the group VI element show a modification of the electronic similar to that found in III-N-V alloys. For example, a dramatic O-induced reduction of the bandgap has been reported in Cdi.yMnyOxTe^ and ZnOxSei_x.85'86 Thus, partial replacement of group-VI anions with more electronegative O atoms in II-VI compounds does have the effect similar to incorporating nitrogen into III-V materials. It has been shown that the electronic structure of these alloys can be well described by that BAC model.

-12-8-4 0 4 8 12 0 0.04 0.08 6 1 k (10 cm ) DOS (1/eV-unit cell) Fig. 9. The calculated energy band structure (left panel) and density of states (right panel) for Zno.88Mn0.i20J:Tei^ with x~0.01. The three possible optical transitions are indicated in the left panel.

New Developments in Dilute Nitride Semiconductor Research

423

The O-induced modification of the conduction band structure of IIVI compound semiconductors offers an interesting possibility of using small amounts of O to engineer the optoelectronic properties of group IIO-VI alloys. One important technological potential of dilute oxides is for photovoltaic solar energy conversion. Efforts to improve the efficiency of solar cells have led to extensive experimental and theoretical studies of new materials and cell designs. To date, the highest power conversion efficiency of -37% has been achieved with multi-junction solar cells based on standard semiconductor materials. ' ' It was recognized over thirty years ago that the introduction of states in a semiconductor band gap presents an alternative to multi-junction designs for improving the power conversion efficiency of solar cells.89"91 It was argued that deep impurity or defect states could play the role of the intermediate states for this purpose. Detailed theoretical calculations indicate that a single junction cell with one or two properly located bands of intermediate states could achieve power conversion efficiencies up to 62% (Ref.90) and 71.7% (Ref.91), respectively. However, difficulties in controlling the incorporation of high concentrations of impurity or defect states have hindered prior efforts to realize such materials. It has been recently found that, with the multiple band gaps that fall within the solar energy spectrum, Zni_yMnyOxTei.x provides a unique opportunity for the realization of the proposed multiband solar cell.92'93 The energy band structure and the density of states for the case of Zn0.88Mn0.i2OxTei_x alloy (with x~0.03) are shown in Fig. 9. An O derived narrow band of extended states E_ is separated from the upper subband E+ by about 0.7 eV. Three types of optical transitions are possible in this band structure: (1) from the valence band to the E+ subband, Ev+=E+(k=0)-Ev(k=0)=2.56 eV, (2) from the valence band to E_ subband, Ew_=E_(k=0)-Ev(k=0)=l.83 eV, and (3) from E_ to E+, E+_=E+(k=0)-E_(k=0)=0J3 eV. These three absorption edges span much of the solar spectrum, indicating that dilute oxide II-O-VI alloys could be good candidates for the multi-band semiconductors envisioned for high efficiency photovoltaic devices. Solar cells made from such a

424

Shan, Walukiewicz, Yu, Wu, Ager, Haller

three-band material have a number of advantages over multi-junction solar cells that include much simpler device structure and much lower cost, as well as higher solar energy conversion efficiency. Acknowledgments This work was supported by the Director, Office of Science, Office of Basic Energy Sciences, Division of Materials Sciences and Engineering, of the U.S. Department of Energy under Contract No. DE-AC0376SF00098. References 1. M. Weyers, M. Sato, H. Ando, Jpn. J. Appl. Phys. 31, L853 (1992). 2. M. Kondow, K. Uomi, K. Hosomi and T. Mozume, Jpn. J. Appl. Phys. 33, LI 056 (1994). 3. C. Skierbiszewski, P. Perlin, P. Wisniewski, W. Knap, T. Suski, W. Walukiewicz, W. Shan, K.M. Yu, J.W. Ager, E.E. Haller, J.F. Geisz, and J.M. Olson, Appl. Phys. Lett. 76, 2409(2000). 4. J.F. Geisz, D.J. Friedman, J.M. Olson, S.R. Kurtz, and M.B. Keyes, J. Cryst. Growth, 195,401(1998). 5. S.R. Kurtz, Allerman, C.H. Seager, R.M. Sieg, and E.D. Jones, Appl. Phys. Lett. 77, 400(2000). 6. W. Shan, W. Walukiewicz, J. W. Ager III, E. E. Haller, J. F. Geisz, D. J. Friedman, J. M. Olson, and S. R. Kurtz, Phys. Rev. Lett. 82, 1221(1999). 7. J. D. Perkins, A. Masceranhas, Y. Zhang, J. F. Geisz, D. J. Friedman, J. M. Olson, and S. R. Kurtz, Phys. Rev. Lett. 82, 3312(1999). 8. K. Uesugi, N. Marooka, and I. Suemune, Appl. Phys. Lett. 74, 1254(1999). 9. J.A. Van Vechten, T.K.Bergstresser, Phys. Rev. 51, 3351 (1970). 10. D. Richardson, J. Phys. C: Solid State Phys. 4, L289 (1971). 11. H.C. Casey, M.B. Panish, J. Appl. Phys. 40,4910 (1969). 12. J. N. Baillargeon, K. Y. Cheng, G. E. Hofler, P. J. Pearah and K. C. Hsieh, Appl. Phys. Lett. 60, 2540(1992). 13. M. Kondow, K. Uomi, K. Hosomi, and T. Mozume, Jpn. J. Appl. Phys. Part 2, 33, L1056 (1994). 14. W.G. Bi, and C.W. Tu, J. Appl. Phys. 80, 1934 (1996). 15. M. Kondow, T. Kitatani, S. Nakatsuka, M.C. Larson, K. Nakahara, Y. Yazawa, M. Okai, K. Uomi, IEEE J. Sel. Topic in Quantum Electronics, 3, 719 (1997).

New Developments in Dilute Nitride Semiconductor Research

425

16. See, for example, Dilute Nitride (III-N-V) Semiconductors: Physics and Technology, ed. M. Henini, Elsevier, London, 2004, Chpts. 1-4. 17. See, for example, J.S. Harris Jr, Semicond. Sci. Technol, 17, 880 (2002), and references there in. 18. M. Kondow, K. Uomi, A. Niwa, T. Kitatani, S. Watahiki, and Y. Yazawa, Jpn. J. Appl. Phys. 35, 1273 (1996). 19. T. Miyamoto, T. Kageyama, S. Makino, D. Schlenker, F. Koyama and K. Iga, J. Cryst. Growth, 209, 339 (2000). 20. W. Shan, K.M. Yu, W. Walukiewicz, J.W. Ager III, E.E. Haller, M.C. Ridgway, Appl. Phys. Lett. 75, 1410(1999). 21. K.M. Yu, W. Walukiewicz, J. Wu, J. Beeman, J.W. Ager III, E.E. Haller, W. Shan, H.P. Xin, C.W. Tu, M.C. Ridgway. Appl. Phys. Lett. 78, 1077(2001); J. Appl. Phys. 90, 2227(2001). 22. J. Jasinski, K.M. Yu, W. Walukiewicz, Z. Liliental-Weber, J. Washburn, Appl. Phys. Lett. 79,931(2001). 23. K.M. Yu, W. Walukiewicz, M.A. Scarpulla, O.D. Dubon, J. Jasinski, Z. LilientalWeber, J. Wu, J. Beeman, M.R. Pillai, M.J. Aziz, J. Appl. Phys. 94, 1043(2003). 24. K.M. Yu, W. Walukiewicz, J. Wu, W. Shan, J. Beeman, M.A. Scarpulla, O.D. Dubon, M.C. Ridgway, D.E. Mars, D.R. Chamberlin, Appl. Phys. Lett. 83, 2844(2003). 25. C.W. White, P.S. Percy, Laser and Electron Beam Processing of Materials, Academic Press, (New York 1980). 26. J.S. Williams, Laser Annealing of Semiconductors. Ed. J.M. Poate and J.M. Mayer, Academic Press, (New York, 1982), p.385. 27. J.M. Olson, S.R. Kurtz, A.E. Kibbler, and P. Faine, Appl. Phys. Lett. 56, 623 (1990). 28. R.R. King, et al. Proc. 28th IEEE Photovoltaic Specialists Conference, (New York: IEEE, 2000)p 998. 29. H.L. Cotal et al. Proc. 28th IEEE Photovoltaic Specialists Conference, (New York: IEEE, 2000) p 955. 30. S.R. Kurtz, D. Myers, and J.M. Olson, Proc. 26th IEEE Photovoltaic Specialists Conference, (New York: IEEE, 1997) p 875. 31. D.J. Friedman, J.F. Geisz, S.R. Kurtz, and J.M. Olson, J. Cryst. Growth, 195, 409 (1998). 32. H.Q. Hou, K.C. Reinhardt, S.R. Kurtz, J.M. Gee, A.A. Allerman, B.E. Hammons, P.C. Chang, and E.D. Jones, Proc. 2nd World Conf. on Photovoltaic Energy Conversion (Piscataway, NJ: IEEE, 1998) p.3600. 33. D.J. Friedman, J.F. Geisz, S.R. Kurtz, and J.M. Olson, Proc. 2nd World Conf. on Photovoltaic Energy Conversion, (Piscataway, NJ: IEEE, 1998) p.3. 34. S.R. Kurtz, A.A. Allerman, E.D. Jones, J.M. Gee, J.J. Banas, and B.E. Hammons, Appl. Phys. Lett. 74, 729 (1999). 35. S.R. Kurtz, A.A. Allerman, C.H. Seager, R.M. Sieg, and E.D. Jones, Appl. Phys. Lett. 77, 400 (2000).

426

Shan, Walukiewicz, Yu, Wu, Ager, Haller

36. S.R. Kurtz, J.F. Geisz, D.J. Friedman, J.M. Olson, A. Duda, N.H. Karam, R.R. King, J.H. Ermer, and D.E. Joslin, Proc. 28th IEEE Photovoltaic Specialists Conference, (New York: IEEE, 2000) p 1210. 37. S.R. Kurtz, J.F. Klem, A.A. Allerman, R.M. Sieg, C.H. Seager, and E.D. Jones, Appl. Phys. Lett. 80, 1379 (2002). 38. J.C. Phillips, Bonds and Bands in Semiconductors, Academic, New York, 1973, p.54. 39. S. Sakai, Y. Ueta, and Y. Terauchi, Jpn. J. Appl. Phys. Part 1, 32, 4413 (1993). 40. G. Pozina, I. Ivanov, B. Monemar, J. V. Thordson, and T. G. Andersson, J. Appl. Phys. 84, 3830 (1998). 41. J. Sik, M. Schubert, G. Leibiger, V. Gottschalch, G. Kirpal, and J. Humlicek, Appl. Phys. Lett. 76, 2859 (2000). 42. A. Ougazzaden, Y. Le Bellego, E. V. K. Rao, M. Juhel, L. Leprince, and G. Patriarche, Appl. Phys. Lett. 70, 2861 (1997). 43. T. Makimoto, H. Saito, T. Nishida, and N. Kobayashi, Appl. Phys. Lett. 70, 2984 (1997). 44. M. Kozhevnikov, V. Narayanamurti, C. V. Reddy, H. P. Xin, C. W. Tu, A. Mascarenhas, and Y. Zhang, Phys. Rev. B 61, R7861 (2000). 45. S.-H. Wei and A. Zunger, Phys. Rev. Lett. 76, 664 (1996). 46. J. Neugebauer and C. G. Van de Walle, Phys. Rev. B 51,10568 (1995). 47. A. Rubio and M. L. Cohen, Phys. Rev. B 51, 4343 (1995). 48. L. Bellaiche, S.-H. Wei, and A. Zunger, Phys. Rev. B 54, 17568 (1996). 49. L. Bellaiche, S.-H. Wei, and A. Zunger, Appl. Phys. Lett. 70, 3558 (1997). 50. T. Mattila, S.-H. Wei, and A. Zunger, Phys. Rev. B 60, Rl 1245 (1999). 51. H.P. Hjalmarson, P. Vogl, D.J. Wolford, J.D. Dow, Phys. Rev. Lett. 44, 810 (1980). 52. D.J. Wolford, J.A. Bradley, K. Fry, J. Thompson, in Physics of Semicondutors, ed. J.D. Chadi and W.A. Harrison (Springer, New York, 1984), p.627. 53. X. Liu, S.G. Bishop, J.N. Baillargeon, and K.Y. Cheng, Appl. Phys. Lett. 63, 208 (1993). 54. Y. Makita, H. Ijuin, and S. Gonda, Appl. Phys. Lett. 28, 287 (1976). 55. E. D. Jones, N. A. Modine, A. A. Allerman, S. R. Kurtz, A. F. Wright, S. T. Tozer, and X. Wei, Phys. Rev. B 60, 4430 (1999). 56. W. Walukiewicz, W. Shan, J. W. Ager III, D. R. Chamberlin, E. E. Haller, J. F. Geisz, D. J. Friedman, J. M. Olson, and S. R. Kurtz, in Photovoltaics for the 21s' Century, ed. V.K. Kapur, R.D. McDonnell, D. Carlson, G.P. Ceasar, and A. Rohatgi, (The Electrochemical Society, Pennington, NJ, 1999), p.199. 57. J. Endicott, A. Patane, J. Ibanez, L. Eaves, and M. Bissiri, M. Hopkinson, R. Airey, and G. Hill, Phys Rev. Lett. 91, 126802 (2003). 58. A. Lindsay and E. P. O'Reilly, Solid State Commun. 112, 443 (1999). 59. E. P. O'Reilly and A. Lindsay, Phys. Status Solidi (b), 216, 131 (1999). 60. E.P. O'Reilly, A. Lindsay, S. Tomic, and M. Kamal-Saadi, Semicond. Sci. Technol. 17, 870(2002).

New Developments in Dilute Nitride Semiconductor Research

All

61. S. A. Choulis, S. Tomic, E. P. O'Reilly, and T. J. C. Hosea, Solid State Commun. 125, 155 (2003). 62. S. A. Choulis, T. J. C. Hosea, S. Tomic, M. Kamal-Saadi, A. R. Adams, E. P. O'Reilly, B. A. Weinstein, and P. J. Klar, Phys. Rev. B 66, 165321 (2002). 63. B. Gil, Solid State Commun. 114, 623 (2000). 64. N. Shtinkov, P. Desjardins, and R. A. Masut, Phys. Rev. B 67, 081202 (2003). 65. P.J. Klar, H. Gruning, J. Koch, S. Schafer, K. Volz, W. Stolz, W. Heimbrodt, A.M. Kamal-Saadi, A. Lindsay, and E.P. O'Reilly, Phys. Rev. B 64, 121203 (2001). 66. P.J. Klar, H. Gruning, W. Heimbrodt, G. Weiser, J. Koch, S. Schafer, K. Volz, W. Stolz, S.W. Koch, S. Tomic, S.A. Chouli, T.J.C. Hosea, E.P. O'Reilly, M. Hofmann, J. Hader, and J.V. Moloney, Semicond. Sci. Technol, 17, 830 (2002). 67. P. R. C. Kent and A. Zunger, Phys. Rev. Lett. 86, 2613 (2001); Phys. Rev. B 64, 115208(2001). 68. G. Szwacki and P. Boguslawski, Phys. Rev. B 64, 161201 (2001). 69. I. Vurgaftman and J.R. Meyer, J. Appl. Phys. 94, 3675 (2003). 70. A. Lindsay, S. Tomic, and E. P. O'Reilly, Solid-State Electron. 47, 443 (2003). 71. R. Bhat, C. Caneau, L. Salamanca-Riba, W.G. Bi, C.W. Tu, /. Crystl. Growth, 195, 427 (1998); 72. L. Malikova, F.H. Pollak, R. Bhat, J. Electron. Mat. 27, 484 (1998); 73. B.M. Keyes, J.F. Geisz, P.C. Dippo, R. Reedy, C. Kramer, D.J. Friedman, S.R. Kurtz, J.M. Olson, AIP Conf. Proc. 462, 511 (1999). 74. W. Shan, W. Walukiewicz, J.W. Ager III, E.E. Haller, J.F. Geisz, D.J. Friedman, J.M. Olson, and S.R. Kurtz, J. Appl. Phys. 86, 2349 (1999). 75. P.J. Klar, H. Gruning, W. Heimbrodt, J. Koch, F. Hohnsdorf, W. Stolz, P.M.A. Vicente, and J. Camassel, Appl. Phys. Lett. 76, 3439 (2000). 76. C Skierbiszewski, Semicond. Sci. Technol. 17, 803 (2002). 77. W. Walukiewicz, Appl. Phys. Lett. 54, 2094 (1989); Mat. Res. Soc. Symp. Proc. 300, 421 (1993). 78. S.B. Zhang, S.H. Wei, A. Zunger. J. Appl. Phys. 83, 3192 (1998). 79. W. Walukiewicz, Mats. Science Forum, Vols.143-147, 519 (1993). 80. K.M. Yu, W. Walukiewicz, W. Shan, J.W. Ager III, J. Wu, E.E. Haller, J.F. Geisz, D.J. Friedman, J.M. Olson. Phys. Rev. B 61, R13337 (2000). 81. J. Wu, W. Shan, W. Walukiewicz, Semicond. Sci. Technol. 17, 860 (2002). 82. K.M. Yu, W. Walukiewicz, W. Shan, J. Wu, J.W. Ager III, E.E. Haller, J.F. Geisz, M.C. Ridgway, Appl. Phys. Lett. 11, 2858 (2000). 83. K.M. Yu, W. Walukiewicz, W. Shan, J. Wu, J. Beeman, J.W. Ager III, E.E. Haller. Appl. Phys. Lett. 17, 3607 (2000). 84. C. Skierbiszewski, P. Perlin, P. Wisniewski, W. Knap, T. Suski, W. Walukiewicz, W. Shan, K.M. Yu, J.W. Ager III, E.E. Haller, J.F. Geisz, J.M. Olson. Appl. Phys. Lett. 76, 2409 (2000). 85. K.M.Yu, W.Walukiewicz, J.Wu, J.W. Beeman, J.W. Ager, E. E. Haller, I. Miotkowski, A. K. Ramdas, and P. Becla, Appl. Phys. Lett. 80, 1571 (2002).

428

Shan, Walukiewicz, Yu, Wu, Ager, Haller

86. W. Shan, W. Walukiewicz, J.W. Ager III, K.M. Yu, J. Wu, E.E. Haller, Y. Nabetani, T. Mukawa, Y. Ito, T. Matsumoto, Appl. Phys. Lett. 83, 299 (2003). 87. P.K. Chiang, J.H. Ermer, W.T. Nishikawa, D.D. Krut, D.E. Joslin, J.W. Eldredge, B.T. Cavicchi, and J.M. Olson, Proc. 25th IEEE Photovoltaic Specialists Conf. (IEEE New York, 1996) p. 183. 88. R.R. King, P.C. Colter, D.E. Joslin, K.M. Edmondson, D.D. Krut, N.H. Karam, and Sarah Kurtz, Proc. 29th IEEE Photovoltaic Specialists Conf. New Orleans, (IEEE, New York, 2002) p. 852. 89. M. Wolf, Proc. IRE, 48, 1246(1960). 90. A. Luque and A. Marti, Phys. Rev. Lett. 78, 5014 (1997). 91. A. S. Brown, M. A. Green and R. P. Corkish, Physica, E14, 121 (2002). 92. K. M. Yu, W. Walukiewicz, J. Wu, W. Shan, J. W. Beeman, M. A. Scarpulla, O. D. Dubon, and P. Becla, Phys. Rev. Lett. 91, 246203 (2003). 93. W. Shan, W. Walukiewicz, K.M. Yu, J.W. Ager III, J. Wu, J. Beeman, M.A. Scapulla, O.D. Dubon, E. E. Haller, Y. Nabetani, and P. Becla, SPIE Proc. Vol.5349,426 (2004).

Ill-Nitride Semiconductor Materials excellent wide band gap semiconductors very suitable for modern electronic and optoelectronic applications. Remarkable breakthroughs have been achieved recently, and current knowledge and data published have to be modified and upgraded. This book presents the new developments and achievements in the field. Written by renowned experts, the review chapters in this book cover the most important topics and achievements in recent years, discuss progress made by different groups, and suggest future directions. Each chapter also describes the basis of theory or experiment. The lll-Nitride-based industry is building up and new economic

developments from these

materials

are

promising. It is expected that lll-Nitride-based LEDs may replace traditional light bulbs to realize a revolution in lighting. This book is a valuable source of information for engineers, scientists and students working towards such goals. Key Features •

Provides useful information to material growers and evaluators, device design and processing engineers

•

Covers recent developments in Ill-Nitride semiconductors materials

•

Predicts future directions for further research and development

Imperial College Press www.icpress.co.uk

9 "781860 946363"

Handbook of Luminescent Semiconductor Materials

Read more

Photovoltaic Materials (Series on Properties of Semiconductor Materials Vol 1)

Read more

Semiconductor Packaging: Materials Interaction and Reliability

Read more

Semiconductor Materials and Process Technology Handbook

Read more

Semiconductor Materials and Process Technology Handbook

Read more

Semiconductor-On-Insulator Materials for Nanoelectronics Applications

Read more

Semiconductor Lasers II: Materials and Structures

Read more

Materials for High-Temperature Semiconductor Devices

Read more

Semiconductor Nanostructures for Optoelectronic Applications (Artech House Semiconductor Materials and Devices Library)

Read more

Semiconductor-Laser Fundamentals: Physics of the Gain Materials

Read more

Semiconductor optics

Read more

Semiconductor macroatoms

Read more

Semiconductor Nanostructures

Read more

Advances in Metal and Semiconductor Clusters (Vol. 4): Cluster Materials

Read more

Semiconductor-Laser Fundamentals : Physics of the Gain Materials

Read more

Semiconductor Materials: An Introduction to Basic Principles (Microdevices)

Read more

Semiconductor Materials: An Introduction to Basic Principles (Microdevices)

Read more

Semiconductor Optics

Read more

Semiconductor equations

Read more

Semiconductor Optics

Read more

Semiconductor Electrochemistry

Read more

Semiconductor Fundamentals

Read more

Semiconductor lasers

Read more

Semiconductor Detector Systems (Semiconductor Science and Technology)

Read more

Semiconductor Memories

Read more

Semiconductor Nanomaterials

Read more

Semiconductor optics

Read more

Semiconductor Electronics

Read more

Semiconductor Junctions

Read more

Semiconductor Device

Read more

Recommend Documents

Handbook of Luminescent Semiconductor Materials

Handbook of LUMINESCENT SEMICONDUCTOR MATERIALS Ed i t e d by Leah Bergman Je a n n e L . M c Ha l e Handbook of L...

Photovoltaic Materials (Series on Properties of Semiconductor Materials Vol 1)

Photovoltaic Materials SERIES ON PROPERTIES OF SEMICONDUCTOR MATERIALS Series Editor: Ron C. Newman Forthcoming Sem...

Semiconductor Packaging: Materials Interaction and Reliability

Semiconductor Packaging Andrea Chen Randy Hsiao-Yu Lo Semiconductor Packaging Materials Interaction and Reliability ...

Semiconductor Materials and Process Technology Handbook

Semiconductor Materials and Process Technology Handbook

Semiconductor-On-Insulator Materials for Nanoelectronics Applications

Engineering Materials For further volumes: http://www.springer.com/series/4288 Alexei Nazarov J.-P. Colinge Prof. Fr...

Semiconductor Lasers II: Materials and Structures

Semiconductor L a s e r s II Materials and Structures OPTICS AND PHOTONICS (formerly Quantum Electronics) SERIES EDIT...

Materials for High-Temperature Semiconductor Devices

...

Semiconductor Nanostructures for Optoelectronic Applications (Artech House Semiconductor Materials and Devices Library)

Semiconductor Nanostructures for Optoelectronic Applications For a listing of recent titles in the Artech House Semic...

Semiconductor-Laser Fundamentals: Physics of the Gain Materials

...